I have brokered health insurance for over a hundred startups. At least half the founders ask the same question within the first ten minutes: “Should we do an HSA?”

They’ve read the blog posts. They’ve seen the TikToks. Triple tax advantage. Tax-free in, tax-free growth, tax-free out. The only account in the American tax code that lets you skip taxes at every stage. Better than a 401(k). Better than a Roth IRA.

They’re not wrong. But they are missing something important.

To open an HSA, you need a high-deductible health plan. And high-deductible health plans make people sicker.

A Brief History of Making People Pay

The intellectual case for HSAs was built on a single finding from the RAND Health Insurance Experiment, a landmark study from 1974 to 1982. RAND found that when people paid more out of pocket, they spent about 30% less on healthcare. Congress loved this. What they cited less often was the other finding: cost sharing reduced effective and ineffective care in roughly equal proportions. People didn’t become smarter consumers. They just avoided the doctor.

The deductible is a blunt instrument. It does not distinguish between a frivolous visit and a necessary one. It only distinguishes between people who can afford to pay and people who cannot.

Congress created Health Savings Accounts on December 8, 2003, embedding them inside a Medicare prescription drug bill. The deal was simple: accept a high deductible, get a tax-advantaged savings account. The theory was consumer-directed healthcare → give patients “skin in the game” and they’ll shop around, skip the unnecessary MRI, think twice before visiting the ER.

Twenty-two years later, 39 million Americans hold HSAs with $147 billion in total assets.

The Triple Tax Advantage, Explained

The mechanics are elegant. Contribute pre-tax dollars, up to $4,400 individual or $8,750 family in 2026. If routed through a Section 125 cafeteria plan, those dollars dodge federal income tax, state income tax, and FICA payroll taxes. That last part is the edge over a traditional IRA: you skip the 7.65% payroll tax too.

The money grows tax-free. Invest in index funds, ETFs, bonds — no taxes on gains as long as it stays in the account. Withdraw for medical expenses at any age, pay zero tax. After 65, it functions like a traditional IRA for non-medical withdrawals. No required minimum distributions, ever.

The catch is the entry ticket.

The Price of Admission

To contribute to an HSA, you must be on a qualified high-deductible health plan. In 2026: minimum $1,700 deductible for individuals, $3,400 for families. No benefits can be paid before the deductible except preventive care and telehealth.

The IRS enforces this with precision. I’ve had founders ask: what if we pair an HDHP with an FSA and use FSA dollars to cover the deductible, effectively zeroing it out? The IRS anticipated this. A general-purpose FSA counts as “other health coverage.” The moment an employee has access to one, HSA eligibility dies. It doesn’t matter whether they use it. Eligibility alone kills it.

You can pair an HDHP with a limited-purpose FSA for dental and vision. You can fund your employees’ HSAs on day one, effectively covering the deductible through the account itself. But between the medical visit and the reimbursement, there’s friction. The employee thinks about the out-of-pocket bill. They hesitate.

That hesitation is the feature, not the bug.

What Hesitation Actually Costs

A 2023 JAMA Network Open study followed diabetic patients whose employers forced them onto HDHPs. Emergency visits and hospitalizations for severe hyperglycemia rose 25%. Each additional year of enrollment increased the risk by another 5%.

Harvard researchers linked HDHPs with a 4.6-month delay in detecting metastatic cancer. In oncology, that is not a rounding error.

Women switched from low-deductible plans to HDHPs experienced delays of nearly seven months to breast cancer diagnosis, across the income spectrum. The researchers expected this for low-income women. They did not expect it for high-income ones. Cost sharing changes behavior regardless of ability to pay.

A paper published in JAMA Network Open just weeks ago found cancer survivors on HDHPs had a mortality hazard ratio of nearly 1.5 compared to traditional plans. The association held after controlling for income, education, and race.

The RAND experiment told us in 1982 that cost sharing cuts needed and unneeded care equally. Forty-four years of research has confirmed it. The deductible does not make people better healthcare consumers. It makes them avoidant ones.

The Payroll Tax You Didn’t Budget For

There’s a second cost no one mentions at the seed stage: operational complexity.

To give employees the full FICA exemption on HSA contributions, you need a Section 125 cafeteria plan: a written IRS document outlining benefits, eligibility, contribution limits, and election procedures. You need a third-party administrator ($2–8 per employee per month). You need annual nondiscrimination testing, which is a particular headache for startups where the founders are the highest-paid employees. You need to file Form 5500 with the Department of Labor. Report employer HSA contributions on W-2s using Box 12 Code W. Integrate a separate HSA custodian with your payroll system.

Skip the Section 125 plan and contribute directly? You’re subject to IRS comparability rules: identical amounts for every employee, with a 35% excise tax if you mess it up.

For a ten-person startup, this overhead is disproportionate. You’re building compliance infrastructure to support a tax shelter that requires putting your team on insurance designed to make them think twice about seeing a doctor.

The Question You Should Actually Ask

The question isn’t “should we offer an HSA?” It’s: what problem are you solving?

If the answer is “minimize my employees’ tax burden”, the HSA is attractive in theory but comes packaged with insurance that discourages utilization. You’re optimizing for tax efficiency while degrading the actual benefit.

If the answer is “I want my team to get care without worrying about cost”, a zero-deductible plan achieves this directly. No tax shelter required. No plan document. No custodian. No nondiscrimination testing.

Congress is currently trying to decouple HSAs from high-deductible plans entirely. If that passes, the calculus changes. Until then, you’re trading your team’s health access for a tax bracket optimization.

For a founder who treats health benefits as a recruiting tool and not a compliance checkbox, that’s a bad trade.

Most Americans cannot explain how their health insurance works. This is not because they are unintelligent. It is because the system was designed by economists, not patients.

Three mechanisms control what you pay when you get sick: the deductible, coinsurance, and the out-of-pocket maximum. They sound like consumer protections. Their origins tell a different story.

How the three mechanisms work

The deductible is the amount you pay before your insurance covers anything. If your plan has a $2,000 deductible, you pay the first $2,000 yourself. Your plan does not contribute a dollar until you cross that line. Preventive care is exempt under federal law, but everything else counts: the urgent care visit for a sinus infection, the blood work your doctor ordered, the X-ray after you twisted your ankle.

Coinsurance kicks in after you meet your deductible. It is a percentage split. A typical arrangement is 80/20: the plan pays 80%, you pay 20%. An MRI that costs $1,200 means you owe $240.

The out-of-pocket maximum is the ceiling on what you pay in a plan year. Once your deductible and coinsurance payments add up to this number, the plan covers 100%. The federal cap for 2025 is $9,200 for an individual. For 2026, it rises to $10,600.

These three mechanisms stack in sequence. To understand what that feels like, consider two people on different plans facing the same year.

Sarah and James both break an arm

Sarah is on a typical PPO: $2,000 deductible, 80/20 coinsurance, $6,500 out-of-pocket max. James is on a high-deductible plan (HDHP): $3,500 deductible, 80/20 coinsurance, $7,500 out-of-pocket max.

In March, both slip on ice and break an arm. The ER visit, X-ray, cast, and follow-ups total $4,500.

Sarah pays $2,000 (deductible) plus 20% of the remaining $2,500 ($500 in coinsurance). Total: $2,500. James pays $3,500 (deductible) plus 20% of the remaining $1,000 ($200). Total: $3,700.

Neither has hit their out-of-pocket max. Both still have exposure for the rest of the year.

In September, both need outpatient knee surgery. Negotiated rate: $15,000. Sarah owes 20% coinsurance ($3,000), bringing her year-to-date total to $5,500. James owes the same $3,000, bringing his to $6,700. The out-of-pocket max has not kicked in for either of them, and the median American household has less than $5,000 in savings.

Now consider David. Same HDHP as James. He has a persistent cough in January. The doctor visit would cost $250. He has not met any of his $3,500 deductible. He decides to wait it out.

The cough was early-stage pneumonia. By February, David is in the emergency room. The bill is $8,000.

David’s plan worked exactly as designed. It made him hesitate to get the early care he needed.

A system designed against you

These mechanisms were not born from medical evidence. They were borrowed from car insurance.

In 1949, Liberty Mutual created the first “major medical” health insurance policy for a group of General Electric engineers. It had a $300 deductible and 25% coinsurance, modeled directly on automobile collision coverage. The logic: a deductible discourages frivolous claims. By 1961, 34 million Americans were enrolled in plans built this way.

The intellectual justification arrived through three economics papers in the 1960s and ‘70s. Kenneth Arrow identified moral hazard: when someone else pays, patients consume more. Mark Pauly argued this overconsumption was rational, and that full coverage was therefore economically suboptimal. Martin Feldstein claimed that raising coinsurance rates would save billions. Together they built a framework that treats healthcare like any other commodity. Give patients skin in the game and they will consume only what they truly need.

The federal government spent $295 million (in today’s dollars) testing this theory. The RAND Health Insurance Experiment ran from 1974 to 1982, randomly assigning over 7,700 people to plans with varying cost-sharing. The headline finding: people on free plans spent 45% more than those paying 95% coinsurance.

The finding underneath the headline was devastating.

Cost-sharing reduced appropriate and inappropriate care in equal proportions. Patients did not become smarter consumers. They did not learn to distinguish a necessary visit from an unnecessary one. They went to the doctor less. Period.

For the poorest and sickest 6%, free care produced meaningfully better outcomes: lower blood pressure, better vision, an estimated 10% reduction in mortality for hypertensive low-income patients.

Every major study since has confirmed this pattern. A 2017 study found that switching employees to a high-deductible plan reduced spending 12 to 14%, entirely through people using less care. Zero evidence of price shopping. Even after two years. Consumers did not learn to find cheaper providers. They learned to avoid providers entirely.

The most damning evidence concerns primary care. High-deductible plans reduce preventive visits even though federal law makes them free. Nearly 20% of HDHP enrollees do not know their preventive care costs nothing. One study found that while HDHP enrollees initially had fewer doctor visits, their hospital admissions doubled three years later.

Doubling copayments for diabetes medications reduced adherence by 23%. For blood pressure medications, 10%. These are not optional treatments. These are drugs that prevent heart attacks.

The economist John Nyman put it simply: conventional moral hazard theory misunderstands why people buy insurance. People do not just buy risk protection. They buy access to care they could not otherwise afford. Much of what the industry calls “overutilization” is patients getting care that improves their health.

Seven decades of evidence point to one conclusion. Cost-sharing at the point of primary and urgent care does not make patients smarter consumers. It makes them sicker.

Why Rivendell plans are different

We designed Rivendell around one principle: the moment a member needs care should never be the moment they face a financial decision.

Every Rivendell plan has a zero-dollar deductible. Primary care is free. Urgent care is free.

This is not generosity. It is plan design that follows the evidence instead of ignoring it.

When you put a price on a doctor visit, people skip it. The ones who skip are disproportionately the sickest, lowest-income, and least able to absorb a surprise bill. The $250 they saved becomes the $8,000 ER visit. The skipped blood pressure check becomes the hospitalization.

Primary care and urgent care are where problems are small and cheap to solve. Every dollar of friction at that level generates multiples in downstream cost. Removing that friction is not just better for members. It is better economics.

The insurance industry built cost-sharing to solve a theoretical problem called moral hazard.

We built Rivendell to solve the actual problem: people who need care and do not get it.

For healthtech founders and investors, these shifts create real opportunities. Here are five predictions for where the market is heading.

1. Consumer-directed healthcare accelerates through HSA expansion and price transparency

The One Big Beautiful Bill Act, signed in July 2025, delivered the largest expansion of Health Savings Accounts in two decades. Starting January 2026, any Bronze or Catastrophic plan on an ACA exchange automatically qualifies for HSA eligibility. Direct Primary Care memberships up to $150 per month now count as qualified HSA expenses. Telehealth access before meeting deductibles is permanently allowed.

This matters because it changes the math for millions of consumers. Bronze plans have always offered lower premiums but higher out-of-pocket costs. Now those same consumers can pair cheap premiums with tax-advantaged savings accounts. Morningstar estimates 3 to 4 million additional Americans will become HSA-eligible in 2026 alone.

At the same time, CMS finalized stricter hospital price transparency rules. Starting April 2026, hospitals must publish actual negotiated rates, not estimates. They must disclose median, 10th percentile, and 90th percentile allowed amounts for services. Hospital CEOs must personally attest to data accuracy.

The combination creates a new consumer segment: people with high-deductible plans, HSA dollars to spend, and actual price information to shop with. Startups building price comparison tools, cash-pay marketplaces, and DPC networks should find a growing addressable market. The winners will be those who help consumers navigate this complexity rather than adding to it.

2. Peptide enforcement scales up

The regulatory question around peptides like BPC-157 is not whether they will be regulated. They already are. The FDA placed BPC-157 on its Category 2 list of bulk drug substances in 2023, citing significant safety concerns and lack of human clinical data. This classification prohibits licensed compounding pharmacies from producing it for human use.

The real question is enforcement. The gray market persists because sellers use “research only” labels to sidestep FDA jurisdiction. Everyone knows consumers are injecting these compounds. The legal fiction protects no one except the sellers.

Enforcement is now ramping. The Department of Justice prosecuted Tailor Made Compounding for distributing unapproved peptides, forcing a $1.79 million forfeiture. State medical boards in Ohio and Florida have suspended licenses simply for storing research-labeled vials in clinic refrigerators.

In 2026, expect continued pressure on the supply chain. Compounding pharmacies face clear prohibition. Clinics prescribing or recommending peptides face malpractice exposure and licensing risk. The wellness industry’s peptide boom is running on borrowed time.

For founders, this creates two opportunities. First, legitimate peptide research pipelines that pursue actual FDA approval. Second, compliance and credentialing tools for clinics navigating the new enforcement environment.

3. PBM transparency mandates force business model changes

The pharmacy benefit manager industry will not be broken up in 2026. The three largest PBMs still control over 80% of the market. CVS Health owns Aetna and Caremark. UnitedHealth owns Optum and OptumRx. Cigna owns Express Scripts. These vertical integrations are not going anywhere soon.

What is changing is how PBMs can operate within those structures.

Arkansas passed the first state law banning PBM ownership of pharmacies, though a preliminary injunction has paused enforcement. Colorado and California enacted “delinking” laws that prohibit tying PBM compensation to drug prices, requiring flat-fee models instead. Massachusetts imposed comprehensive licensing and transparency requirements. Utah mandated rebate pass-through to consumers at point of sale.

At the federal level, the House passed legislation in December 2025 requiring PBMs to report detailed prescription drug spending data to plan sponsors. The FTC continues investigating PBM practices around spread pricing and rebate opacity.

Meanwhile, transparent PBMs are proving the alternative model works. SmithRx reports 30% average cost reductions compared to legacy PBMs through its pass-through pricing model. Mark Cuban’s Cost Plus Drugs continues expanding. Employers are increasingly willing to switch.

The opportunity here is not in trying to break up the incumbents. It is in serving the growing segment of employers and health plans that want out of the opaque system entirely.

4. OpenAI and Apple compete to become the healthcare front door

Over 230 million people ask health and wellness questions on ChatGPT every week. That is not a typo. OpenAI’s consumer health engagement already exceeds most telehealth platforms by orders of magnitude.

