Commit Early, Commit Often

Earnest Sweat — Sun, 27 Sep 2026 07:11:32 GMT

Founder Signing Day. Invented firms, no real logos.

In February 2010, a seventh grader named David Sills V gave a verbal commitment to play quarterback at USC. He was 13. Lane Kiffin offered him a scholarship after watching video sent over by Sills’ private quarterback coach. I was in my twenties when stories like this started making the rounds, and I remember thinking what everybody thought. What are we doing? He hadn’t played a down of high school football. He hadn’t hit his growth spurt, or maybe he had, which is somehow worse, because now you’re projecting a kid who might already be done growing. Nobody knew. That was the point. The offer wasn’t a bet on what Sills was. It was a claim on what he might become, staked before anybody else could stake it.

Here’s how it went. Kiffin was fired in September 2013, before Sills ever got to campus. The next coach honored the offer and then recruited other quarterbacks anyway. Sills decommitted in 2014, went to West Virginia, and got moved to wide receiver. He spent a season at a junior college trying to stay a quarterback, came back to West Virginia as a receiver, and in 2017 he was an All-American who tied for the national lead in touchdown catches. He went undrafted and has spent the years since fighting for NFL roster spots. The talent was real. The position was wrong. And the coach who made the offer was gone before the kid showed up.

I’ve been thinking about Sills since a call this week with a friend who runs a community for emerging managers and founders. I was telling her how desperate this market has gotten to find founders before they even know they’re founders. She’d been hearing it too. In the last month alone, she’d heard three times from groups thinking about starting an academy, and one of them was looking at refurbishing an old university campus, dorms and all. Then she said the line I haven’t been able to shake: “I almost feel like they’re scouting the womb now.” I asked if she remembered when middle schoolers started getting football offers from Alabama. She said, “And he’s 14.” I said sometimes 12. Then I told her I was going to have to write about it, so here we are.

Nothing is proprietary

In my essay, The Souls of Seed Folk, I mentioned a founding partner at one of the large platform firms told me maybe 90% of venture dollars now chase momentum, and maybe 10% of the business still does the old craft of finding and building companies. This essay is about the 10%. I used to think of the craft as the quiet corner of the business, the place where patience and relationships did the work. It isn’t quiet anymore. It’s a knife fight, and the biggest firms in the world walked into it carrying the biggest knives.

The reason is simple. Proprietary deal flow, at almost every stage, isn’t a thing anymore. Nothing is proprietary. Like, literally nothing. You can find people very easily. The only question left is whether you can find your exceptional people, judge them well, and get them to take your money.

The tools did this. Harmonic says it tracks more than 195 million people and covers companies from the moment founders file for incorporation. Its own blog walks through how to spot a stealth founder: somebody leaves an AI lab or another high-signal employer with no announced next role, changes their LinkedIn title to “stealth” or “building something new,” registers a domain, files the paperwork, or waits out an earn-out after an acquisition. Specter sells more than a thousand talent signals a day to over 400 firms. Evertrace watches incorporations, domains, GitHub, and patents. Harmonic even named its AI agent Scout. The scouts didn’t get to keep the job title.

Harmonic's guide to finding and tracking stealth founders, June 2026.

When everybody has the same radar, the radar stops being an edge. Konvoy said it plainly in a January piece titled “Deal Flow is Now Table Stakes.” In a meeting with an LP and I landed in the same place: the edge now is finding founders before they’re visible to the AI sourcing systems. Which means before they leave. Which means before they decide. Which means, if you follow the logic all the way down, before they know they want to be founders at all. We used to play Who Wants to Be a Millionaire. The new game show is Who Wants to Be a Founder, and the host is hoping you don’t know the answer yet.

The recruiting board

Here’s what the recruiting looks like right now. I’m going to use sports analogies, because that’s what it feels like.

