For discussion only
Thesis
At the frontier, the idea is illegible. The bet is on the person.
A working memorandum on what top technical investors actually screen for in founders — the qualities that get funded once the business plan can't be trusted to predict anything. Drawn from the trajectories of Paradigm's Matt Huang and ByteDance's Zhang Yiming.
Technical skill is the price of admission, not the edge. Above it sit six repeatable signals — clarity, obsession, slope, agency, patience, and mission-alignment — and one hard distinction that separates a lifelong learner from a founder: the willingness to expose work to a real market. This note lays out the screen, the evidence, and the co-founder analogue.
Two slopes run in parallel. Only one of them capitalizes a company.
Consuming information at high volume is a genuine, fast-rising curve. It is not the curve that ships. The two are routinely confused, because the first is measurable, safe, and rewarded everywhere.
Learn · read · absorb · prepare
Measures how much is taken in. Rewarded by school, conferences, talks. No rejection, no failure — which is precisely its comfort.
- Papers read and dissected
- Courses and books completed
- Frameworks and models collected
Make · ship · expose · trade
Measures what changes in the world as a result. A specific person adopts it, pays for it, or walks — feedback that can't be argued with.
- Someone's behavior actually changed
- An open artifact strangers picked up
- A real user, not a benchmark, reacted
Technical skill is the ticket in. These are what get bet on.
The screen a leading technical investor applies to founders. Notably, almost none of it is innate — it is choice and habit, which is what makes it observable and testable.
Clarity
Thinking a hard thing down to the bone
Not talking the most — seeing the most. The strongest operators tend to listen; when they speak, they have already absorbed the full complexity. The first signal read in a great founder is extreme clarity about what, precisely, is being built.
In practiceThe test is compression: state, in one sentence, what is being built, why now, and why this team. An inability to compress is an inability to see clearly yet.
Obsession
Not effort — the inability to stop
Investors screen for obsession, not diligence. The distinction matters: disciplined effort can be summoned; obsession is fascination that won't switch off — nights that run together, not because they should, but because the problem won't let go.
In practiceLocate the problem worked on unbidden and late. Its absence is itself a signal — better to search for it than to grind on something merely tolerated.
Slope over intercept
Rate of growth beats current position
The stated preference is for the slope of a person's trajectory, not the intercept — the title, pedigree, or present level. It is what justifies backing an unproven 17-year-old over a decorated résumé.
In practiceCurrent level is a weak predictor; velocity over the next six months is a strong one. Optimize for the second, in others and in oneself.
Agency
Redefining what counts as hard
A working tenet at the frontier: the things people call hard usually aren't — they only appear so when one lacks control over the tools. The response is to build the missing tool rather than wait for it. Great teams teach themselves what has no textbook.
In practiceSeparate genuinely-hard from tools-not-yet-owned. It is almost always the latter. "Can't" resolves to "haven't yet."
Patience
Surviving back-weighted returns
The temperament suited to frontier technology is one that tolerates payoff arriving late — buying in the winter, planning for scale that looks absurd at the time. Judgment on long horizons, undisturbed by short-term noise, is the rarest and most trainable of the six.
In practiceDeliberately hold assets that compound but show nothing for months, and accept the early boredom. Move the evaluation window from the week to the year.
Mission over money
A floor, not a bonus
The most expensive lesson in the sector: technical brilliance without shared mission is a liability, not an asset. Alignment became an explicit filter afterward. Subtract the upside and the question is simply — would they still do this?
In practiceA screen for others and for oneself. Whoever performs only when rewarded will, at the decisive moment, choose the short-term win that costs the long-term one.
Two technical founders. One pattern beneath the surface.
Both engineers by origin — no accident; technical people back technical people. But the technical layer is only the entry ticket. What was bet on sits above it.
Matt Huang
Investor
- Firm
- Paradigm
- AUM
- ~$12B
- Role
- Co-founder, MP
- Education
- MIT, Math
- Prior
- Sequoia Capital
- Signal bet
- ByteDance ~10,000×
- Model
- Research-as-investing
A mathematician entering crypto wasn't a leap — it was a homecoming. The field sits on three pillars he was raised at the intersection of: cryptography (math), incentive design (economics), and distributed systems (technology).