In January 2026, OpenAI launched ChatGPT Health, a dedicated space where users can connect medical records and wellness apps like Apple Health, Function, and MyFitnessPal. The platform helps users interpret lab results, prepare for doctor appointments, and navigate insurance decisions. OpenAI also launched ChatGPT for Healthcare, an enterprise product for hospitals and health systems. Early adopters include Boston Children’s Hospital, Cedars-Sinai, Stanford Medicine Children’s Health, and Memorial Sloan Kettering.

Apple is pursuing its own path. Bloomberg reports that Health+, Apple’s AI-powered health coaching service, remains on track for 2026. The service will reportedly offer personalized recommendations on nutrition, exercise, and chronic disease management based on Apple Watch data.

These two companies are not partnering. They are competing for the same position: the intelligent layer between consumers and the healthcare system. Apple has device integration and privacy positioning. OpenAI has conversational AI capabilities and massive user engagement.

For healthtech founders, this creates both threat and opportunity. The threat is obvious. If ChatGPT becomes how people navigate healthcare, many point solutions become features rather than products. The opportunity is in the infrastructure layer. Someone needs to build the integrations, data pipelines, and clinical guardrails these platforms require.

5. The payer-provider AI billing war intensifies

Health insurers had a brutal 2025. Medical loss ratios spiked. GLP-1 medications like Ozempic and Wegovy drove pharmacy costs through the roof. Medicaid unwinding pushed previously covered populations into commercial plans with higher utilization patterns. Medicare Advantage margins compressed under CMS rate adjustments.

Insurers blame providers, specifically their adoption of AI-powered coding and billing optimization tools. Centene’s CFO said it publicly: hospitals have gotten better organized around AI for coding than payers have for claims review.

The response is predictable. Cigna rolled out automated downcoding policies in October 2025, flagging physicians who bill a higher proportion of complex visits than their peers and adjusting claims downward. Other insurers are deploying their own AI to scrutinize claims, increase prior authorization requirements, and identify billing patterns to challenge.

This creates a destructive equilibrium. Providers optimize coding. Payers optimize denials. Administrative costs rise on both sides. Patients get caught in the middle.

The escape valve is value-based care. Capitated models have doubled since 2021, now representing 14% of all healthcare payments. When providers take on financial risk for total cost of care, the billing war becomes irrelevant. Both sides share incentive to reduce unnecessary utilization rather than fight over how to code it.

For founders, the near-term opportunity is in the arms race itself. Revenue cycle management tools, prior authorization automation, and denial management software all have eager buyers. The longer-term opportunity is in infrastructure that makes value-based care actually work: risk stratification, care management platforms, and analytics that help providers succeed under capitation.

The through line

These five predictions share a common theme. The healthcare system’s traditional gatekeepers are losing control. Consumers have more tools to shop and pay directly. Regulators are forcing transparency on PBMs and hospitals. Tech giants are inserting themselves between patients and providers. And the payer-provider relationship is fracturing under the weight of AI-powered adversarial billing.

For anyone in the idea maze and not knowing what to focus on, ping me and I can give you a dozen ideas that I have faced while building a vertically integrated health plan from scratch!

Let’s go build 🔥

I was wrong.

The quotes came back at nearly $1,300 per person per month for a decent plan. For a team of 10, that’s over $150,000 annually before any other employee benefits. Meanwhile, I knew founders at larger companies paying 20-30% less for equivalent coverage. Same city. Same network. Same benefits.

What I discovered was a regulatory architecture that systematically forces small businesses to subsidize a broken system, while large corporations opt out entirely.

How the Small Group Market Actually Works

If you have fewer than 100 employees in most states (50 in some), you’re in the “small group market.” This market operates under strict rules established by the Affordable Care Act:

Community Rating: Insurers cannot charge you more based on your team’s health status. Sounds fair, until you realize it means your healthy 26-year-olds are priced identically to a group with chronic conditions.

Guaranteed Issue: Insurers must accept all small groups. Again, sounds good, but it means the risk pool includes everyone who couldn’t get coverage elsewhere.

Mandated Benefits: Your plan must cover a comprehensive set of “Essential Health Benefits,” plus whatever your state mandates on top. In New York, that’s a long list.

The result? You’re not buying insurance for your team. You’re buying a share of the entire small group risk pool in your state. And in states like New York, that pool is expensive and shrinking.

New York’s “Pure Community Rating” Problem

New York goes further than federal requirements. While the ACA allows insurers to charge older employees up to 3x more than younger ones (reflecting actual actuarial risk), New York prohibits any age rating at all. A 25-year-old engineer pays the same rate as a 64-year-old.

This is called “pure community rating,” and New York adopted it in 1992—two decades before the ACA. The policy goal is social solidarity: everyone shares the burden equally.

The practical result is that young, healthy startups massively overpay relative to their actual risk, while older, sicker groups get a discount. Your Series A is subsidizing someone else’s COBRA.

The ICHRA Option: Better, But Not the Solution

When founders complain about group plan costs, someone inevitably suggests ICHRA - Individual Coverage Health Reimbursement Arrangements. The pitch: give employees a tax-free allowance, let them buy their own plans on the individual market, and avoid the small group market entirely.

ICHRA has real advantages:

• Predictable employer costs

• Employee choice and portability

• Works well for distributed teams across multiple states

But ICHRA has a fundamental limitation: you’re just shifting to a different broken market.

The individual market (healthcare.gov or state exchanges) uses the same community rating principles as the small group market. Your employees aren’t escaping the risk pool, they’re joining a different one, often with similar pricing problems and narrower networks.

ICHRA solves the administrative headache of multi-state coverage. It doesn’t solve the cost problem. In New York, individual market premiums are just as inflated as small group rates. You’re rearranging deck chairs.

Level Funding: The Alternative You’re Not Allowed to Use

Here’s what larger companies do instead: they self-insure.

A self-insured (or “self-funded”) plan works differently. Instead of paying premiums to an insurance carrier who assumes all risk, the employer pays claims directly from their own funds. The company bears the risk, but also captures the savings when claims are low.

For a Fortune 500 company with 50,000 employees, this makes obvious sense. The law of large numbers means claims are predictable, and they have the cash reserves to handle volatility.

But what about a 30-person startup? That’s where level-funded plans come in.

How Level Funding Works

A level-funded plan is a hybrid structure designed to make self-insurance accessible to smaller employers:

1. Claims Fund: You pay into an account that covers expected medical claims, based on your specific group’s health profile (yes, underwriting).

2. Stop-Loss Insurance: You buy a policy that caps your downside. If any individual has claims above $20,000 (the “specific attachment point”), the stop-loss carrier pays. If total group claims exceed 125% of expected, the “aggregate” stop-loss kicks in.

3. Administrative Fees: A third-party administrator handles claims processing, network access, and compliance.

The “level” part: you pay a fixed monthly amount, just like a fully-insured premium. But here’s the key difference, if claims come in lower than expected, you get a refund. In a traditional plan, the insurer keeps that as profit.

Why Level Funding Saves Money

The savings come from two sources:

Underwriting: Unlike community-rated plans, level-funded arrangements can price based on your actual group’s health profile. If you have a young, healthy team, your rates reflect that reality instead of subsidizing the broader pool.

Transparency and Incentives: You see exactly where money goes - claims vs. administration vs. stop-loss premium. This visibility lets you make informed decisions about plan design, wellness programs, and cost containment. You’re no longer just writing checks into a black box.

Brokers report typical savings of 15-30% compared to fully-insured renewals. For a startup spending $100K+ on health insurance, that’s real money.

The National Picture

Level-funded plans have exploded in popularity over the past decade. In states with permissive regulations: Texas, Ohio, Florida, and most of the country—small businesses have embraced them as an escape valve from ACA market costs.

The structure works. Claims data shows that level-funded groups maintain coverage, access the same provider networks, and achieve meaningful cost reductions. It’s not “junk insurance”, it’s the same self-funding strategy that large employers have used for decades, adapted for smaller scale.

Why You Can’t Do This in New York or California

If level funding is so effective, why isn’t every startup using it?

Because in New York and California, regulators have made it effectively illegal for small businesses.

The mechanism is subtle but powerful. Under federal law (ERISA), states cannot regulate self-funded health plans directly, that’s federal jurisdiction. But states can regulate insurance products sold within their borders, including stop-loss policies.

So these states weaponized stop-loss regulation to block level funding.

New York: The Outright Ban

New York Insurance Law (Sections 3231 and 4317) prohibits the sale of stop-loss insurance to any employer with fewer than 100 employees. Full stop.

No stop-loss means no level funding. Without that financial backstop, a single catastrophic claim could bankrupt a small company. The state didn’t ban self-insurance (they can’t under ERISA), they just made it economically suicidal.

The policy rationale, articulated by the Department of Financial Services, is explicit: prevent “adverse selection” that would destabilize the community-rated small group market. If healthy groups exit the pool, premiums rise for everyone remaining, potentially triggering a “death spiral.”

California: The $40,000 Toll Booth

California took a different approach with Senate Bill 161 (2013), championed by State Senator Ed Hernandez.

Instead of banning stop-loss outright, California set the minimum “attachment point” at $40,000 per individual. This means if you self-insure, you’re responsible for the first $40,000 of claims for each employee before stop-loss kicks in.

For a 10-person company, that’s $400,000 in potential exposure before your insurance activates. For most small businesses, that’s unacceptable risk. The high attachment point functions as a de facto ban, not by prohibition, but by making the economics impossible.

In Texas, by contrast, you can get stop-loss with a $10,000 or even $5,000 attachment point. The risk is manageable. The option is real.

The Philosophical Divide

The defenders of these restrictions make a coherent moral argument: health insurance should be a vehicle for social solidarity. The healthy subsidizing the sick isn’t a bug, it’s the point. If you allow risk segmentation, you abandon the most vulnerable to unaffordable coverage.

But this solidarity is selectively enforced.

Large corporations self-insure freely under ERISA, exempt from state mandates and community rating. Amazon doesn’t subsidize the New York small group pool. Neither does Goldman Sachs or Google.

The policy creates two classes: enterprises large enough to escape the regulated market, and small businesses trapped inside it. The bodega subsidizes Walmart’s healthcare savings.

The Results Are In: Restriction Hasn’t Worked

If the goal of these policies was to stabilize the small group market and protect consumers, the evidence suggests failure.

New York’s Shrinking Pool

From 2020 to 2024, the number of covered lives in New York’s small group market dropped by 24%, from 960,000 to under 740,000. Businesses aren’t staying in the pool and paying high premiums. They’re dropping coverage entirely or finding other workarounds.

The “death spiral” that regulators feared from adverse selection is happening anyway, just driven by unaffordability rather than segmentation.

Premium Explosion

New York premiums have increased dramatically despite or perhaps because of the restrictions:

• Rochester area: +45% over four years

• NYC and Long Island: +36-38% in the same period

California’s small group premiums have grown at 7% annually since 2022, outpacing national averages.

The Comparative Data

An Urban Institute analysis found New York’s average lowest gold premium was nearly double that of states with competitive markets. The “protection” of the risk pool hasn’t translated to affordable coverage. It has created some of the most expensive small group insurance in the country.

Meanwhile, in deregulated states, level-funded plans are saving small businesses 15-30% while maintaining coverage. The market that regulators feared would collapse is functioning.

The Federal Fix: What’s Actually Moving in Congress

There’s reason for cautious optimism. The Self-Insurance Protection Act (H.R. 2813 / H.R. 2571) is working through Congress with bipartisan support.

The bill would amend ERISA, the Public Health Service Act, and the Internal Revenue Code to explicitly exclude stop-loss insurance from the definition of “health insurance coverage.” This technical change would have massive practical implications:

It would preempt state laws like California’s SB 161 and New York’s stop-loss ban.

By establishing a federal definition that separates stop-loss from health insurance, the bill would strip states of their authority to set prohibitive attachment points or ban the product for small employers.

The bill has been reported out of the House Committee on Education and the Workforce. It faces opposition from the NAIC (the state insurance regulators’ association) and some Democratic leadership, who argue it would destabilize state markets.

But the affordability crisis is creating strange bedfellows. Small business associations spanning the political spectrum are unified in support. The Chamber of Commerce and NFIB are aligned. Even some traditionally ACA-supportive groups are acknowledging that current policy isn’t working for small employers.

What This Means for Founders

If you’re running a startup in New York or California, you’re currently trapped. Your options are:

1. Pay the community-rated tax: Accept that your healthy team subsidizes the broader pool, budget accordingly, and treat it as a cost of doing business in these states.

2. Use ICHRA: Shift the administrative burden to employees and access the individual market. Doesn’t solve the cost problem, but provides flexibility for distributed teams.

3. Relocate key functions: Some companies incorporate entities in permissive states to access level funding for portions of their workforce. This is complex and may not be worth it at small scale.

4. Wait and advocate: The federal legislation could change the landscape within 1-2 years. If the Self-Insurance Protection Act passes, level funding becomes available nationwide regardless of state restrictions.

In the meantime, understand what you’re paying for and why. The $1,200/month premium for your 26-year-old engineer isn’t actuarial reality. It’s policy choice. A choice made by regulators who decided that small businesses should bear the cost of social solidarity while large enterprises opt out.

That’s worth questioning.

We have arrived at a paradoxical destination. We spend more on healthcare per capita than any other nation, yet a growing number of Americans are unhappy with their outcomes, and many are opting to go uninsured in 2026, returning us to a pre-ACA precariousness. To fix this, we must stop patching a broken model and instead introduce a structural revolution: a government-backed safety net for the catastrophic, paired with a robust free market for the routine.

The Unintended Consequences of the ACA

To understand the solution, we must diagnose why the current intervention failed. The ACA was built on the noble intention of expanding coverage. It mandated that fully funded health policies cover a massive list of procedures to prevent bankruptcy. However, by mandating that every policy covers everything, the floor price of insurance rose for everyone.

Perhaps the most counterintuitive failure lies in the Medical Loss Ratio (MLR) rule. This regulation capped insurance profits as a percentage of premiums (forcing them to spend 80-85% of revenue on care). While intended to curb greed, it created a perverse incentive.

Consider a simplified example: If an insurer sells a policy for $1,000, their profit is capped at $150 (15%). If they want to double their profit to $300, they cannot simply become more efficient; under the MLR rule, they effectively need the premium to double to $2,000. By inflating the cost of care, they inflate the absolute dollar amount of their allowable profit. Furthermore, it drove insurers to acquire hospital systems to capture profits on the provider side—profits that were capped on the insurance side.

The result? Massive consolidation, reduced competition, and skyrocketing costs. We are now in a fiscal trap where premiums are so high that individuals cannot afford them without massive government subsidies. The government is effectively underwriting the healthcare costs of a massive percentage of the population, creating a fiscal time bomb that is unsustainable.

A New Philosophy: Safety & Choice

We need a reset. Healthcare policy should aim to achieve two distinct goals:

1. Security: No American should face bankruptcy due to a life-threatening emergency.

2. Affordability: Routine care and supplemental needs should be subject to market forces to drive down prices.

The current system tries to mash these two goals into a single, bloated insurance product. We should separate them.

Pillar 1: Universal Catastrophic & Chronic Maintenance Coverage

Instead of funding the bureaucratic maze of Medicaid, Medicare, and ACA subsidies, the government should shift its focus to a single, streamlined mandate: Universal Catastrophic Coverage.

This would be a basic right for every citizen, regardless of age, income, or location. If you are in a car accident, suffer a heart attack, or face a sudden, life-threatening crisis, you are covered. Crucially, this extends to chronic maintenance for serious conditions, ensuring that those with long-term needs are not left behind.