The academy. On September 22, Andreessen Horowitz announced the Horowitz Andreessen Academy, a private, full-time, in-person school in San Francisco for young people coming out of high school. Admission runs on proof of work, meaning what you’ve built, shipped, or earned, not your grades. It’s unaccredited, so no degree and no transferable credits. The founding class is about 50 students, free, starting in fall 2027, with $50,000 in compute credits each and co-ops at companies like Anduril, Anthropic, Nvidia, OpenAI, and Stripe. The school raised $42 million in a round led by a16z, and the announcement promises that Jensen Huang, Satya Nadella, and Patrick Collison will come by to lecture.

a16z's announcement of the Horowitz Andreessen Academy, September 22, 2026.

I have to be honest about my first reaction. It reminded me of the Osborn Academy. If you missed the 2017 Spider-Man cartoon on Disney XD, here’s the plot. Norman Osborn’s son gets suspended from the genius high school in town, so Norman opens his own genius school at the top of Oscorp Tower, holds auditions, and sets it up as the rival. I’m not saying anybody in this story is the Green Goblin. I’m saying I’ve seen this episode. And to be fair to a16z, they called their shot. Their Big Ideas list last December predicted an AI-native university. Most firms predict things. They’re opening one.

The camp. My friend Aditi Maliwal, a GP at Upfront, posted a few days ago that applications are open again for the Upfront Founder Fellows. It’s eight weeks for technical students, recent grads, and dropouts. Fellows get one-on-one mentorship with an Upfront partner, a seat at the Upfront Summit, fireside chats with portfolio founders, a standing invite to dinners in LA, New York, San Francisco, and Boston, and $75,000 or more if they incorporate a venture-backable startup. Her post opens with the idea every recruiter eventually lands on, that the best ones don’t wait for permission. She’s right, by the way. That’s the problem. Everybody is right about that, so everybody is standing outside the same gym.

Upfront Founder Fellows, second cohort.

The camp circuit runs deep. The Thiel Fellowship pays $250,000 over two years to people 22 and under who leave school. Z Fellows gives $10,000 for a single week and takes high schoolers. Neo will give a college student $40,000, no equity, to take a semester off and spend it in San Francisco. South Park Commons writes $400,000 for 7% into people at what it calls the “-1 to 0” stage, before there’s an idea. Kevin Hartz said last fall that close to 20% of his bets over the prior year went to teenagers, up from about 5% two years earlier.

The early decision. This one made me laugh, because it’s the verbal commitment with better branding. Y Combinator’s Early Decision tells students to apply now and do YC after they graduate, and they get funded the moment they’re accepted. That’s a high school junior telling Kentucky basketball he’s coming after senior year. The difference is YC actually wires the money.

Y Combinator Early Decision. Apply now, do YC after you graduate.

The foundry. Afore has an open position for its foundry program right now. Its Founders-in-Residence page doesn’t bury the lede: People Before Ideas. It runs two months, typically invests $250,000 to $750,000 with a floor of $100,000, and takes about ten teams per cohort, with cohorts starting in July, October, and February. Afore isn’t alone. Eclipse just raised $1.3 billion and plans to use part of it to build physical AI companies in-house. General Catalyst has a whole creation practice. Greylock Edge takes founders before they have an idea. The foundry is the prep school. Bring the kid in early, house him, coach him, and make sure he’s wearing your jersey when it counts.

Afore's Founders-in-Residence page.

The beacon. Then there are the firms posting their position needs. Pear publishes Requests for Startups, and its August edition came from operators instead of partners. Ali Ghodsi at Databricks wants an agent that interviews employees to capture what a company knows. Tomer London at Gusto wants an AI app builder for kids. Y Combinator’s fall list has 13 ideas, including one on the future of American defense written by the Secretary of the Army. In The Souls of Seed Folk, I told the story of a friend who put a beacon out on a problem he’d been chewing on since his operating days, and a few operators from a big marketplace company answered and came to hang out at his firm for a month. Same move, smaller megaphone. A request for startups is a coach telling every high school in the state, we need a left tackle.

Pear's August 2026 Request for Startups, written by operators.

The radar, hand-built. The friend from that March lunch has his own version of the radar, and his is smarter than most. He goes back to last cycle’s hot companies, the ones that were rocket ships and then kind of stalled. The people there are still great. They were in the building for the inflection, and they learned from other people’s mistakes on somebody else’s dime. Nobody’s looking at them, because everybody is poking around trying to get anyone from the two or three hottest companies. He also told me most firms do references so poorly they can’t tell who was the reason a company inflected and who just picked a good rocket. Then he said the line I wrote down twice: “You’ve gotta find people who don’t know they’re pretty.”

