Deep-tech companies die of organizational failure, not technical failure
How fast a startup can iterate determines how long it can keep winning — and iteration speed comes from "infrastructure" like procurement, hiring and performance review. Moving approvals up to the budget layer and replacing reviews with eigenreviews is worth more investment than agonizing over any specific technology.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Line-item approval doesn't control spending; budget buckets do
When he tried to personally sign off on every purchase, a $3,000 power supply was enough to stall the founder: should you wait three days for the auction and get it at half price? But you're burning $100,000 a week — waiting a week to save half the cost of a power supply is already wiped out by that week's spend. So he offers a rule: line-item approval is the wrong place to control spending; the review has to move upstream to the budget. Divide the money into buckets first and let well-paid employees make their own tradeoffs inside a bucket, rather than getting stuck on a $3,000 purchase.
— Max HodakExperiments look free because nobody has priced a single run
The hidden cost in hardware startups is that gases, resins and culture media get bought in bulk and then split across countless experiments; when nobody knows the true cost of a single experiment, experiments look free — they don't cost dollars, they cost "the media in the fridge." Science built Helix in-house for exactly this, wiring procurement all the way through to manufacturing and batch records, and only then could it calculate that each round of wafer iteration costs $40,000. His arithmetic cuts deeper: even if you raise $20 million in a Series A, with a 20-person team and 20,000 square feet, headcount is half the burn and rent eats another $750,000 to $1 million, leaving only about $3 million a year that can actually go into research — a few iterations and it's gone.
— Max HodakHR shouldn't screen résumés first; the whole company should vote
He turned hiring into a four-step funnel: the first step isn't HR screening résumés, it's the system picking seven or eight employees from across the company with similar backgrounds to vote, with options running from strong yes to strong no — which both prevents any one group from becoming a bottleneck and replaces a single person's taste with an average judgment. The actual funnel: 17% of applicants reach a phone screen; about half of phone screens get a take-home; onsite-to-offer conversion has to stay at 25% at minimum. The phone screen doesn't test specific skills, it tests three things: judgment, horsepower, and agency. And they try to design AI-resistant homework — for instance the minimum-cycle GPU kernel challenge Anthropic has used.
— Max HodakAnnual 360 reviews should be replaced by a monthly re-vote
The once-a-year 360 review, in his view, is a low-feedback, high-trauma ritual. The replacement is an internal mechanism called eigenreviews: every four to six weeks, randomly selected colleagues answer the same question — "knowing how this person has performed since, would you still vote to hire them today?" The answers are woven into a graph covering the whole company, then recursively weighted in the manner of eigenvector centrality: people the group endorses should have their votes count for more. To guard against vote collusion, each run does a thousand rounds of dropout, randomly removing a portion of the edges; if the scores come out bimodal, that suggests a voting clique may exist and warrants a manual look. Feedback lag drops from a year to about a month, and the annual HR process gets cancelled along the way.
— Max HodakWhat decides life and death isn't the technology, it's the weekly iteration rate
The whole conversation converges on one line: iteration rate is what separates the winners from the losers. If you can accumulate one useful lesson a week and your competitor manages one a month, that competitor will never really threaten you. Between two technical paths, even if the other one has quite a bit going for it, a markedly shorter iteration cycle is reason enough to prefer this one, because the compounding is that brutal. Hodak specifically notes that deep-tech companies rarely die because "the technology doesn't work"; more often they scale to a few hundred people and a few hundred thousand square feet without building the corresponding systems, so strategy can't connect to execution. He has seen too many teams of scientists and engineers with sterling pedigrees die on exactly that — execution failing to keep up.