By consolidating this essential coverage under a unified framework, the government also gains the leverage to negotiate drug prices directly with pharmaceutical companies. Much like other nations, this monopsony power would allow the US to bring down the astronomical costs of novel drug treatments and life-saving medications, preventing price gouging while maintaining access to medical breakthroughs.

This model is far less financially taxing than the current Medicare system, which faces looming insolvency. Yet, it remains available to every senior and every citizen, regardless of their ability to pay. By stripping this coverage down to necessities—preserving life and limb—we ensure that the safety net is fiscally sustainable and truly universal.

Pillar 2: The “Health 401(k)”

With the catastrophic risk handled by the public sector, the private sector can return to what it does best: innovation and competition.

Currently, employers choose your health plan, insulating you from the actual cost of care. We should replace this with a defined-contribution model, similar to how the 401(k) replaced the pension. Employers would contribute a set percentage of an employee’s income—say, 10%—directly into a Health Savings Account (HSA).

For the unemployed or those currently on Medicaid, the government would provide minimal HSA contributions to fund their essential expenses. This ensures that even the most vulnerable are empowered to participate in the market economy rather than being relegated to a separate, often inferior, system.

The individual then controls the capital. With these tax-advantaged funds, they can choose to:

• Pay for routine medical expenses out-of-pocket.

• Purchase a supplemental insurance plan (similar to Medicare Advantage) that covers non-emergency needs.

• Save the funds for future health needs.

The Outcome: A True Market

The power of this shift lies in price signals. When patients spend their own HSA dollars, they become consumers. They will ask, “How much does this X-ray cost?” before the procedure.

To ensure this market functions effectively, we would introduce a “Federal No Surprise Bill Act.” This legislation would require providers to disclose exact costs ahead of visits. By mandating that services be paid for ahead of delivery, we effectively eliminate the predatory practice of surprise billing later.

This demand for transparency will force providers to compete. Hospitals and clinics will have to post prices and compete on value, just like every other sector of the economy, from LASIK eye surgery to veterinary care. When providers compete for patient dollars, prices fall, and quality rises.

Proof of Concept: The Singapore Model

This is not a theoretical fantasy; it is a proven model. Singapore has successfully implemented a system based on personal responsibility and government safety nets for decades. By requiring citizens to save for their own routine healthcare (via their MediSave program) while providing a catastrophic backstop, Singapore has achieved health outcomes that are superior to the United States at a fraction of the cost.

In recent years, Singapore’s life expectancy has hovered around 84 to 86 years, significantly higher than the US average of roughly 80 years. Even more damaging to the American status quo is the cost comparison: Singapore spends roughly $4,000 per capita on healthcare annually, while the US spends nearly $14,000. They achieve better health for less than one-third of the price by correctly aligning incentives: consumers shop for value, and the safety net catches those who truly fall.

Conclusion

We are currently worse off than we were 15 years ago because we attempted to regulate our way to affordability. It backfired. By simplifying the government’s role to providing a catastrophic floor and unleashing the power of consumer choice through HSAs, we can build a system that is solvent, affordable, and humane. It is time to stop subsidizing a broken system and start building a functioning marketplace.

The wind cuts through my hair as I pedal across the Williamsburg Bridge at night. The city skyline towers over the East River, a monolithic wall of light that makes my chest tighten. It is a mixture of anxiety and awe, a physical reminder of exactly where I am.

It has been two years since I moved here from Berlin. I came looking for a new personal challenge, but I found something I didn’t expect: a crucible. I have grown more in these twenty-four months than in any other period of my life. I have completely fallen in love with this city, but loving New York is not like loving a person; it’s like loving running a marathon. It will hurt and suck along the way, but it’s so much fun, exhilarating and will make you so much stronger at the other end.

The Crime of Trying

New York rewards a behavior that Berlin punished: trying hard.

I remember back in Berlin, my founder friends would ask me whether I am joining them for the Sunday morning rave. When I told them I only take Saturdays off and work Sundays, they treated me as a mood killer. Ambition there often felt like a social faux pas. To care too much, to strive too hard, was to be uncool. There was a comfortable ceiling to my life there.

In New York, that ceiling was blown off the moment I arrived. Here, mediocrity makes me invisible. The city inspires me through its sheer, crushing extremes. It demands that I be more. More charming, more attractive, wealthier, smarter, better in every way you can imagine.

It is a binary existence: If I stop trying, the city spits me out. If I keep pushing, it might just turn me into a diamond.

The Texture of Struggle

This intensity, however, came with a social cost I wasn’t prepared for. The dynamics here often feel jarringly performative compared to my time in Europe.

I recall meeting a German founder back at a networking event; we bonded over what was breaking and not going well in our companies. After a while, he told me that it had been a while since someone shared what was not going well. There is an intimacy in shared struggle.

But I remember my first month here, talking to an American founder who was pre-funding, pre-product, and clearly in the pivot-hell. Yet, he pitched me like he’d just raised a Series A. He needed his narrative like water in a desert.

Sometimes, talking to locals feels like interacting with NPCs in a video game. Be that countless “how do you do”s in a day or office kitchen chatter about everything and nothing at the same time. The conversations sound personal, but have no depth nor connection. The dialogue feels scripted, sheltered from the texture of real vulnerability. Back in Europe, I felt like people value knowing others at a level that matters. Only connections worth having are deep and personal ones. In New York, I found that most people who move here, particularly in the professional world, often only talk to others to climb. Social connection becomes transactional leverage where every conversation falls flat.

The Zero-Sum Game

There is a saying that people in San Francisco want to live in New York, but no New Yorker wants to live in San Francisco. The two cities are very different, but I’ve noticed a strong distinction in the “game theory” of the cities.

San Francisco feels like everyone is playing a positive-sum game - the logic of rising tides. You never know where the person you meet might be in a year so better treat them well. But New York feels zero-sum. Space is limited. Tables at Torrisi are limited. Status is limited. My win is often viewed as someone else’s loss. That’s also why conversations often end up being shallow and transactional.

Finding Home in the Noise

So, how do I settle for good in a place that treats friendship as a transaction and life as a competition?

I had to find the others. I had to find the people who had stopped pitching. Finding “home” in New York wasn’t about finding a nice apartment in the West Village - it was about finding the pockets of reality amidst the performance. It was finding the friends who would sit on a stoop with me and admit that they are tired, that the rent is too high, and that they are scared they won’t make it.

When I found that tribe, the transaction costs vanished. The city stopped being a battleground and became a playground. The city became a source of inspiration and connection.

The Cost of Admission

I watch people come and go. People come, stay for a few years. Then, they pack up for softer cities like Austin, Lisbon, London, where life does not feel like a constant competition. Some leave because the cost, financial, emotional, physical, eventually grinds them down.

But on those nights when I am cycling over the bridge, the skyline impossibly dreamy, I understand why I am here. New York doesn’t let me settle. It forces me to become a diamond.

And this has shaped what I want for my future. I no longer look for a safety net or optimize for hedging the downside. Now, I want to be the truest version of myself, the most ambitious version of myself and leave nothing on the table. If I look back at this period of my life, it should not be comfortable. It should be epic.

For me, there is no question whether the struggle is worth the cost. I am here to stay.

When I started my sabbatical in February this year I knew I wanted to do something that was truly me. Something where it feels hard to others, but trivial to me. It was easiest to follow ideas that have stayed with me for years and health optimization was the most obvious choice for me. I have gone through a journey of always being very aware of the inputs to my health and trying my best trying to optimize the outputs. Over the years I have developed habits that have helped me achieve peak performance energy, motivation and fulfillment without much effort. I knew I wanted to help others achieve the same. It was easy for me, but I knew many others where struggling with it. I knew I had a secret that I needed to share with others to help them reach greatness. It was the impact I want to have on the world. To help more people reach their greatest potential.

My journey quickly led me to explore the space of wellness apps and employee benefits that encourage people to reach peak productivity. The trickiest part for me was always how to provide people the tools to make the journey easier and natural. Wearables and wellness classes are all things people have to pay out of pocket meaning only a small part of the population will ever have access to them. Employee benefits such as mental health apps etc reach more scale, but it’s always harder to justify the spend as the employers when the performance improvements and better retention is questionable. Thus, my journey led me to an inevitable employer spend that I had been so frustrated about at my last company, which is health insurance.

I had to pick a health insurance plan for my team at my last company. Trying to provide the best benefits, I asked AI, searched Reddit and asked fellow founders - there was no one option for health insurance that caters for startup needs. Not any smarter, I picked the most expensive health insurance plan for my team. Everything broke. Insurance cards arrived months late. Medical procedures got denied. Our assigned PCPs didn’t exist. Preventive care wasn’t covered. The provider search didn’t work. I’d paid premium prices for a non-existent product.

It became clear that if I wanted to provide the largest amount of people access to preventive health and help them live the most fulfilling lives, the solution was starting a new health insurance company. The people get the awesome benefits, employers don’t have to pay more than they currently do. Next, how to start a health insurance company?

In March, I started talking to people in the industry and learning as much as I could. Why do the existing insurance companies suck so much. Why don’t companies vote with their wallets? Why are there so few options in California and New York? Why is preventive care not covered? And obviously, why are the plans so expensive for generally healthy, young teams?

I read every book I found about the health insurance market. I read books and articles about regulation and why the market worked the way it did. I read a book how you can reduce healthcare costs by not paying your hospital bills. I read a book that described how Obamacare was passed and which parties lobbied for which provisions. I read a book about how it came to be that health insurance ended up becoming an employer-sponsored benefit. I talked to health insurance experts and slowly started to catch up to them. I sounded knowledgeable. At some point someone said “you’re one of the few person that understands why American healthcare system is so broken”. I felt like I had become an insider.

In May, I started talking to investors about my vision to provide every American access to unlimited preventive health. I was so excited. They were not. Investors hate insurance businesses. They did not care that I am not taking underwriting risk. They did not understand the why now. They did not believe in the team, which was me. They said I don’t have any healthcare background. They asked why would you succeed where Oscar and Pearl failed. They did not believe in the distribution model. They did not believe in the market size. Despite the market being the largest part of the US economy.

Rejection was plentiful. I wavered and lost belief myself in what I was doing, but I could not give up. I knew how much I had hated my health insurance. How I was willing to pay for a premium product and how that premium product did not exist. I feel like the world was against me. However, even in the darkest hour, I had support from a few that never wavered. There were friends. There was my family. And there were my potential customers.

Customers. Whenever I told founders like me what I am working on, they got visibly excited. They loved the mission and they felt the pain. They wanted the product. Now. Without asking how much it costs. That gave me the confidence that I have seen the future. I have a secret that the rest of the world does not yet know. I knew I was right, despite most people around me not believing me.

The first conversation where it clicked for me was in San Fransisco. I was at a Nordic founders event and told another founder that I am building a new type of health insurance plan for startups. He immediately became curious. He told me a story how his emergency department visit bill was denied, despite having a great employer-sponsored plan. He told me his employees have no clue how to find providers in the network and avoid care due to that. He was beyond excited about there being a better way how to do insurance. His passion for something that was still an idea in my head showed me that I am truly onto something that people care about.

Then I had more conversations with other founders. People told me how they did not care which plan they had. They did not know which carrier they have, what their premium is or even where their insurance card is. They knew the product they had purchased was useless for them unless they get into an accident. Anything that they actually cared about was not covered by insurance. More and more conversations would end with “When can I sign up for this?” without asking me how much it would even cost.

So, I kept going. I kept going because I knew this had to exist. I knew this is the largest problem for me to work on right now. This is my life mission.

I also realized this was the point where I had quit or pivoted in my previous startup adventures. When the going got tough, I had looked for a way out. I had looked for a way out since the world did not agree with me. This time was different though. I knew I was onto a secret. My prospective customers were pulling the product out of me.

Fast forward to June. I decide to put the fundraising behind me and learn everything there is to learn about health insurance. I enrolled in health insurance broker courses and learned every regulation and rule there is out there. I completed 40 hours of studying separately for California and New York. I passed the insurance exams in both states with flying colors and finally became a legally licensed health insurance broker. I could finally start helping startups find the best suitable health insurance plan. I learned about their pain and helped them the best I could. My secret sauce was being a licensed broker in an old-school industry AND understanding the needs of startups. A unique combination very few others have.

I found an engineer to help me build the brand and consumer app. I found a partner for managing prescription drugs. I found a partner for managing claims processes. I found a partner for pricing the stop-loss policies. I found a partner for giving our members unlimited access to primary and urgent care. It’s all possible. You just need to keep on going. I talked to ten reinsurance partners before the first one was willing to take a chance on me. Most companies in the insurance industry hate innovation. They want tried and tested. It took me persistence to convince a single carrier to do business with me. They have doing business for 40 years the same way, but they found my idea compelling enough to take a chance on me!

“A year ago, I would have quit by now. When investors said no, I always found a way out. I’d tell myself ‘the market isn’t ready’ or ‘the timing is wrong.’ I’d pivot to something easier.

Not this time. Because this time I had something I never had before: customers begging me to build it.

When a founder tells you ‘I want this now, I don’t care what it costs’ - that’s not feedback, that’s a command. When three teams start using your product before it’s finished - that’s not luck, that’s validation.

I became a licensed broker because customers needed it. I built partnerships because customers demanded it. I’m launching in January 2026 because customers are waiting.

You can just do things. Not because you believe in yourself - I’ve believed in wrong things before. You can just do things when customers are pulling the product out of you. That’s the only permission that matters.

Don’t wait for investors. Don’t wait for advisors. Don’t wait for the industry to accept you. Find the people with the problem. If they say ‘yes, now,’ everything else is just logistics.” First, it’s impossible. Then, it’s inevitable.

Health insurance is tied to your employer in the US and thus, most employees expect to get a health insurance plan along with a monthly salary. For tech startups, health insurance ends up being the second largest cost, roughly being 15% of salary costs. I have seen founders repeatedly pick a plan at random or worst, delegate the decision to their payroll provider salesperson who does not know anything about their team.

This guide breaks down exactly how to structure health insurance as you scale from 1 to 500+ employees, with specific guidance for the NYC and California markets. I’m a licensed health insurance broker in New York and California, and I’ve structured plans for dozens of startups navigating these decisions. What follows is the pattern that actually works at each stage.

At any stage of a startup, your employees will expect benefits and will likely have 100% employer contribution with their previous employer. To be on the safe side, I would suggest reimbursing 100% of your employees health plan costs if you can afford it so you don’t have to worry about losing a great employee because of it. Every employee needs to be offered the same benefits so changing that number down the line is painful. So, if you have raised, pay the full cost of the plan and roll with it. When it comes to reimbursing spouses and families of your employees, there is less of a standard. It’s safest to reimburse 50-100% of spouse/family costs. However, it’s unlikely somebody would reject your offer if you don’t do that. Go with your gut, how generous you want to be with family health benefits.

Which Plan Tier Should You Offer?

The ACA metal tiers: Bronze, Silver, Gold and Platinum represent actuarial value, not quality. Platinum plans cover roughly 90% of costs, Gold covers 80%, Silver 70%, and Bronze 60%. The difference shows up in deductibles and copays, not in which doctors or hospitals are covered.

For NYC and California tech startups, the safest choice is to go with Silver since it’s quite standard. I’ve seen companies try to save money with Bronze plans, and it might get a weird look from some candidates. Your candidates are comparing your offer against companies offering Gold/Platinum PPOs with $1500 deductible. If you’re offering a $6,000 deductible Bronze plan, you’re signaling penny-pinching on employee health, which might not be a problem worth explaining.