The same friend who said “targeting the womb” told me about a firm that has built what sounds like a heat map of founders coming out of universities. She also told me about a GP who runs his fund in two buckets, core checks and talent checks. The talent checks go to university kids. He loves the dropout, and not just the kid building apps in his dorm room, the one in the engineering program. He’s in a handful of companies where, as my friend put it, 21-year-olds are running $3 billion businesses.

The street agent. Every recruiting system has a shadow version. My friend told me about an investor, she knows who bought memberships to all the major coworking spaces. He’d go on Saturday nights to see who was still working, tell everybody he was a scout for one of the big-name firms, and befriend the people grinding around the clock. Then he’d coyly switch it up. Oh, I was scouting for a friend over there. I’m actually with this other fund. That’s how he found the hottest deals. It’s shady. It also worked, which is the part that should scare you.

The scout used to be the respectable version of all this. Sequoia started its program around 2009 and gave operators and founders money to write small checks into companies they found, which is how scouts ended up early in companies like Uber and Stripe. The idea was to have other people out there doing our jobs for us, getting our name around, keeping our tentacles in rooms we weren’t in. That was proprietary deal flow back when it existed. Now the scouting report is a SaaS subscription, the tentacles are an API, and some of the people still doing it in person are lying about who they work for.

Does the analogy hold?

Mostly. And the places where it breaks are the interesting part, so stay with me.

It holds on commitments. A verbal doesn’t bind anybody. Sports Illustrated tracked 500 top-100 recruits from the 2007 to 2011 classes and found that 14.6% decommitted. In the 2024 class, it was roughly 19% of Power Four (the top football programs) commits. Nobody publishes the decommit rate for fellowships and residencies, but I promise you it isn’t zero. A fellow is eligible for a check. A founder in residence is invited to stay. The only thing anybody signed was paperwork for a company that doesn’t exist yet.

It holds on the coach leaving. Sills lost the coach who recruited him. Tate Martell committed to Washington at 14, watched Steve Sarkisian leave for USC, and bounced from a Texas A&M commitment to Ohio State, Miami, and UNLV before retiring in 2022. Michael Avery committed to Kentucky basketball as an eighth grader in 2008, Billy Gillispie got fired, the new staff pulled the scholarship, and Avery played Division II at Sonoma State. None of that is on them. They were kids. If you’ve raised money in this business, you know the story from the other side: the partner who championed your deal leaves eighteen months later and you’re the orphan on the portfolio page. Now imagine the partner who recruited you at 18 into an academy leaves before you ever start the company.

It holds on the target. Being the best 12-year-old is hard. Staying the best 13-, 14-, 15-, 16-, and 17-year-old while everybody knows your name is much harder, because every camp you walk into has somebody trying to prove he’s better than the kid with the offer. Founders are going to feel this. The kid who gets into the academy at 18 is going to have a scouting report on her before she has a customer.

It even holds on the data, and the data cuts both ways. On3 looked at more than 16,000 recruits across five NFL drafts. Five-stars got drafted 63% of the time, four-stars 22% of the time, and three-stars about 5% of the time. So the early rankings aren’t nonsense. Pedigree predicts. But three-stars still made up 40% of everybody who got drafted, because there are so many more of them. The hit rate lives at the top. The volume lives in the middle. If you’re a Series A investor, you should read that twice.

The all-star games tell the same story. ESPN looked at McDonald’s All-Americans from 1993 to 2012 and found that 42% played fewer than 25 NBA games and only about 8% made more than one All-Star team. So when people ask whether tech should have its own McDonald’s All American game, I have news. We have one. It’s called Forbes 30 Under 30, and its alumni include Martin Shkreli, Sam Bankman-Fried, and Charlie Javice. Forbes had to publish its own Hall of Shame. Demo day is the dunk contest.