— Max HodakAI and regulation are a better combination than people assume
Medical devices can't escape the quality system — a pile of standards granular enough to cover how a lithium battery plugs into a PCB, circuit board insulation, even whether the corners of a shipping-package label will lift in a vibration chamber. It used to take regulatory specialists finding every applicable standard one by one and building compliance evidence tables, routinely burning months; AI can search the standards comprehensively and generate the evidence tables quickly. Hodak's counterintuitive conclusion: AI and regulation are a better combination than people think — most regulation is "written in blood" and fundamentally reasonable; the hard part is only that humans are bad at executing it. He also stresses that a company should be AI-native: hand the agent all the context, quality records, batch records and the rest. AI's current role for a team is a multiplier, not a replacement.
— Max HodakEven long-horizon companies should chase revenue earlier than instinct says
"Some ideas are worth trying at $50 million or at $0, but you can't find out with $5 million." His fundraising advice for long-horizon companies: define the next key value inflection point, price the experiments you need to complete, then raise twice that amount and tolerate roughly 20-30% waste. More contrarian still: even wearing the deep-tech long-game label, you should pursue revenue earlier than instinct suggests. Hodak describes a company as beating "money cancer" back into remission one round at a time — but it shouldn't be that way forever. With revenue, a company gets valued on its long-term roadmap rather than its probability of survival, and it opens up a set of investors who could never have participated otherwise.
— Max HodakThe strongest organizations hold two kinds of people who look down on each other
Observing SpaceX from the periphery, he saw roughly 20% of employees were "committed Mars colonists" and 80% were serious engineers who thought those people were insane and just wanted to build the highest-performance methane engine in the world — and both cultures are indispensable. Science deliberately replicated this: 30% are openly transhumanist, 70% are clinicians and scientists who think the first group is crazy. Hodak believes the company can stand at the edge of the Overton window doing frontier research while also running clinical trials in six countries, obtaining European marketing authorization and publishing results in NEJM precisely because those two groups tolerate each other. To actually reshape the future, you have to be able to handle both temperaments at once.
— Max HodakIn their own words · checked verbatim
It is mostly not that we are smarter. It is infrastructure like this.
Max Hodak6:14
if you can learn one thing every week, and there's a competitor that's learning a thing every month, they will never matter.
Max Hodak18:46
whenever you exert action into the universe, that creates information, like in a very fundamental sense.
Max Hodak22:51
If you get them wrong, you'll end up like spending five million dollars a month and feel like you have very little control over it.
Max Hodak23:54
I think that the combination of AI and regulation is, is a better fit than people think.
Max Hodak37:18
I think that people should push for profitability sooner than they often think that they need to.
Max Hodak44:36
When you start a company in this space, you're committing to a decade of your life that you will never get back, no matter how it turns out.
Max Hodak56:00
Figures
| Hiring funnel: share of applicants reaching a phone screen | 17% | 12:32 |
| Hiring funnel: share of phone screens reaching a take-home | about 50% | 13:34 |
| Hiring funnel: minimum onsite-to-offer conversion rate | ≥25% | 14:38 |
| Runaway monthly spend when the infrastructure isn't built (figure of speech) | $5 million/month | 23:54 |
| Science's internal culture mix | about 30% openly transhumanist / 70% serious clinical staff | 52:51 |
| SpaceX culture mix (Hodak's observation) | about 20% committed Mars colonists / 80% serious engineers | 52:51 |
Glossary
- eigenreviews
- A performance-review mechanism that periodically collects peer votes and then weights them with a PageRank-like graph algorithm to work out who is genuinely trusted.
- AI-resistant homework
- Take-home tests designed so AI can't simply do them for you, with a clear optimization target, used to tell an applicant's real ability apart.
- vibe coding
- Writing software fast by collaborating with AI in natural language; this is what first made building internal tools in-house worth it for startups.
- tropicalization
- The European term for the surface encapsulation technology on an implanted device, which must survive long-term exposure to bodily fluids without harming the body.
How to listen
Deep-tech founders and early employees whose pace is being dragged down by procurement, hiring and performance processes; also useful for investors watching the brain-computer interface space.
If you don't care about neural interfaces, fast-forward 40:29 to 52:00; the culture argument at 52:51 at the end is worth pausing for.