The better funded you are, the more it makes sense to move up the Tier ladder. If you just raised 25M, it would be a weird look to offer Bronze or Silver. The moment you compete with Big Tech for talent, health benefits really matter.

Which Carrier Should You Pick?

In New York, you’re primarily choosing between UnitedHealthcare, Aetna, Empire Blue Cross Blue Shield, Oxford (a UHC subsidiary), and Cigna. Empire has the strongest network density in the five boroughs, particularly with major hospital systems like NYU Langone and Mount Sinai. Oxford is solid for Manhattan-centric companies but thinner in the outer boroughs and Westchester. UnitedHealthcare and Aetna have the broadest national networks, which matters if you have remote employees or people who travel frequently. Cigna tends to price aggressively but has a smaller NYC footprint. If you care about a nice UX experience, Oscar Health, is a solid option, but expect a smaller network.

The network question isn’t just about size, it’s about which specific providers your employees care about. It will likely not matter much at earlier stages. If half your team however lives in Brooklyn and uses NYU Langone for everything, losing that network access will cause immediate complaints. Your broker should run a network disruption analysis against your current employee addresses and known provider preferences before you switch carriers. Don’t assume all networks are equivalent.

California is dominated by Kaiser Permanente, and you need to have a position on it. Kaiser is an integrated delivery system, they’re both the insurer and the provider. You see Kaiser doctors at Kaiser facilities, use Kaiser labs, and fill prescriptions at Kaiser pharmacies. This model is genuinely cheaper, often 20-30% below comparable PPO coverage, because Kaiser eliminates the fee-for-service inefficiencies. The quality is generally excellent; their outcomes data is strong.

The problem is some people viscerally hate it. They don’t want a closed system. They have an existing relationship with a non-Kaiser specialist and don’t want to switch. They perceive it as restrictive, even though Kaiser’s network is large. You’ll get pushback. The standard solution is offering both Kaiser HMO and a PPO option from Anthem Blue Cross, UnitedHealthcare, Aetna, or Health Net. Employees who prioritize cost choose Kaiser; employees who prioritize flexibility pay more for the PPO. Expect roughly 60-70% Kaiser take-up if you offer both.

One thing that doesn’t show up in rate sheets but matters enormously: claims processing quality and member services. Some carriers are competent at adjudicating claims correctly the first time, others routinely deny valid claims and force appeals. Some have knowledgeable member services reps who can explain EOBs and help navigate the system; others have offshore call centers reading scripts. Ask your broker which carriers have the fewest escalations and member complaints in their book of business, that might help you make a better decision.

1-10 Employees: Just Use a PEO plan

At this scale, trying to be clever about health insurance is a waste of time. You’re better off joining a PEO like Justworks, Rippling, or Sequoia One and get access to solid benefits through their pooled risk. You’ll pay roughly $800-1,200 per employee per month all-in, which covers payroll, benefits, and workers comp. The entire administrative burden disappears for onboarding, compliance and payroll integration.

The tradeoff is straightforward: you’re locked into whatever carrier relationships your PEO has, and you can’t customize much. But at 5 employees, you don’t need customization. You need to ship product.

There’s an alternative if you have a distributed team or want tighter budget control: ICHRA. Your employees pick their own plans on the individual marketplace, and you reimburse them tax-free. Set your contribution at $750 in California or $1000 in New York per employee and you’re done. Platforms like Thatch make this administratively trivial. The downside is your employees now have to understand deductibles and networks, which many resent.

The value that you get from ACA marketplace plans is really bad since they bundle everyone into the same risk pool and you don’t get the benefits from being a group. However, at this stage, overpaying for bad benefits is not the end of the world. The startup trick here is often to pick the cheapest plan from the marketplace (catastrophic or bronze if you’re over 30) and use the rest of the ICHRA budget for things you actually care about. Just keep in mind that you will need to pay whatever healthcare costs you have out of pocket since the deductibles will be high!

What you absolutely should not do: ask employees to buy their own plans and reimburse them with the company. This way, the money you spend on health insurance is not tax deductible and health insurance will cost you 30-40% more after taxes.

If you are pre-funding and don’t have money for a plan, there is one more clever way how to get employees coverage while spending a lot less. There are innovative solutions that work as healthcare cost crowd funding platforms, the most popular being Crowdhealth. This will be a post-tax expense for you, but you will be able to provide your team a catastrophic coverage for roughly $200 a month per employee that will work just fine at this stage.

10-25 Employees: Decision Point

Somewhere between 10 and 25 employees, you need to decide if you’re staying with your PEO or graduating to broker-sourced coverage. If your growth trajectory is uncertain or you’re comfortable with your current setup, stay put. But if you’re reliably at 15-20+ and want to control your plan design, it’s time to work with a benefits broker.

Note: a benefits broker is not an insurance agent. Different incentive structures, different expertise level. The broker sources fully-funded plans from carriers like Aetna, UnitedHealthcare, Anthem BCBS, or Kaiser if you’re in California. With fully-funded plans, you pay a predictable monthly premium and the carrier assumes all claims risk. You’ll typically see $800-1,200 per employee per month for a gold or platinum tier plan, though NYC and California run 15-25% higher than national averages.

Plan types matter here. PPOs give employees maximum flexibility to see any provider but cost more. HMOs lock them into networks and require PCP referrals but reduce premiums. HDHPs paired with HSAs shift costs to employees through high deductibles while giving them tax-advantaged savings accounts. Most startups default to a PPO if they can afford it, your team members don’t want to deal with referral bureaucracy.

One thing to understand about California and New York: small group pricing is community-rated, meaning everyone pays the same regardless of individual health status. This sucks for younger, healthier, teams, but the trouble of using self-funding might not make sense yet at this team size.

25-500 Employees: Level-Funded is the Play

Once you hit 25 employees, you should start considering level-funded plans. Here’s why the economics change: with fully-funded plans (traditional), you’re paying the insurance carrier to assume risk you probably don’t have. If your population is young and healthy (which most tech startups are), you’re mostly subsidizing other healthcare claims through pooled pricing.

Level-funded plans unbundle this. You pay a fixed monthly amount that covers three things: a claims fund, stop-loss insurance, and administrative fees. If your actual claims come in below projections, you get the surplus back at year-end. If claims spike catastrophically, stop-loss coverage kicks in, typically at $30-50K per member. The result is you keep the insurance company’s margin, which usually translates to 20-30% savings if you have a healthy workforce.

The catch is complexity. You now have fiduciary responsibility under ERISA, you need to fund the claims account monthly, and you have to manage stop-loss certificates and claims reporting. This absolutely requires an experienced broker, don’t try to DIY this. Your level-funded plan admin will take care of ACA reporting, HIPAA compliance, and COBRA administration. The administrative burden is real, but at this scale you should have ops infrastructure anyway.

The upside is meaningful plan design flexibility. You can optimize deductibles, copays, and network configurations in ways that aren’t possible with off-the-shelf fully-funded plans. You can even offer benefits that are not normally covered by health insurance like wearables and gym subscriptions. This is what makes level-funded plans powerful, along with their lower cost of insurance.

500+ Employees: Go Fully Self-Funded

At 500+ employees, you’re leaving money on the table with traditional insurance. The carrier’s margin, typically 15-20% of premium, is pure waste. Cut them out entirely.

With fully self-funded plans, you pay all claims directly. There’s no insurance carrier; instead you hire a third-party administrator (TPA) to handle claims adjudication, network access, and member services. Advanced TPAs like Rightway, Collective Health, or Bind offer reference-based pricing models. You can also use carrier-based TPAs like Aetna ASO or UnitedHealthcare if you want their networks.

The economics are compelling: you typically save 30-50% compared to fully-insured plans because you’ve eliminated the carrier margin and can optimize plan design aggressively. But you’re taking on real risk. Monthly claims are volatile, and while stop-loss insurance can protect you from catastrophic scenarios (you can get specific stop-loss at $1M per member etc), you should have cash reserves set aside to weather this.

The real advantage isn’t just cost. It’s control. You can now do things like contract directly with primary care providers, establish centers of excellence for surgery and oncology, implement reference-based pricing where you pay Medicare rates plus a markup instead of billed charges, carve out pharmacy benefits with transparent PBMs, and integrate virtual-first solutions like Maven, Spring Health, or Teladoc.

This requires real infrastructure: a full-time benefits manager or fractional benefits consultant, data analytics capability for claims trend analysis, and board-level approval for the risk tolerance. You won’t see savings for 12-18 months. But once it’s running, the cost advantages and customization options are substantial. Every large American company is using a self-funded health insurance plan. It’s here to stay.

What about HSAs?

HDHPs with HSAs deserve mention because they’re financially clever but culturally difficult. The tax advantages are real: employees contribute pre-tax dollars, investment growth is tax-free, and withdrawals for medical expenses are tax-free. For high earners who can max out contributions ($4,300 individual, $8,550 family in 2025), it’s basically a secondary 401(k). But employees hate high deductibles - $3,000-5,000 before insurance pays anything - and many won’t have the cash flow to fund the HSA properly. Unless your entire workforce is financially sophisticated and highly compensated, expect pushback. The typical compromise is offering an HDHP alongside a Gold PPO and letting employees choose.

Regional Specifics You Can’t Ignore

New York has the strictest community rating in the country. Demographics barely move the needle on small group pricing. Also, when using ICHRA, all of your employees are forced into the same risk pool with everyone between 30-60 years old, which means you will be paying a lot for mediocre benefits. You also have to deal with NY’s mini-COBRA continuation coverage rules, which are more generous than federal COBRA. And Paid Family Leave integration with disability carriers is mandatory, so make sure your broker understands the interplay. New York has also banned stop loss insurance for groups less than 100 so you’re better off being part of PEOs for longer if you want the best benefits.

California is dominated by Kaiser Permanente’s integrated delivery model. It’s cheaper than PPO equivalents but locks employees into Kaiser’s network, which some people love and others hate. California does allow stop loss for small groups, but you can only get specific stop loss insurance starting from 40k onwards, meaning you still need to take some risk as an employers when moving over to level-funding.

If you still have question about health insurance, feel free to reach out!

From the early 1900s to the 1930s, the family doctor was typically seen at home, paid in cash or even barter. The village doctor that my grandparents grew up with wielded immense authority, their advice rarely questioned. The Flexner Report of 1910 ushered in an era of formalized medical training in North America and cemented physician authority, reinforcing this patriarchal trust.

By the 1940s and 1950s, primary care shifted from homes into hospitals and offices. Employer-based insurance became widespread due to wage controls during World War II, institutionalizing medical care. Though doctors still retained authority, the relationship grew more transactional.

The 1960s saw the birth of “primary care” as a specialty, with Medicare and Medicaid expanding healthcare access. Family medicine emerged in 1969, rekindling continuity and lifelong patient-doctor bonds within a formalized setting.

In the 1970s and 1980s, managed care reshaped primary care again. PCPs became gatekeepers controlling costs and specialist referrals, sometimes straining patient trust. Meanwhile, patient-rights movements and informed-consent laws empowered patients, gradually shifting toward a shared decision-making model.

My most recent health plan was a HMO which meant I needed a referral for every specialist visit. Since I had freshly moved to New York City, I was randomly assigned a PCP by my new insurance plan. I wanted to do a sleep study and for that I needed a referral from my PCP. When I tried calling the office where my doctor was supposed to work, I was told that the next appointment is in three months. I could not be bothered to wait three months to get a referral so I paid for the study out of pocket. For most Americans, this is not an option and they will postpone or ignore necessary specialist visits.

By the 1990s, the internet introduced patients to an overwhelming amount of medical information. “Dr. Google” transformed the PCP’s role from unquestioned authority to interpreter and guide, responding to patients who arrived with printouts and questions. Google became the top-of-funnel solution people tried before needing to go to a doctor. There was definitely also a cost factor, since Google is free and people gladly tried to solve their problems for free before needing to want and pay for a PCP visit. Surprisingly, people searching their symptoms online did not lead to a decreased number of in-person visits, but it did lead to more informed discussions during those visits.

Today, ChatGPT can assist patients with routine tasks remarkably well. Have a late-night question about a common rash or cold symptoms? ChatGPT answers instantly, clearly, and without the inconvenience or expense of an office visit. In fact, a tool like ChatGPT is often better than a human at providing rapid, reliable, standardized responses to straightforward queries: 24/7, without fatigue or wait times. Furthermore, AI often has memory now, which means that it can follow your symptoms for days/weeks/months and give you a much better diagnosis than a PCP who only sees a snapshot of your health during a visit. Both methods however need good data in and the user needs to know which prompts to use and what data to include to get accurate symptoms. With more data, better reasoning models and improving IQ for foundational models, ChatGPT is overtaking a lot of traditional PCP responsibilities.

Yet, today's PCPs still have important responsibilities: preventive care like screenings, vaccinations, and lifestyle counseling; diagnosing and treating common illnesses; coordinating care among specialists; educating patients and advocating for their needs; and building long-term relationships based on trust and understanding.

Many of these tasks are already being replaced or significantly enhanced by AI. Routine tasks such as refill requests, reviewing normal lab results, scheduling appointments, and initial symptom assessments are now being handled faster and more accurately by algorithms. AI-powered scribes, like Nabla, Abridge and Nuance’s DAX, have freed physicians from hours spent typing into electronic health records, allowing them to reengage directly with patients.

Moving forward, AI will and should take on an even larger role, especially in preventive care and chronic disease management. AI-driven health platforms can continuously analyze data from wearable devices and health apps, providing personalized insights and timely interventions to prevent illness. AI's ability to track complex chronic conditions in real-time, predict flare-ups, and adjust treatment recommendations proactively can substantially improve outcomes and quality of life. The adoption will likely happen in two phases: first AI creating treatment plans that doctors need to approve and over time taking over simpler workflows completely. Just like in B2B SaaS - first you have the copilots and then come the agents.

This technological advancement will significantly expand access to high-quality healthcare, particularly benefiting rural and underserved areas where PCP shortages are common. Patients previously limited by geography or healthcare availability will have constant, expert-level guidance at their fingertips.

However, even the most advanced AI cannot replace certain human roles. Accountability, intuition, ethical decision-making, and building trust through genuine personal relationships will always remain uniquely human capabilities. Human doctors that have more longitudinal data about their patients can sense hidden anxieties, understand unspoken fears, and provide nuanced emotional support and reassurance during critical health moments. It’s easier for human doctors to rely on their gut feeling and push back to patients when they feel they are not being fully honest. It will take some time for AI to develop such instincts.

For purely data-driven clinical decisions, AI is already superior to doctors. I have been helping a friend’s parent understand how their chemotherapy is going and whether the sessions are improving their situation. For such a simple data crunching exercise, I feed the research model all data and I get a very detailed report how the biomarkers are changing and how the treatment is working. The report summary has always been very similar to the doctor’s comments, but the depth of the report would have taken a human a full day to compose. AI is already better than any human doctor, but because health is so sensitive, we still need a human in the loop for many clinical decisions.

The same applies to having the last call for clinically more dangerous procedures like high-risk prescription drugs. With enough real-life training data about clinical outcomes, AI can start to prescribes drugs with a great safety profile, but we want to always have the clinicians in the loop for more dangerous drugs and diagnosis. Where to draw the line legally will be a tricky challenge to solve.