What looked illegible to others was legible to him — which is why he could underwrite conviction when few peers could. The discipline that follows is contrarian by construction: commit capital in the winter, when the signal-to-noise is highest and the price is lowest.
He backs the person, not the idea. On the signal bet, he judged the idea likely to fail — but read, through a translator, someone unusually capable, obsessed, aggressive yet balanced, with extreme clarity and total ambition. The return was on the order of 10,000×.
His model reframes venture itself: what if a firm didn't merely back the category winner, but built it? Its researchers are investors — inventing mechanisms and open-sourcing infrastructure that much of the industry now runs on. He describes the firm as an academy for unusual technical talent: recruit the exceptional, then build around their strengths rather than force standard roles. He is not the one who writes the papers — he is the one who can read every paper in the room and judge which research, and which person, is worth the bet.
Zhang Yiming
Operator
- Company
- ByteDance
- Products
- Toutiao · TikTok
- Origin
- SW engineer
- Education
- Nankai University
- First role
- 1st eng., Kuxun
- Core trait
- Long-term patience
Genuinely technical — a programmer by training and early career. Known in his university's technical circles; first engineer at a travel-search startup, running a 40-person team by his second year.
But a pure coder does not become the founder of a company at this scale. What he stresses in his own account is not code — it is patience: long-term judgment, undisturbed by short-term factors, waiting for the designed outcome to arrive.
The connective thread: a technical investor backs technical founders, but reads for what sits above the technical layer. This operator's self-description — patience, long-term thinking, obsession, mission over money — maps onto the six-signal screen almost line for line. Technique gets a founder seen. The rest gets them bet on.
Ashley Smith
Comparable · Solo GP
- Fund
- Vermilion Cliffs
- Vehicle
- Fund II · $25M
- Edge
- GTM operator
The other valid modelOpposite scale — a $25M fund against a $12B platform. Concentrated conviction and build-alongside on one end; an operator's edge (go-to-market for developer and security tools), smaller checks, wider portfolio on the other. Different shapes, one shared floor: capital alone is insufficient — a firm must offer one concrete thing a founder cannot get elsewhere. Both back only technical founders. Every firm must answer the same question: why would a founder choose you?
The heaviest, least-reversible bet on the cap table.
A fund hedges a bad pick across a portfolio; a founder cannot. The wrong co-founder ends the company, and the separation is far harder than a layoff. The bar is therefore higher than for any founder the same investor would fund.
Mission alignment
The most expensive lesson, applied to a partnership. Subtract the money and the prestige — do they still want this? Only the true believer stays on the worst day. Evidence: what they have done when there was no reward.
See them fail first
Anyone partners well in good times. Go through one real failure and one sharp disagreement, and observe whether they break, deflect, or stay honest. Partner with someone you have carried weight with — not merely dined with.
Fill the avoided gap
The benchmark pairing is deliberately complementary and split evenly. The comfortable pairing — same instincts, same strengths — is the dangerous one; it charges into a shared blind spot.
Pick the fastest learner
Do not be dazzled by the résumé. Two static people cannot drive a company through violent change; a slope person needs a slope person beside them.
The same four criteria a good co-founder uses to screen the other side. The move is not to hunt for someone who fits them — it is to become someone who fits them first.
Every signal above stays inert until it meets the arena.
Founders are not made in the practice room. They are made in the arena — work exposed, in front of a real market, with a real chance of rejection and no do-over. Preparation is necessary and infinitely postponable; the exposure is the single act that converts a lifelong learner into a founder. Below is what one real instance of it requires.
A real counterparty
Who
Someone with an actual problem the work can touch — not a reviewer, not a peer, not a benchmark.
Minimum viable exposure
What
Not the finished product — the smallest version usable tomorrow. Then listen, and say nothing.
A commitment date
When
"Soon" is preparation. The moment it has a date, it becomes the arena. Exposure only counts once it is real.
Primary references.
Note on sourcing
Certain biographical details — family background and the specifics of the signal investment — derive from a single profile and one podcast interview, and should be read as reported rather than independently verified. Links above are provided for that purpose. This document is a study note compiled from public sources; it is not investment advice.