Here’s where it breaks. When Alabama offers a 14-year-old, everybody knows what sport he’s going to play. Football exists. The field is 100 yards. There’s a depth chart. When a firm recruits a pre-founder, there’s no sport yet. The market the kid is going to build in might not exist on the day she signs. That’s why the requests for startups exist. They’re an attempt to hand the recruit a position before the season starts. But Sills is the warning. The talent was real and the offer was for the wrong position. Most of the programs on the recruiting board are recruiting quarterbacks. A lot of those kids are going to turn out to be receivers, and some of them are going to turn out to be the best offensive coordinator you’ve ever met.

It breaks on the calendar, too. An athlete has a clock. You’re a high school senior, then a freshman, then draft eligible, and then it’s over. Founders don’t have a signing day, and nobody ages out of starting a company. The best founder I back in the next five years might be a 42-year-old who’s been an early employee three times.

And it breaks on the referee. College sports got so crazy that the NCAA eventually stepped in and pushed the contact date back. In 2017, Division I lacrosse banned recruiting contact before September 1 of junior year, because eighth and ninth graders were committing. By 2019, most Division I sports couldn’t start recruiting conversations until the summer after sophomore year. Venture has no NCAA. Nobody is pushing the contact date back. a16z just moved it to the day after high school graduation.

NIL and the portal

The part I didn’t expect to find is what happened after the scholarship offers. The analogy doesn’t stop at the eighth grader. It keeps going, and the next two chapters might be the most useful ones for us.

The first chapter is NIL. In June 2021, the Supreme Court ruled 9-0 against the NCAA in Alston, and on July 1 the NCAA let athletes get paid for their name, image, and likeness. Within a year, boosters had pooled their money into collectives, and the collectives became pay-for-play with extra paperwork. In the 2023-24 school year, collectives accounted for roughly 82% of all NIL pay. Last year the House settlement let schools pay athletes directly, up to $20.5 million a school, and the whole market is now estimated at around $4.5 billion a year. And yes, it went younger. Nike signed a 13-year-old soccer player in 2024. A 14-year-old football player in D.C. hired an NIL agent the summer after eighth grade.

I wanted to know whether the early offers led to NIL. Not directly. State legislatures, antitrust lawyers, and the Supreme Court did that. But the early-offer era proved how much the talent was worth to everybody except the talent. The money was already moving. It was just moving under the table. When the FBI arrested ten people in the college basketball case in September 2017, including four assistant coaches and an apparel executive, one scheme routed money through middlemen to recruits’ families in exchange for commitments, and the other paid coaches to steer players toward advisers. The NCAA’s own commission said the rot started in youth basketball. NIL didn’t create the market. It moved the market above the table.

Venture already has NIL. We call it grants, stipends, and compute credits. Paying the name before there’s a product is exactly what $50,000 in compute, a $40,000 equity-free semester, or $250,000 to leave school is. You can see the pricing in the terms, too. Z Fellows’ $10,000 can go in at a $1 billion cap. That isn’t an investment so much as a recruiting pitch. The lesson from sports is who wins once you can pay the name directly, and the answer is the biggest collective. In venture, the biggest collective is the platform firm with a balance sheet big enough to fund a school. My friend’s advice to the smaller firms thinking about academies was the right one: combine forces, pair up with the other ten groups, and do it together. That’s a collective. I’d be surprised if we don’t see one by next year.

Am I too old for an NIL deal? I checked. Texas only allows high school NIL starting at 17, and nobody has written the rules for a guy with two podcasts and a third on the way.

The second chapter is the transfer portal, and it’s the twist. The portal launched in October 2018, and in 2021 the NCAA let first-time transfers play right away. Men’s basketball went from 957 portal entries in 2019 to 2,320 in 2025, more than 40% of all players. More than 10,500 college football players entered during a single two-week window this January. Then something happened that nobody in the scholarship-for-seventh-graders era saw coming. Once schools could pay players and players could move freely, a lot of the money moved later, not earlier. College basketball coaches started saying freshmen were a worse investment than proven transfers. Going into July 2025, only nine players in the ESPN 100 for the class of 2026 had committed, down from 17 a year earlier.