As we embrace AI's potential, more and more control and decision making will move to the patient. They will always have the most data, be most invested and soon have the best models in the world to guide them in their health journeys. This does not mean that the human connection should be completely forgotten. Most Americans lose the connection to their PCPs after leaving their parents’ homes. You can keep your doctor if you don’t mind doing virtual visits , but physically moving around and changing jobs makes it challenging to keep the same PCP and develop a relationship with them. Right now we are getting the worse of both world, where we don’t have a connection to our family doctor anymore (thus loosing the data that doctors have about our parents and grandparents), nor can we use the data from our wearables and advanced diagnostics like polygenic risks from DNA sequencing. The primary care will need to be more mobile and data-led to give modern families a more holistic coverage.

Even if you are ok with sticking to your childhood doctor using televisits, it’s a lot harder to keep your doctor if your new employer insurance plan does not have them in the network. Good solutions here could be either people using direct primary care providers and pay out of pocket for their subscription. That’s kind of what OneMedical and formerly Forward Health provided - same PCP experience no matter where you lived. Most of their customers paid out of pocket for the membership and visits. It will be interesting to see the direct primary care evolve as AI will take over lots of their responsibilities. A likely scenario here is that the AI-led tools will become financially a lot more accessible and the out-of-pocket costs for the rest of the PCP interactions will be manageable to a larger part of the population.

My dream is that AI will enable the best of both worlds for primary care: more consumer-led healthcare, better access to primary care and a stronger bond between the PCP and patient going forward. Our laws will need to evolve to enable AIs to take some of PCP responsibility and free them from a lot of unnecessary work. Right now, we are focused on the copilots in healthtech and soon we can start expanding access to better primary care. Every American having access to a concierge like primary care service does not seem like a distant dream anymore. It’s an exciting time to be a consumer and builder in healthcare!

That illusion of safety shattered the night my friend called me from New York.

To illustrate that with a story, I will tell you what happened to my friend Dom. He has a history of heart arrhythmia and went through treatment last year in Hungary, where he is from. He spends a few months each year in the U.S. and has travel insurance just for emergencies. He is super active and very health‑conscious, so he rarely sees a doctor. When his heart started racing, he called me and said, “I need the ER—how do I avoid bankruptcy?” not “Where’s the nearest cardiac unit?” I was scared for him, then angry that cost anxiety came before care. Domo’s story isn’t an outlier; it’s a symptom of a badly aligned market.

Why is care so expensive in the US?

Salaries are high, but not twice those of other Western nations. In theory, private competition should push prices down. Instead, every layer is mis‑aligned. Employers choose plans that must cover government‑mandated essential benefits. Insurers must price policies without knowing individual risk, so younger, healthier workers subsidize older, sicker ones. Employers rely on brokers who steer them toward plans that maximize the broker’s commission. Nobody—employer, broker, or insurer—feels true accountability for cost or outcomes.

Given the price tag, you would at least expect world‑class service. You don’t get it.

I learned the same misalignment first‑hand at an urgent‑care clinic in SoHo. One winter morning I woke with piercing ear pain, searched for the closest clinic, and called to confirm they “accepted all insurance.” Fifteen minutes later I flashed my card—only to hear, “Marketplace plans aren’t in network.” My company was paying $1,600 a month for the top‑tier policy, but none of that mattered.

I stayed anyway, desperate for antibiotics. After a brief look in my ear, the doctor veered into mood questions, vaccination upsells, and back‑pain screenings. Five minutes later I walked out. Months afterward I discovered charges for each unnecessary question—$15 for one antibiotic pill, $30 to swallow it in the office. The fee‑for‑service model incentivizes more CPT codes, not better outcomes.

Prevention should be valued—yet it isn’t

Ironically, I’m the type of patient insurers should court—obsessed with prevention. I tested Oura, Supersapiens CGM, EightSleep, Levels, gut‑biome kits, regular bloodwork, VO₂‑max labs—you name it. Thanks to diet and exercise, my reactive medical costs stay under $1,000 a year. But insurers profit from a member for only 1.5 years on average, the “actuarial risk horizon.” They have no financial reason to fund long‑term prevention when I could switch plans next year. Employers’ horizon is only four years. The only stakeholder rewarded for staying healthy is me—yet I can’t pick a plan that rewards me back.

Trying to fix it for my own team

Even when I tried to solve this for my team, the system punished us. My last startup used individual coverage health‑reimbursement arrangements (ICHRA), an allowance model that lets employees buy any Affordable Care Act (ACA) marketplace plan and keep leftover dollars for other health expenses. After hours of research I chose an Anthem platinum plan with a $600 deductible. I learned the hard way that marketplace plans primarily serve lower‑income Americans and are price‑capped by state committees. To hit those caps, insurers shrink provider networks.

Healthy employees “hack” ICHRA by picking the cheapest bronze plan they never intend to use, pocketing more allowance dollars. But that still leaves a $10k deductible wall if anything bad happens. If this is the peak of employer innovation, the outlook is grim.

To understand why every modern fix feels like a band‑aid, we have to rewind to 1942. Roosevelt froze wages during WWII, so employers dangled health coverage as a perk. The 1954 Revenue Act made employer‑sponsored insurance tax‑free, cementing the link between job and coverage.

Commercial health insurance grew with minimal regulation. High‑risk individuals paid higher premiums, so the uninsured rate hit 15%, and medical bills became America’s top cause of bankruptcy. Over time, drugs, procedures, and administrative layers inflated costs; admin alone now adds ~25% to every bill.

In 2010 the Affordable Care Act banned risk‑based pricing, created subsidized exchanges, imposed essential‑benefit rules, and capped insurers’ overhead at 15% of premiums. Premiums and benefits were regulated; underlying prices were not. Result: a “private” market with government rules and runaway costs.

All those layers culminate in today’s unsustainable math. Marketplace networks keep shrinking while per‑capita spend tops $14k—2‑3 × other rich nations. Medicare and Medicaid now eat 21% of the federal budget. Fifteen years after the ACA, the uninsured rate has inched from 15% to 11%, and medical debt still drives most bankruptcies. Something has to give.

Is there a better model?

Many Americans point to Canada or Europe’s single‑payer systems, but converting the U.S. wholesale is fiscally and politically improbable. If the U.S. seems trapped, look at how Singapore escaped a similar bind.

Singapore spends one‑quarter what America spends per person yet boasts world‑leading life expectancy and a mostly private delivery system. Employers pay 10% of salary into mandatory health‑savings accounts (HSAs). Prices are transparent, and patients pay cash from their HSA, creating downward pressure on providers. Public hospitals are subsidized so no one is denied basic care, while a universal insurance layer handles catastrophic costs.

Why am I telling you this? The U.S. already runs a parallel idea—the 401(k). Make HSAs universal and portable, push routine spending to cash, and market forces will reward prevention. Larger HSA balances become a retirement asset, directly aligning health and wealth.

Enter Rivendell

Rivendell borrows Singapore’s two‑tier logic—cash‑driven prevention plus catastrophic back‑stop—and adapts it to U.S. regulations. I’m not running for president, but I am tackling the biggest problem of our generation by building this company. Our mission: give every American real incentives for preventive care, while shielding them from financial ruin. If that vision excites you, join us. We need missionaries and artists alike. Let’s build the system America deserves.

After my journey with my previous startup, Acapela, I joined a venture fund as a partner. My goal was to source deals and take time to figure out whether and how I’d start another venture.

Each month, I deeply explored a new topic and met potential co-founders to kickstart something new. Despite previously experiencing two challenging co-founder relationships, I believed I had learned enough to identify a great match next time.

Fast forward to early 2023: ChatGPT had just launched, becoming the fastest-growing internet product ever. Coming from productivity tooling and developer infrastructure, I researched how companies could better leverage large language models (LLMs). To avoid crowded productivity spaces, I focused on the less glamorous but critical problems enterprises faced with LLM deployments. In Europe, compliance and observability seemed like perfect angles.

By March, I’d collaborated unsuccessfully with four potential co-founders on different prototypes and ideas. With a strong thesis about LLM observability but no partner in sight, I faced a dilemma: start immediately alone or wait and potentially miss the opportunity. I decided to dive in solo, confident I had both the technical and sales skills necessary—and believing I could always bring in a co-founder later if needed. Well, it turned out to be harder than I thought.

Fundraising, however, was smooth. I raised $1 million in just two weeks from supportive Solo GPs and angels, starting the journey positively.

Next came product development and team building. Hiring freelancers from my previous team, we rapidly built features to keep pace with the swiftly evolving market and customer demands. Revenue appeared, but users struggled to articulate exactly why they paid us. Despite apparent growth, true product-market fit remained elusive.

Loneliness quickly became a reality. The freelancers clocked out promptly at 5 PM—a tough reality for a pre-seed startup founder. It hurt to see I was building something my customers and team didn’t deeply care about. I even doubted if I genuinely cared about the problem myself.

Change became inevitable. When the European AI Act got delayed, the compliance angle evaporated overnight. It was clear I needed to pivot significantly, changing both the market and the team.

Fortunately, I already had a US work visa. Wrapping things up in Berlin quickly, I said an emotional goodbye to great friends and headed west, choosing New York over San Francisco simply because it felt magical and better aligned with my new market—financial institutions.

In the US, I shifted the product toward sales engineers and began actively searching for a co-founder. Solo-founder life had proved emotionally exhausting, and sacrificing equity to relieve that burden seemed entirely worth it.

Finding a co-founder mid-journey, however, proved difficult. What should a trial period look like? Should they build something new or continue with the existing product? I experimented with two talented individuals. One wasn’t a perfect fit, and the other—who initially felt ideal—decided after three months he’d prefer starting something entirely new. After tough investor conversations, I realized I wasn’t prepared to discard all we’d built.

After a final desperate pivot, I recognized that true product-market fit and substantial growth were unlikely. Mentally preparing to wind down, I received an unexpected offer from a customer to license our product. Seeing this as the graceful exit I needed, I swiftly wrapped things up and redistributed the remaining funds to investors, taking a necessary break.

Conversations with investors went easier than expected. One expressed surprise I’d persevered beyond a year; another pinpointed a core issue—a lack of a clear problem. Both observations resonated deeply.

Reflecting on these two challenging years, several lessons stood out clearly:

Firstly, investing in my health was crucial. No matter how difficult things got, maintaining my fitness routine provided energy and mental resilience to overcome lows.

Secondly, raising funds from supportive pre-seed investors made a huge difference. They understood startup struggles and provided invaluable support throughout.

Thirdly, nothing replaces a true co-founder. Regular check-ins with investors and peers helped but didn’t eliminate loneliness. You can’t replicate the emotional support a committed co-founder provides.

Fourthly, selling to the US market proved far easier than Europe. American enterprises actively engage startups, anticipating future disruption—a stark contrast to Europe’s skeptical stance.

If I could redo it, I’d begin immediately—even with an imperfect co-founder—instead of waiting. Progress, even imperfect, is invaluable.

A key takeaway: never optimize a startup purely for venture capitalists. My VC background influenced my decision to pursue ideas that seemed attractive to investors. However, truly great businesses rarely start with obvious ideas aligned with existing VC theses. They’re usually niche, deeply personal ventures driven by intense founder curiosity.

Another personal challenge was the fear of publicly admitting changes or failures. As a second-time founder, I mistakenly believed everyone expected me to have everything figured out. This prevented me from openly sharing my struggles, limiting potential outside help.

Most importantly, pick an industry you’re genuinely passionate about and willing to commit to long-term. Successful startups often emerge from deeply personal interests, initially considered niche or obscure.

After ending Commonbase, a trusted friend reassured me I wasn’t a broken founder and emphasized the importance of pursuing personal interests. Commonbase hadn’t truly reflected my passions, making sustained enthusiasm difficult.

Moving forward, I’m focusing on something deeply personal: longevity and healthcare, inspired by my experiences navigating healthcare in the US. To my surprise, I found no startups effectively addressing preventive healthcare access. This gap has become my new passion.

As the saying goes, “A healthy man has a hundred wishes; an unhealthy man only has one.” Health will be at the heart of my future efforts.

Thanks for following along on this reflective journey. Stay tuned for what’s next!

1. Living with too many rules rarely leads to

   happiness.

2. Say yes to invites - it shows you care and you never know who else you will meet.

3. Hard conversations, easy life. Easy conversations, hard life.

4. Seek mentors early. Keep seeking them.

5. Don’t chase the hype.

6. The greatest source of energy is curiosity.

7. Doing high-status activities rarely leads to interesting outcomes.

8. Venture further. You can always come back home.

9. Be more weird. Weirdness attracts interesting friends.

10. It’s ok to not be a founder.

11. Read broader. Before reaching for another self-help book, ask yourself whether you want to be more like the author.

12. Read physical books. And keep them. They become part of your mind.

13. It’s ok to let go of people if your values drift apart.

14. Lean into your roots. They will help to keep you grounded.

15. Move to the US. The slope is especially important at the start of your career.

16. See your best friends at least once a year in person. You both will appreciate it.

17. Revisit friends you have drifted apart from. Sometimes you will be surprised.

18. Work on interesting problems to attract interesting people.

19. Not every company succeeds. Every founder succeeds at some point.

20. It’s not easy to build a reputation. It’s very easy to ruin it. Keep it in mind especially during tough times.

21. Your best friends should be your harshest critics.

22. Never lose the ability to make fun of yourself.

23. Keep most of your money in ETFs. Stock picking is an expensive hobby, not fixed income.

24. You will never regret investing in your health. It’s probably even more true when you feel like you lack time or money.

25. Before making big decisions, ask yourself how your future self would look back at it in five years.

26. Verbally saying out what’s the worst case scenario often takes away half of the stress.

27. Burnout comes from working on something you don’t care about, not from working too hard.

28. Don’t save money on high-quality beddings. Paying for high-end pillows and covers guarantees a happy moment before falling asleep.

29. Trust people generously until they give you a reason not to trust them.

30. Embrace being a beginner. Not wanting to look stupid will otherwise hold back a lot of personal growth and funny experiences.

I am sure that the current version of me will appreciate lots of these points going forward. I am not sure whether every point would hold true in ten years of time. Let’s live and learn.

So what changed? Land war is back on European soil. Europe has experienced a relatively peaceful experience since the end of WWII. During the Cold War, Europe experienced uprisings in the Soviet block, but there was no active warfare. After the fall of the Soviet Union, we had a few more minor scale wars in the Balkans, but apart from that, things were relatively quiet until 2008, when Russia invaded Georgia. In April 2008, NATO promised to consider Georgia’s bid to join the alliance, and in August 2008, Russian forces invaded. The war was quickly over, but Georgia lost a small part of its territory, gave up its NATO bid, and Russia got away without any repercussions. Due to no reaction from the Western nations, it was the beginning of the undermining of the USA’s global hegemony.

Fast-forward to the end of 2013. Former Ukrainian President Viktor Yanukovych decided to drop the Ukrainian association agreement with the EU, which sparked the Euromaidan uprising. The former President fled the country in March 2014, and a few days later, Russian troops entered Crimea, declared a referendum for Crimea to join Russia, and by the end of the month, Russia had officially annexed Crimea. The result? Russia had once again occupied a piece of another nation’s territory without any consequences from the Western countries. It was clear by now that attacking independent nations and occupying pieces of their territory was tolerated by the Western nations since the cost of going to war was too high. An active warzone war was created in the Donbas region, but the oil kept on flowing from Russia to Europe, and nobody batted an eye.

In February 2022, Russia invaded Ukraine and started a full-scale modern land war on European soil between two major nations. No Western European nation believed Russia would attack and scoffed at Eastern European nations sending Ukraine military support well before the war broke out. In Jan-Feb 2022, Germany was accusing the Baltic countries of being naive and warmongering for sending Ukraine support. On February 24, those accusations stopped because the war broke out. Everyone expected Ukraine to lose the war quickly, but due to good luck and a strong will to fight the much bigger enemy, Ukraine managed to fight back and not lose its independence. This time, the response from Western nations was more substantial, and they decided to support Ukraine financially and with military equipment.

Since the start of the invasion, the Western nations have promised to support Ukraine with all kinds of weapons. Despite being much more powerful economically than Russia, Europe has not been able to provide Ukraine with the support it needs. EU’s GDP is more than ten times larger than Russia’s. Despite that, the combined military equipment output is often short of providing Ukraine with the same army equipment Russia can produce. To make matters even worse, also with US military support, Ukraine has roughly 3-5 times less military equipment than Russia can deploy on the field. Where do we go from here?

At the start of 2024, Ukraine is running out of ammunition and men to fight the war effectively, not considering taking back lost territory. The European economic support package was stalled behind the Hungarian veto, and US military support was blocked behind the Republican party veto. Once the EU’s 50B support package was approved, it amounted to 0.02% of the block’s yearly budget. That level of commitment does not seem to signal any urgency or importance. There is no reason to point fingers, but Ukraine will be forced to negotiate peace at less-than-ideal terms without EU and US support. The result of Russia winning a proxy war against the entire Western alliance will show the world that the power of balance is not as straightforward as it used to be at the end of the Cold War. The Western world can no longer play the world police, and it might make sense to reconsider your alliances. It will be a significant geopolitical win for Russia.

To make the situation worse for Europe, Russia now produces roughly three times more military equipment than at the start of the war. What happens after the current conflict ends? It is estimated that it would take Russia 4-5 years to rearm and refill their ammunition stockpiles completely. After having won a proxy war against the Western nations once, it’s pretty clear that its ambition will not be limited. The question of whether its next target could be Moldova, Finland, the Baltic nations, or Poland is anyone’s guess. It’s very unclear whether the US would uphold its NATO commitment, especially if there is a leadership change in the White House in January 2025. Where does that leave Europe going into the second half of the 2020s? Alone and vulnerable.

The sad truth is that the European industry needs to be more capable of producing enough missiles to upkeep our promises to Ukraine. Right now, Ukraine is shooting 2k 155mm missiles a day, and Russia is shooting 10k in a single day (Ukraine used to shoot 6k rockets at the start of the war, with Russia sh. To match Russian firepower, Ukraine would need 300k missiles a month. Currently, the EU can produce 14k missiles monthly, and the US 28k missiles. Even with combined efforts, the Western nations are coming more than five times short of the number of rockets Russia can deploy. The massive demand for 155mm missiles has also pushed their price from $2100 at the start of the war to $8400. There’s a tremendous financial incentive to produce more, but nobody seems to pick up the slack. The US plans to increase its production to 80k missiles a month by 2025, but that’s still not close to enough to sustain another long-term land war. To rub salt into the European wound, while the EU managed to supply Ukraine with 300k missiles in 2023, North Korea supplied Russia with 1M missiles. With an economy 500 times smaller than the EU’s. The emperor clearly has no clothes, yet nobody dares say it out loud.

When it’s clear that Europe does not know how to produce enough war machinery to defend its borders, perhaps we can supplement our shortcomings in hardware with startups instead. Some progress was made in this space when the NATO Innovation Fund was announced to invest in early-stage defense startups in NATO countries (the US did not participate in the fund). While this is a step in a great direction, the fund has not made a single investment in a year, so death by stagnation seems to be the case at all levels of our national security. Unfortunately, European venture capital firms are often not allowed to invest in defense tech startups, making the funding environment even tricker. If financing was not tricky enough for startups, selling to different European countries should be easy, right? Not so fast. To sell your defense products to a government, you first need to go through field testing and get a green light from the soldiers about the efficacy of your product. It is only then possible to open a procurement process in which the startup can participate. You would expect that once you sell to one nation, the rest should be easy. That’s not how it works, though. With every country, the process starts from scratch, and you need to have someone open the battlefield testing door for each army separately. That means extremely long sales cycles that are not repeatable, and there’s a very fitting term in defense tech about the time between the first POC contract and getting deployed on the field: the Valley of Death. You can figure out why it’s called so yourself.

We have a looming war on our borders, and we don’t seem to be the least prepared for what’s coming for us. What can we do today to be ready to deter the threat looming over us? On a broader level, the nations should finally wake up and start practicing realpolitik again. I feel like some European countries are so blindfolded that they expect Russia not to continue their “military operations” because it’s immoral. I mean, what the actual fuck. When did we replace realpolitik with woke national security policies? A balance of powers built by Otto von Bismarck in Europe created one of the longest-lasting peace periods in Europe from 1871 to 1914. The only way to get back our national security is to create a new balance of power that would prevent any aggressive country from using tanks to exercise their foreign policy. How can Europe return to a balance of power in as short an amount of time as possible?

Europe must become united, innovative, and nimble to restore its national security guarantees. That combines becoming energy independent, creating more local production capabilities, a unified procurement system, and better funding sources for defense startups. Let’s start with energy independence. Instead of relying on cheap imported energy from the US and Russia, Europe needs to focus on affordable local energy sources. That will likely mean a combination of solar, wind, and nuclear energy in the short term. Those energy sources must be cheaper than the imported alternatives, so we must deregulate new production capabilities to reduce costs. The only reason why nuclear power is so expensive is regulatory costs. Despite ambiguous chances of success, Europe should try to lead the path in fission energy research. So far, most of the research has been funded by private and research money - there is no reason why Europe should not win the race to build the first fission power plant against the US. Making energy an order of magnitude cheaper in Europe will make it possible to become a competitive manufacturing powerhouse again. There are alternatives to relying on cheap Russian gas for our factories - we must push to invest and build cheaper energy sources of tomorrow.

Next, moving on to ammunition production. Unlike the US, Europe relies more on private companies to produce most of its weaponry. That is currently the task of four large manufacturing companies in Europe: Germany’s Rheinmetall, Britain’s Bae Systems, France’s government-owned Nexter, and Nammo, owned partly by the Norwegian and Finnish governments. These producers are unwilling to increase production as quickly as needed because they first want governments to commit to decades-long munition orders. Completely understandable. Which is why we need an alternative to these behemoths. Ammunition requirements change often, and production needs to be elastic. Europe should focus on building independent blocks of munition factories that are easy to duplicate and refocus. The best analogy here is Tesla Gigafactories. Instead of a monolith, the Gigafactories consist of many identical smaller manufacturing lines that can be quickly scaled up and down or switched to a different model. Ammunition production should work the same - as a collection of smaller factories distributed across Europe that can be quickly repurposed and scaled up and down. Both to bring down costs and make the production more elastic.

Next up is defense procurement. Right now, most members do their procurement independently for their armies. NATO also has procurement capabilities, but they are being used on a smaller scale. The way procurement works in defense is that suppliers need to find a way to do a field test with the responsible military unit, which tests and writes a report about the battle efficacy of the product. A government procurement project is created to buy the product if it fits their purpose. That process needs to be replicated by each member country separately. That slows down the commercialization process of defense companies, and it’s the main reason US defense tech does not bother selling to Europe - the contract sizes compared to the effort is a rounding error compared to selling to the Pentagon. To fix that, European nations need a unified procurement alliance where international units test new products, and the procurement system works the same way across the countries. Once an army unit tests a new product on the field, each country can create an order for the product if they find it useful for their militaries. This way, the defense companies only need to go through field testing once and can sell to each country participating in the system afterward. Since the European Union is a non-military alliance, such a procurement alliance will likely need to live independently next to NATO. It’s improbable that the US wants anything to do with such initiatives, and thus, creating something stand-alone will likely increase the iteration speed and adoption rate of countries. Making Europe the fastest adopter of new defense tech solutions will help us attract the best founders to focus on the European market first. It’s important to note here that an organization called Defence Innovation Accelerator for the North Atlantic (DIANA) was created to mentor Defense tech startups, but judging from their performance in the first few years of operating, it seems to be a classical government organization mostly wasting startup time and not providing them with the most critical asset - revenue.

Finally, we get to defense startup funding. Next to increasing the production of standard ammunition, we need our defense startups in Europe to build the latest defense systems specialized for the European theater of war. Defense tech has historically been a taboo topic in Europe. Due to their LP commitments, most venture capital firms cannot invest in dual-use or military-only companies. We need more funding sources and founders building Europe’s latest wave of defense tech. The NATO members (apart from the US) created a 1B NATO Innovation Fund to finance European defense tech startups, but it does not serve its purpose well. It feels like another government initiative with zero urgency and no real motive to provide excellent returns to governmental LPs. Governments channeling capital to startups is significant only when distributed through hungry venture capitalists motivated to pick the best startups to increase their carry. Thus, Europe needs a much larger capital vehicle as a fund of funds to scout new money managers who want to focus on investing in defense. We need many orders of magnitude more defense funds investing in the next wave of European defense founders. This fund of funds will likely need to be 10B+ in size and keep on backing the best money managers who reliably invest in the best founders in Europe. Since defense tech is still a rapidly evolving topic, Europe still has the chance to play a globally significant role here. The US funds focusing on national security investments (such as the American dynamism fund by a16z) are looking at deals in the US. By creating a new class of great defense investors in Europe, we could make both the best funding and commercialization environment in the world for startups focused on national security.

Europe, the historical birthplace of most of Western culture, is now in danger of being destroyed by another major war. The Russo-Ukrainian War did not wake many European nations to the looming threat. Donald Trump threatening to encourage Russia to attack weaker NATO members does not seem to matter. The world is changing, but many still do not get it. Europe needs to get its act together. Truthfully, the European Eastern flank is actively preparing for the next war, but the rest of Europe must also step up. Funding another GDPR-compliant ESG-reporting startup will not help us when a tank rolls across our border. Let’s focus on defending everything beautiful that Europe stands for to ensure the European continent remains the best place to retire for the next generations. There’s a lot to be done, but I am confident we have the people and resources to make Europe feel safe again. We lack only urgency and willpower. I stand with Europe.

Many new startups have now decided to open-source their source code. Looking at the latest YC batch, roughly a third or so of the companies focusing on developers as a target group are open source. Approximately 5% of all companies from the batch are open source, including many that do not target engineers. Most startups are doing it to go along with the trend and defend themselves against another group of college grads applying to YC with your product’s Open Source X version. A great example of this would be Helicone, which started as a closed-sourced LLM monitoring tool, and once YC had accepted a few open-source versions of the same thing, they also decided to go open-source.

There are valid reasons for being open-source. The one already mentioned, just avoiding competition of open source versions of the same product, is primarily a fad. I would strongly argue that most companies adopt the best tool out there, not a shittier open-source version. Supabase is winning over developers from Firebase because they have a better product, not because they are open source. Being OS allowed them to capture a lot of marketing hype, but that’s all there is. Where open source does shine, though, is helping with hiring. Engineers often prefer working for open-source companies, and finding new employees amongst your contributors helps with the hiring funnel. If someone is a great OS contributor, they will likely also make a great full-time engineer. Another positive factor in the B2B enterprise space is creating trust with your open-source solution concerning security practices and continuity. Some deals might be easier to close if you tell the buyers they can look at the code to ensure it’s secure (they never do and mostly still want a SOC II, but it can help) and tell them they can self-host if you go out of business. Does it make a huge difference? No. While there are upsides to being public with your code, there are also downsides that people rarely don’t talk enough about.

The biggest downside to being open-source is converting your users to paid customers. Many OS projects initially start down-market and get a lot of adoption amongst hobby hackers and early-stage startups who would not make great customers anyway. Making revenue from those users is often a challenging process. While GitHub sponsors, etc, do help with making some money from sponsors, it’s rarely enough to pay even a single full-time engineer a salary. So, while many open-source projects create a ton of value for the engineering community, they rarely capture any significant part of it. I know too many great builders that got a few thousand stars for their repos but dropped the effort after six to twelve months since they failed to monetize their projects. So what’s the alternative?

The alternative is returning to the roots and making money before building anything. It’s fun to build cool shit and get lots of GitHub stars for the thing that you are building. Stars, however, are a terrible approximation of creating value. Thus, to avoid the issue of false signs of progress, builders should focus on making revenue instead. There is always a chance to open source your repo later if that’s important for you. If they paid for your closed-source project, they will also pay for your open-source project!

So, what’s my advice to builders deciding whether to build open source? One, I would not open source before you have made sure you can make money with what you are building. Second, I would not go the OS route too early. It’s cool to have people report a bunch of errors and feature requests, but it can also lead you down the wrong path. In the early stages of company building, listening to valuable feedback is more important than listening to everyone. Feedback from paying customers is gold. Feedback from solo hackers can lead you down building stuff that does not matter at all. Ignoring requests and bug reports can be mentally tough, though, since being responsive and “listening to customers” feels productive. Ignoring the instinct at an early stage can be very helpful.

It’s important to note that going open source is more or less a one-way door. Close sourcing an existing solution with users can piss off people very hard. I have rarely seen it being done successfully. Second, if you go OS, do so smartly and know what you are trying to get out of it. Avoiding open source X competition from YC batches is not a good answer. Finally, if you think of best-in-class software products (Linear, Figma, Framer), none is public. It’s just another testament that nothing great has been built by a committee.

The name “love languages” is a bit of a misnomer since the concept applies a lot more broadly than just romantic relationships. I would instead call the theory “modes of appreciation,” but it does not sound so catchy, so let’s stick to “love languages.” Everyone has their own ways of perceiving being appreciated, which applies to dating, friendships, and work. Just like when we might perceive being most loved when our partner tells us they love us, the same concept applies to work. It’s very likely that for the same person, the best way to show appreciation for their excellent work might be to announce in front of their colleagues what a fantastic job they did. Alternatively, if someone really values receiving flowers from their partner, they might feel most appreciated at work when they receive a bonus or raise. Being aware of this pattern helps us sense more appreciation from those around us and makes our intentions more evident to our friends when we are trying to show that we care about them.

If love languages are so essential to how we perceive other people interacting with us, where do they come from? It’s always a question of nature vs nurture, and in this instance, I have developed a small theory of my own that love languages are entirely nurture-based. More specifically, our dominant love languages are formed in early childhood based on how our parents showed us their love. It seems to hold true based on anecdotal evidence from my circle of friends. In my childhood and more-or-less until today, my parents have always shown me their love by doing small things for me when I needed them (also when I did not need them). Simple things include picking me up from the airport or always having my favorite drinks (Coke Zero, if you wondered) in the fridge whenever I return home. I do not remember any occasion when my dad would not have picked me up from the airport whenever I arrived, and my mom ensured all my favorite things were there when I got home. They seem trivial, but they affect us for the rest of our lives, so it’s better to be aware of them. Perhaps it’s time to ask yourself how your parents showed you their love in your childhood?

I have often wondered whether love languages change throughout your life. I have too little data to make any firm conclusions here, but the most significant service we can do is to be aware of the preferences of the people around us and make sure we are conscious of it when raising our children. Maybe it is not worth it to put your children through the same childhood trauma and never hug them or tell them you love them. It seems to be especially cruel in Nordic cultures, where people don’t tend to be very physically affectionate. Anyways, we often try to act in the opposite way of our parents, which is probably one area where such behavior might make sense. Especially if you felt you lacked specific modes of appreciation as a child. Telling your loved ones you love them is one of the most precious gifts you can give.