Look at that from a venture seat. When the money got real and the movement got free, the smart money went to proven production. Not to the eighth grader. To the player who already did it somewhere else and was ready for a new jersey. That’s what my friend’s hand-built radar is. It’s the transfer portal. It’s also the part of the market I’m built for. I’m a Series A investor. I don’t get to take the 14-year-old to dinner, and by the time a company gets to me it’s already been recruited three times. My whole job is to know who was actually on the floor when the thing worked.

Who coached them?

Here’s the question I keep coming back to. If everybody is chasing the phenomenal amateur, who’s getting left behind?

Go back to Sills one more time. Lane Kiffin didn’t discover him at a camp. He saw video from Sills’ private quarterback coach. The coach was the node. In basketball it’s the AAU coach. In football it’s the trainer, the seven-on-seven coach, the high school coach who’s sent 30 kids to Division I. It’s the parents, the uncle who drove to every tournament, the teacher who gave him a key to the gym. Recruiting has always run through the people around the athlete, and once real money showed up, a lot of the power moved to them. The FBI case proves that more than I’d like it to. The influence was real enough that people broke the law to buy it.

In venture, the people around the founder are the early employees. The first 50 or 100 people at great companies are the coaches, the teammates, and a lot of the future founders. They sat next to the person who became the CEO. They know who was the reason the company scaled and who just picked a good rocket, which is exactly the reference my friend says nobody does well. They’re who founders call when they need a head of sales who’s done it before, an advisor who’s lived through a down round, or a board member who’s actually shipped something. And a lot of them, to borrow the line, don’t know they’re pretty. Some of them should start companies. Many of them shouldn’t, are exceptional at what they do, and have never had a place in the founder story. Everybody wants to call their firm a founder magnet. Magnets work on what’s close. The early employees are what’s close.

That’s Exceptions, and I want to be explicit about it here. Exceptions is the third podcast show I’m building, after Swimming with Allocators and Carry On, and it comes with a dinner series. The podcast is about the early employees of historically great venture-backed companies, the people who rarely get interviewed, in a handful of deep, researched episodes a year. The dinners are small, eight to twelve people, half pre-seed and seed founders and half standout early employees. The founders get a room with the people who could be their advisors, board members, and first great hires. The early employees get a room with the founders worth betting on. And I get what every recruiter wants and nobody can buy from a database: signal on where the best operators land next, who’s about to spin out, and a back channel on who was actually on the floor.

If everyone else is handing out scholarships to fifth graders, I want to know the AAU coaches. That’s not a playbook. It’s where I think things are going, and it’s the one part of this knife fight where being the most analog person in the room is an advantage. I’m looking for people to co-produce and build it with: operators who want to host, studios and producers who care about this story, and companies and firms that want to be around this community early. If that’s you, reply to this or find me.

What are we building?

I don’t want to pretend I have this figured out. I have more questions than answers, so I’ll leave you with the ones I can’t stop asking.

If a firm finds you at 18, puts you in its school, places you in a co-op at a company in its network, hands you its list of ideas, and funds your first company, whose company is it? I don’t mean the cap table. I mean the ideas. When every academy reads the same requests for startups, we might get a generation of founders who are very good at building what they were recruited to build. That’s not the same as building what only they could see. A request for startups list asks for an AI app builder for kids. There’s a version of the future where the kid who grows up on it is sitting in an academy in 2035, building the next thing on somebody’s list.

What happens to the kid who says no? In sports, the kid who decommits gets called disloyal. In venture, the pre-founder who turns down the fellowship, joins a startup as employee number twelve, and learns the job might be the best founder of the decade in 2032. Are we going to recognize her then, or is she a three-star forever because she didn’t take the offer at 19?

And what did we give up on the way here? For a decade this business said founders first and founder friendly. That was a promise about how we’d behave after the check. Commit early is a promise about how fast we’ll find you before it. One is about the relationship. The other is about the claim.

Commit early, commit often is good advice for code. You save your work in small pieces so you can roll it back when something breaks. I’m not sure it’s good advice for people, because you can’t roll back a 19-year-old. So I’ll keep doing the thing I’m good at. I’ll keep asking who coached them.

With gratitude,
earn

GROUNDWORK: Preferred Equity Research