After all this theory, how can one apply this mental model? First, figure out your dominant love languages and compare how people around you are showing you their appreciation. In a work environment, people often have “how to work with me” documents, which should be a crucial part. Also, talking about love languages early with your partner helps avoid many unintended misunderstandings. Often, when we don’t feel appreciated in a relationship, it’s due to a mismatch of love languages and not due to a lack of care. The opposite is also crucial - how aware are you of the close people around you? Do you know your partner’s or best friend’s love language? The most common assumption people make (me included) is that people perceive love the same way you perceive love. Thus, if my dominant love language is acts of service, that’s also typically how I show people that I care. I have yet to develop a foolproof method to learn about other people’s love language when I meet them. Generally, people who read the love languages book are also thrilled to share theirs. So far, talking about it has yet to backfire. Such a small thing, yet so powerful. Let’s share our love with the world the way it understands.

I started both Mirage and Acapela in the most classical 50-50 cofounder equity split. You find a partner in crime, you divide areas of responsibility and you hit the road. This setup can work well, but there are a few things to look out for.

First, there should always be an agreement between the founders who can call the final shot when there is a deadlock in making an important decision. I would go even further and split the equity at least with one share difference to make that symbolic distinction. Cofounders overriding each other when making decisions should not happen often (weekly etc), but making decisive decisions quickly is the main competitive advantage of a startup and this helps to keep the decision momentum. One of the common causes for slow deaths for startups is having a CEO, who is slow to make decisions and communicate them clearly. Both as a cofounder and an employee, I would much prefer a strong-willed CEO, who makes decisions quickly and is willing to correct course when new data emerges.

Moving on to how to choose that one-in-a-million cofounder to get married with after a few months (hopefully) of cofounder dating? I have gone through plenty of cofounder conflicts and also had to part ways in the past which has taught me a few things about what not to do. Never compromise. If you have doubts about your cofounder, keep on looking. It’s very similar to a romantic relationship - it’s much better to be single than be with a not-so-great-fitting partner. By and large, I have learned to look for two key qualities in my cofounders. I need to admire them professionally in at least one area or in other words, they need to have a superpower in one skill set where I could not become so good even if I tried very hard. That keeps you on your toes and helps you appreciate your cofounder through the lows and highs. Starting to feel like you could do a better job than your cofounder in their role is a highway to a breakup. The second thing I look for in any potential cofounder now is wanting to spend time with them also in a non-work related setting. They do not need to be my best friends, but being able to laugh at stupid shit and go hiking or whatever together to release some steam is crucial to managing your relationships through the lows. Again, I would draw many parallels between a romantic relationship here, where being friends with your partner is paramount to living a happy life together. Even when the love is not there (or financial success in the case of a company), you still stick together due to personal affection towards one another.

Now that finding this perfect partner is out of the way, why have a cofounder in the first place? After having founded Commonbase by myself, I realized just how lonely the journey is. Even when you are very strong-willed and put in the same amount of energy working solo, the journey will always be lonely since you can never have a cofounder-like relationship with your other team members. This alone might be a strong enough reason to have a cofounder next to you for the whole journey.

Another aspect of talking against solo founding would be not having someone to bounce ideas off of who cares just as much and who has almost as much context as you. As I have been working on the current company as a solo founder, I have tried to use my investors and friends as this sounding board, but unfortunately, your investors are not as deep in the weeds and your friends just don’t care about the success of the company to the same level as you do. Thus, they are not really great alternatives to having a cofounder.

What are scenarios then, where solo founding makes sense? Honestly, not so many. The only one I was able to think of is the case where you really need to start a company immediately and you don’t have that ideal person next to you when you kick things off. As I said before, having no partner is better than just a placeholder. Also, incorporating a company alone does not rule out bringing in another person next to you at a later point. In many cases, it might be even easier to convince the right person once there is already a little bit of progress made.

What would I recommend to folks who end up being solo founders nevertheless? You need to replace all emotional and balancing aspects of a cofounder with other people around you. Have a circle of other founders around you to release steam constantly. Ask your investors to be “hands-on” and actively question your decisions so that you would have an active sounding board before you waste too much time working on the wrong things. It’s also easier to attract amazing talent when you create a larger-than-normal employee pool, for example, 20% or 30% of total shares to make early joiners care more. My final advice to any solo founder (or friend in general) would be to have an amazing romantic partner next to them. Having someone who listens to you and understands all the emotional rollercoasters of a founding journey cannot be replaced. A great partner can make the difference between success and failure - as PG said, do not forgo love for the sake of your company. You will be doing both yourself and your company a disservice.

Finally, if you are a solo founder going through this journey right now, please do reach out. Even though I might not know what you are struggling with every moment, I will understand what you are going through emotionally. That’s worth more than gold.

I had my first chance of living in the US when I was 16 and got an opportunity to represent Estonia at a student exchange program in North Carolina. That was my ticket to the US, I thought. Promised Land, at last. Thank God almighty I am free at last.

I headed over from Europe and landed right in the middle of nowhere in the US. Well, it’s North Carolina after all. I spent half of my time there on the university campus and then some in two host families. My first experience in the States was a colorful one - my host families tried to convert me, taught me how to shoot automatic rifles and lots more. It was fun. The whole experience of living and studying there made me less starry eyed about my childhood vision of moving there permanently. I took it as my next goal to go to college in the US. In reality, there were only two colleges that I was interested in, Stanford and MIT.

The next chapter in my adventures was less rosy. I took all the necessary tests and ended up applying to MIT early. I could not go to Stanford since they did not have full financial support for internationals… After a few months of waiting came my first childhood trauma - my application was not accepted. Thus also ended my dream of going to a US college.

My next chance of going to the US came when I started my first startup, Mirage, in Zurich, Switzerland. We wanted to move the company to California and decided to postpone taking VC funding in order to get into YC before any other external funding. We ended up applying and getting interviewed by YC three times. Three times flying to Mountain View for a ten minute interview just to get rejected each time and fly back empty-handed. Another time when the opportunity felt so close, yet so far.

The next few years made me feel much more settled in Europe and I developed a really good rhythm in Berlin. The idea of moving never came up during the covid years, especially since I was building a remote work company with both my co-founder and partner living in Berlin. After my last company, I had to ask myself the existential question of where to live again and the idea of moving over to the US came up again. Initially I was very happy with life in Berlin despite feeling like I had explored everything the city has to offer. After starting my latest company in the LLM infrastructure space, I initially started building the team and doing sales in Europe, but I was slowly frustrated. Frustrated by a lack of ambition and risk-taking.

That leads me to the current chapter. I submitted my visa application the moment I started Commonbase since I knew I might want to move the business to the US at some point. What I did not realize was that point came quicker than I expected. From a business perspective, American companies are much further ahead in their LLM adoption which moves my dream market there. Personally, I wanted to try living in a more intense and ambitious city than Berlin. It was clear that it was time for me to finally make the move. Here I am, writing this on the day my O-1A visa was stamped into my passport, giving me the ability to live and work in the US for the next three years. So what’s next?

I am moving permanently to NYC at the end of October in order to start the next chapter. I spent quite a bit of time this year in SF and realized it’s not for me. The city is expensive, dangerous, echochamber and despite catering well for super early-stage companies, not that great for later stage companies I am selling to. NYC on the other hand is closer to Europe, more international, diverse and just a hotspot for all kinds of people at the top of their field. Why move to the US at all after spending most of my life in Europe?

The US is by far the largest and most dynamic country for building global tech companies in the world. If you look at the most successful consumer products built in the last few decades, all but one have been built in the US (Spotify is the only counter-example). It is the country where the most impactful companies get built, where the most exciting research gets done (75% of 2023 STEM Nobel prizes live in the US) and where the most ambitious people move to if they want to reach the absolute top of their fields. I am excited to compete with the best and try to win in the big league. Being surrounded by the tops in their respective fields will be my personal gunpowder. Lots of adventures ahead - both personally and professionally.

To the next ten years chasing my American dream 🥂

But what’s really at stake? And do businesses capitalizing on unprecedented consumer demand for AI have more than just their interests in mind?

Regulating AI has many different stakeholders, but the largest one is society in general. Many jobs will be changed completely and new jobs will be created by the technological changes that come along with that. OpenAI for example hires a big workforce inhouse to fine-tune their foundational models with reinforcement techniques where the humans give the algorithms feedback which responses work best. Also, many companies are now hiring stand-alone prompt engineers in addition to prompt engineering becoming a part of most white-collar jobs. 

The topic was not as relevant until late last year when OpenAI publicly released their latest GPT-3.5 model. The results of the model surprised everyone since nobody was expecting the technology to feel so human-like. Finally, the moment arrived where people were both in awe and scared about what might come next. This started a general discussion around the quick progress in the performance of generative AI and its impact on the society in general.

In the last decade most people thought that AI will first impact mechanical jobs by replacing them with robots, then influencing very repetitive digital jobs and only then coming for white-collar jobs and finally replacing software engineers. The reality turned out to be different. The best early use cases turned out to be the most highly-paid ones - code generation, summarizing/explaining legal texts, doing sales outreach or upselling to customers based on their recent activity. The quick adoption of the technology (ChatGPT reached 100 million users in 5 days) and its potential impact on jobs and privacy quickly started discussions amongst the regulators. That culminated in a Congress hearing with the CEO of OpenAI, Sam Altman, and two other technology leaders from the industry. The Congress went as far to suggest creating a new regulating entity that would be run by the CEO of OpenAI… It is very clear that regulation in the US is moving just as fast as the generative AI industry itself with the goal of protecting user harm. It is not a question of if, but a question of when the regulators will want to make sure that the technology will benefit the society as a whole, not just a few private research companies.

Besides the general society, there are also a lot of industry experts that are very aware of the rapid progress in AI and that are calling for more oversight and regulation to slow down technological progress. In March 2023, many famous experts and entrepreneurs signed a petition to stop the research on state of the art AI models in order to put better guardrails in place to reduce the risk of building a dangerous general artificial intelligence system. They claimed that we need better regulations and oversight over the leading research institutions to make sure that in the course of the AI race we would not accidentally create a system that will be harmful to humanity. The memorandum was not officially aimed at any single company, but it was very clearly meant to target the latest breakthroughs created by OpenAI. Even more strangely, when the CEO of OpenAI testified in front of Congress and was asked about how he would regulate AI, he proposed to create an international agency that would be allowed to regulate and audit all the leading research institutions. It seems as if the most powerful people working on pushing the progress of artificial intelligence seem to agree that we need more regulation to oversee the entire industry. It is very rare to see private enterprise ask to be more tightly regulated - what is so different about AI?

The main reasons why most people in the industry favor early regulation are there just being a realistic risk of a very bad outcome for humanity in case things go wrong and lessons learnt from the past decade of failed regulation in the digital currencies industry. The first reason, summarized as the AI risk, is the possibility of an artificial general intelligence system turning against humans. We don’t really know how big the risk of that happening actually is, but since it is a possible outcome, most people in the industry are very open to having plenty of international oversight early to make sure that humanity as a whole is in control of what’s being developed in those research labs. Here, trading a little bit of transparency and overhead for the sake of avoiding the worst outcome when moving towards general artificial intelligence is definitely worth it. The other reason for more regulation is much less spoken about. Digital currencies have existed since 2008 and ever since they were created there have been countless stories of how the technology is being misused and how consumers are being victims of countless fraud cases. Despite this, there has been very little clear legal guidance from any major Western jurisdiction. Neither the European Union nor the United States has created clear legislation on how digital currencies and other blockchain-related products are treated legally. This has slowed down the adoption of the technology by every mature industry and rightly so, since large companies have reputation and profits to lose in case they do not comply with “non-existing” rules. The lack of regulatory clarity in the blockchain industry pushed founders away from the US to start companies elsewhere, where the regulation is more clear (or there is no oversight). The companies that stayed in the US have been hit with heavy fines despite trying their best to comply with all the existing rules and interpreting them as best as they can. This has led to a weird situation where some of the leaders in the crypto industry in the US (for example CEO of Coinbase, the largest crypto exchange in the US) sued the SEC because they are not responding to Coinbase’s questions nor willing to give any clarity how the existing laws are being applied to companies. The leaders in the AI space have definitely learnt from that and welcome clear rules around data privacy and the development of more advanced intelligent systems. In the long run, this will increase the adoption of the technology by larger companies and thus push the entire industry forward.

Despite no regulation slowing down adoption of new technologies by large companies, too broad regulation can also slow down entire industries as we saw with GDPR. Even though data privacy laws had to be updated, GDPR arguably went too far where internet users globally have to click through cookie banners that nobody reads nor cares about. With regulation you always have to get the balance right between protecting consumers and not stifling innovation. The regulators should learn from the parts of GDPR that did not have the desired effect on protecting consumer data and design the upcoming AI Act so that every constraint to developing and deploying AI models would allow oversight over the models without slowing down new innovators. Unfortunately, larger companies have the resources to comply with regulations, which makes the smaller companies the ones that most suffer from new regulation. Therefore, drawing a clear parallel from GDPR, we should make sure that the upcoming regulations for AI do not block open source projects from launching competing models. We need open source to thrive in the age of AI since that’s the only way to make the newest models available to all market participants, not only the ones who are willing to pay the closed research companies. Also, having competing open source models will make it easier for companies to fine-tune their models and self-host them to make sure customer data does not leave their premises. We should learn from previous regulation waves and make sure the rules protect consumers while not slowing down the less well capitalized players in the market.

We find ourselves at an interesting moment of time, where the EU is actively working on the AI Act and the US is slowly drafting their own legislation to control the AI industry. It is very likely that those laws will be passed already this year and come into effect by 2024. Even though the laws are not finally written, it seems likely that every company building and providing a service that uses advanced AI will need to go through audits and continuous monitoring. Even though that might seem like a high price to pay in order to use generative AI, it will speed up the adoption of the technology by even the most conservative industries and companies. Such regulatory compliance will be harder for smaller companies than incumbents due to needing more financial resources. Therefore, cheaper and easier solutions are needed to make companies of any size stay within the new rules. We have yet to figure out how generative AI will affect every job, but it is inevitable that every company and individual will be affected by this technology in the next few years. It will significantly increase the productivity of every worker and who knows - maybe the 4-hour work week dream will finally become a reality.

To give a little bit of context to each lesson, we went through five major product iterations during our two years of building Acapela. In chronological order they were:

• Stand-alone threaded platform for having asynchronous meetings to replace repetitive video calls with your colleagues.

• Automated meeting agenda builder that would always ping the creator and/or participator of the meeting to create an agenda and summary of a video call.

• Slack competitor for having all of your company’s collaboration in spaces and rooms always dedicated to a specific conversation.

• Collaborative todo-app for Slack which makes sure that every request you make on Slack is always assigned to someone in the channel.

• Unified inbox for all your notifications.

Our first product was meant to offer companies a better way how to replace unproductive meetings with async collaboration spaces. While our value prop seemed to resonate with users just like Kitkat, we failed to realize how our new tool would fit into the existing meeting culture and tooling.

Our very first lesson learned about meetings was that every meeting is useful for at least one person in the room. Even when everyone always complains about how unproductive their meetings are (this meeting should have been an email) there is always at least one person in the room, for whom the meeting is very valuable. They probably did not figure out a better way to make a decision within the organization and this is why they scheduled a meeting to force action by the end of it. Most people in meetings might feel like their time is being misused, but sometimes they find themselves on the opposite side of the table where they are the ones organizing a meeting just to force action by having a 30-minute slot. By the end of it, you need to come up with a decision. In such a working culture you end up in a situation where everyone complains about how many unnecessary meetings they have, but in reality, they are happy to put up with them, because you never know when you need to force action using a 30-minute blocker in everyone’s calendar. For example, I remember talking to a head of product at a famous fintech company which is known for its highly efficient working culture. He showed me his calendar on our user research call and it became clear that he was spending 40+ hours in calls during his week. When I asked him about getting work done outside of the meetings and how productive he felt the calls were, he was very satisfied with his time usage and productivity. He was also not bothered about how the meetings were affecting his colleague’s work since in his view work was getting done. The first lesson learned is don’t trust what the user is complaining about. Check whether they are actually trying to solve the problem they complain about.

Lesson number two - your product has to fit into the existing tooling amazingly well or completely replace them. Whenever you start building a product it should be totally clear whether it will live next to existing tools (such as Slack in our case) or whether you intend it to be an addition to current tools. That will very strongly influence your product decisions and how nice you want to play without tools (to integrate or not). That’s also a part I often saw many future-of-work startups get wrong where they did not have a strong stance on this and the product ended up having some integrations into existing tools, but then also building their own sources of knowledge. We decided to build a tool that should replace Slack and not build any integrations. Our solution was clearly not solving a big enough problem to justify introducing an extra silo of knowledge within an organization while keeping Slack around. Also during user feedback calls for our collaborative todo app, it became very clear that users wanted their todos to live in Slack. People did not care about our fancy web app which we had put so much effort into building. Even though the web app and the Slack integrations worked in perfect sync, every user told us during user feedback how we could improve the Slack integration, not what was missing from the web. Most future-of-work startups in some way or another compete with Slack and they have to figure out whether they want to compete or complement. For example, threads.com clearly decided to compete whereas Loom went out of its way to make the Loom experience within Slack as smooth as possible.

The third lesson stuck with me from every product iteration we tried to push to the market - it is not possible to solve bad culture with better tooling. A lot of the problems were the result of bad company culture. Often, managers and founders try to make their teams more productive by introducing more tooling to keep up good habits, but this rarely works if the company itself is not already trying to walk the talk. I remember talking to a CEO of a fully remote company that had only one weekly recap call within the company. She was so deliberate about the work culture at their company and this single weekly call was a great testament to that. She immediately understood the problem we were trying to solve with our async meeting platform, but she just did not have any meetings to replace. Another example about tools not fixing cultural problems is from companies deciding to install Clockwise for more efficient meeting coordination. It will have very little effect internally if the company actually does not care about respecting other people’s focus time more than always scheduling calls to discuss problems. The same experience stuck with me with every product iteration we had. Forcing people to fill in meeting agendas with tooling did not work if there is no culture of preparing meetings ahead of time in the entire company. Asking employees to make their Slack messages more actionable by always tagging at least one responsible person also does not make the team more productive if there is no habit of always having at least one person responsible for every initiative. This lesson applies to both leaders, but also builders. If you want to make your teams more productive, start with culture. And for founders, if you are planning to build productivity tools, make sure you are targeting customers that are already taking action which your tool will just make ten times simpler for them.

Next lesson is a short one and mostly applies to unit economics - it is very hard to build a venture backed business building a single-player tool that has no virality nor network effects. There are two different concepts in network-based tools: virality and network effects. Virality means that using the tool in its normal way will attract more users to the product. A great example would be document editing or video call tools where using the tool will naturally introduce the tool to more people potential users since they are by default collaborative. Network effects play out in products where more users using the tool with in the same subatomic network increase the value of the tool for other users. Great examples are Slack and Tinder, where respectively your colleagues or single people living around you increase the value of the tool for you. If your product has neither, it will be extremely hard to make the unit economics work for a venture-scale business. Both virality and network effects make sure that your long-term cost of customer acquisition has a chance to stay below customer lifetime value. We noticed it very clearly when building our last product iteration - a unified inbox for all your work notifications. The tool was a single-player tool and we quickly realized that by only charging single users while having to attract new users via paid marketing had negative unit economics from the very beginning. You can obviously hope for crazy word of mouth growth since your product is so much better, but in my opinion that should be a given for any startup. The word of mouth growth will help your unit economics, but it will not fix it.

The last lesson is a somewhat more complex beast - we experienced that adding network effects to a product retroactively is extremely hard to pull off if it is not a natural part of the product. In the B2B SaaS space that often means trying to turn a single-player tool into a collaborative tool used within a team. Our unified notification inbox was quite naturally a single-player tool and due to unit economics we had to try to make the tool collaborative. We considered introducing shared inboxes, creating more value for the users by cross-referencing files from your colleagues notifications and creating stateful notifications (sent, received, confirmed, done), but all of those efforts were rather far-fetched. I am sure we would have figured out a way how to provide more value to the user in case their colleagues were also using the product but the initial product experience would have suffered and it was clearly another zero-to-one effort, not just incremental improvements. I see the same thing play out now with a number of other originally single-player tools (for example Raycast) and it will be very interesting to see how they manage to introduce multiplayer features into a product that was started as a single-player tool.

We had a very exciting journey with Acapela and I am still rather bullish on the asynchronous collaboration space. There were many hard lessons learnt trying to find PMF in this space. I am sure many companies will be built in this space and I hope our learnings will help future founders avoid some of those mistakes. If you are actively building a related product, please reach out and I would be happy to provide more insights into our journey and how you could learn from that.

The first part of my life was all about starting to look and act like an adult. The part that followed after that was all about exploring myself as an individual and trying to figure out what I value, desire, and enjoy. The part that I am transitioning into now is focused on doubling down on the things that I love, am good at, and how I want to be remembered as an individual.

I grew up in a small nordic country mostly spending my childhood amongst myself. I spent a lot of time around my grandparents because they clearly had more time to put up with my endless curiosity and energy. My grandparents were amazing people who showed their love and support at every moment, but by and large, let me explore the wonder of life by myself. I ended up spending a lot of time alone playing around in the woods or with different electronics that my grandfather had. I was allowed to try out almost anything and it was rare that my grandfather tried to stop me from failing. I remember an occasion when I tried to recharge the car battery myself and I ended up killing the battery since I tried to charge it with 240V instead of 12V. I was obviously ashamed to kill the battery of my grandfather’s Lada, but such lessons strengthened my curiosity and gave me more confidence to try out anything.

Another part of my childhood that shaped me very strongly as a person was how my parents shared their love and affection with me. I grew up in a typical Estonian family where physical touch and words of affirmation are seldom to come by. On the other hand, small acts of service and activities together were always an important part of our family life. I had not thought about this aspect of my childhood a lot until I had a conversation with my partner about why it is that my dominant love language is acts of service. Funnily enough, the same holds true also today. My parents still show their affection by always being there when I go home. That might literally be picking me up from the airport or getting my favorite food to make me immediately feel at home.

The last part that stood out from my childhood was people being treated like an adult. I was always included in the planning of things and was asked to help with all the chores from a very early age on. That could have meant helping carry food home from the market or always washing the dishes. Weirdly enough, I developed a strong appreciation for washing dishes as a child. It always felt like an easy way to help out my grandparents after cooking and it has stuck with me to this day. I often still do the dishes by hand, even though I have a dishwasher. Go figure… Another part that is related to responsibility was also my parents trusting me with tasks that were not trivial. I remember my grandfather letting me switch gears on a two-hour car ride from Sillamäe to Tallinn just because I was so curious to learn how to change gears. The same kind of trust also carried over to my school years where I was never asked to show my grades. I was just trusted to take care of it and in a weird way, it actually motivated me to push even harder at school to excel. When you are trusted with something, you are very likely to do your best not to screw it up.

It is always a smooth continuum but it’s quite clear which motivations dominate the action in each part of life. The second part between the teenage years and the end of the twenties can be characterized by trying out a ton of stuff and seeing what sticks. It applies to both careers and personal relationships. Neither are linear learning curves where you can make steady progress. Using an analogy from machine learning, this part of learning is similar to gradient descent with a very steep learning rate to avoid getting stuck in local optima. I am sure many people do get stuck in local optima who do not take enough risk and experiment enough with a wide range of topics. This also seems to be the main reason why so many people experience personal crises realizing they have been climbing a path that is not making them happy.

For me personally, both my school and university time was a time of great professional exploration. I was always someone endlessly curious and choosing a professional path was challenging for me. There were so many things that were exciting and I could have imagined picking up many different professions. In high school, instead of doubling down on one subject and excelling at it, I decided to try to compete in competitions in pretty much every subject that I had. I remember taking part in 11 different olympiads in the 11th grade, which was clearly an overkill but helped me push my curiosity in many different directions. The same applied to choosing a university. Since I was not sure where I wanted to study next, I decided to apply to every university which sounded exciting. I ended up applying to a total of 20 or so universities in every part of the world. It was honestly a total coincidence that I ended up at Imperial studying computer science. I was very close to choosing mathematics at NYU or applied mathematics at the University of Singapore. Looking back at that choice, university selection definitely has a huge impact on the rest of our lives, but perhaps not in obvious ways. It is much more likely that I would still live in the US had I decided to study there, but I am not sure whether that life path would have made me any happier than the one I took. The same applies to what we choose to study. I am very glad that I ended up studying computer science. It definitely worked as a badge of technical capabilities, but I am rather sure I would have ended up having an awesome career also had I studied political science or philosophy.

Having said that your field of study does not affect your professional outcome much, it is rather important what type of people you surround yourself with. That applies especially during university and work. It is important to surround yourself with people who are curious and ambitious. I noticed that while studying at Imperial I was surrounded by people who were very driven, but all in the same direction. Everyone wanted to achieve the exact same outcome. Most of my coursemates saw being a senior engineer at Google as their ultimate professional goal. It also affected me and even though my gut feeling told me that this would not be my dream, the social pressure around me made me at least want to try out working at such prestigious companies. I tried working at an investment bank and I tried working at Google and only after having gotten this experience did I realize that this is not my dream. I don’t regret having tried those things, but I do realize looking back how strongly we are affected by the ambition of the people around us. I am sure that had I studied at Stanford, I would have been much more likely to stumble upon startups faster. To summarize, be aware of what people around you look up to. You might value the exact same things, but it should be a conscious choice, not just a mimetic desire.

Even though our professional desires change during our careers, there are things that become more clear as we have more work experience. After finishing my master's studies at ETH Zurich, I worked at five different companies, ranging from sizes of 10 to hundreds of thousands of employees in a number of different industries. What I had not tried out was starting my own company. So I started a company called Mirage straight after graduating. I started the company with one of my colleagues from my research lab and we were convinced that we wanted to get on the machine learning hype and build something in this space. Looking back, it was a bad idea to start a company for the sake of starting a company, but it did teach me two important lessons. I was very happy being a founder and having responsibility and control over my work environment. I also learned that I don’t really enjoy working on deep tech products where most of the work is in research and not really visible to the end user. That led me to found Acapela, which was much more user-facing and where the key differentiating factor was UX and the product, not the tech.

During the two years that I spent working on redefining what asynchronous collaboration should look like, I realized that I enjoy doing design work much more than engineering work. I started the company as the technical founder and spent the first few months mostly building and hiring. It was quite clear to me that I did not enjoy spending most of the day coding. Speaking to users and iterating with potential UX solutions came much easier for me. As the technical team grew and I mostly moved to a management role, I started to enjoy the product work even more. Jamming about potential feature implementations and UI details with engineers was a lot more fun than the actual implementation. It took us a very long time to find a solid designer, which forced me to spend more and more time on the actual UI of the product. It was definitely also possible because we had a very senior and independent engineering team, which I am very proud of. Over time I ended up spending more and more time working on UX and UI as opposed to technical discussions. I would have never considered myself a designer before (and I still don’t), but I grew to love the actual hands-on work so much that I naturally gravitated towards going into the details of design and being more high-level in engineering discussions. This is perhaps just to prove the point that despite having studied and practiced engineering, I discovered that I like working on products and design even more. This helped me understand even better how much I care about building the best product and also how much fulfillment I get for being recognized for building stunning and delightful products. This will have a big impact on what I will also build next.

I now feel like I am getting much closer to the third stage of life, where I have a higher confidence about which “hill” to capture both personally and professionally. The definition of a hill in the career sense here can be rather vague, but for myself, I have figured out that the rough criteria are:

1. Working in a high-growth technology company

2. Work in a product-focused role combining design, product, and engineering

3. Have a highly independent role which allows me to alter the product roadmap and company culture

4. Being surrounded by people I admire and can form close relationships with

5. Building a product where UX is a key selling point

6. A work environment that combines both remote and in-person work

In addition to having some quite set criteria, there are also lots of points where my level of confidence is much lower but I still have some gut feeling about what those things could be. For example:

• Work towards giving non-technical people capabilities that currently only engineers have (increase the number of builders in the world).

• Enable individuals to start taking more health decisions into their own hands by giving them suggestions based on aggregated biometrical data from their smart devices.

You might have noticed that the points in the second category (low confidence level) can be opposed to each other. That’s why they are low confidence ;)

The same principle should be applied also to other areas of life besides career choices. For example, for me, it holds true very much for where I want to live. I have lived in a bunch of different cities in the last 10 years and have gained a rather high level of confidence about which aspects of a living environment are most important to me personally. For example:

1. Have at least a few best friends living close to me (ideally within walking distance). That can obviously be built up in any city but is vital to me for happiness.

2. Have a high density of people around me that are ambitious and optimistic about the future. That strongly helps with finding friends but also being inspired by others around me.

3. Have plenty of sunlight throughout the year. I am definitely a climate refugee (originally from Estonia) and have realized that having sunlight even on a cold day just makes me a much happier person.

4. Be close to a body of water and ideally mountains. The older I get, the more I realize how important living close to water is for me. I am also a big hiker so having mountains close by is a huge plus for me.

5. There should be a wide variety of restaurants in the city that are at least somewhat authentic and also affordable so that I could visit them often with anyone I meet.

6. The local population should be somewhat tolerant of foreigners. I have always done the best I can to blend into the culture where I live, but in certain countries, it’s almost impossible to become “one of them” even despite your best efforts.

7. The local laws and values should put a high emphasis on tolerance, liberty, and progress. This perhaps covers some of the previous points, but there are clear examples where the local population is tolerant and progressive despite local politics (Berlin is a good example).

I have not found a place that would tick all of those boxes and usually, every location comes with strong trade-offs in at least one of those areas. So far in my travels, I can identify with and feel most at home in three locations where I have previously lived: Tallinn, Berlin, and Zurich. I am sure other candidates might fit the bill where I have not permanently lived yet (for example SF or Cape Town) but I have already found three strong candidates that make me feel very secure.

The same principles should also be applied in relationships. The older I get the more picky I become about which people I want to surround myself with. There is no good reason to do one business deal with someone who you would not want to do business with for the rest of your career. I also want to make friends who I believe at least in the moment that I would love to be friends with for the rest of my life. With more social experience it is also easier to rely on your gut feeling about people you meet - if you feel like something is off, then you can be rather sure that there is a good reason for that.

The biggest gains come from long-term commitments in every area of life: relationships, business, and skills. The third stage of my life is all about commitment and doubling down on the things that I love, value, and am potentially good at. That does not mean that there is no more room for exploration - just the space for exploration becomes more limited as we gain more confidence in ourselves.