How to Spot Exceptional Talent Before Everyone Else

Summary

• The a16z crew argues that peer endorsement can function as a powerful credential alongside institutional ones. Erik Torenberg shares David Perell’s line that his Twitter account is worth more to his career than his college education; David Booth’s thought experiment asks what is more impressive: Stanford attendance or being followed by Patrick Collison, Elon Musk, and Marc Andreessen. Most people, he argues, would pick the latter, and Cosign aims to productize that “peer-to-peer credential” as a free directory of people and companies for the startup community.
• Olivia Moore’s investor case is the episode’s strongest commercial claim: “the most valuable data in all of Silicon Valley is who has conviction in whom.” a16z investors still ping people daily asking whether they know someone who worked with a candidate. Someone evaluating five startup offers usually cannot access that signal, even though it could mean the difference between joining a rocket ship and a company that is not doing as well.
• The AI timing thesis is that as AI-driven outreach and interest scale, human endorsement becomes the scarce asset. Olivia posted one Google Form for portfolio hiring and got 2,000 responses — “it’s gonna be 10,000 a month from now” as browser usage improves, “it’s gonna be 100,000” — so “the scarce thing is the human endorsement or the human conviction.” AI also helps address the cold-start problem: Cosign assembles Wikipedia-style profiles from public data rather than waiting for sign-ups.
• Josh Elman, who worked at LinkedIn in 2004 and 2005, delivers the incumbent anatomy: Reid Hoffman’s “growth, usage, revenue” sequencing built a powerful long-term network effect on a single degraded signal — “are we connected?” The trusted graph decayed into collecting (“we met once at a party, we high-fived”), while a population of more than a billion users, including many “mildly happily employed” people, makes LinkedIn difficult to displace head-on. Olivia adds that switching costs are reinforced by LinkedIn’s “famously good” bot detection. The opening is granularity — did you manage them, work on the same team, or work there at the same time — which LinkedIn treats as one flat connection.
• The product’s deliberate constraint is positivity-only, which the team frames as both wedge and quality filter. Erik says “this product is gonna win on people using it to show off,” and Olivia notes endorsements self-police because “if you endorse someone who’s not great, that reflects poorly on you.” Private signals — “I’d fund them,” “I’d hire them,” or “I’d buy this company” — can match people on mutual interest while reducing the social cost of asking.
• David Booth’s talent-spotting track record is part of the pitch: seven years ago he emailed hundreds of people for their “person to watch”; one answer was Russell Kaplan, then a 22-year-old Tesla engineer and now president at Cognition. The closing argument stacks the canon: Paul Graham cosigning a pre-exit Sam Altman around 2009, Ryan Hoover cosigning Erik, and Dwarkesh, whom Erik says was 25 and had never had a job as far as he was aware. The claim is that if someone is talented and smart, a respected person can discover, cosign, and accelerate their career.
• Erik answers the alpha-hoarding critique with a long-game argument investors should note: don’t cosign only the founder raising now; cosign the one raising two or five years from now. When that founder eventually raises, Erik says, they may remember who believed first and call that person first. His kicker: people remember, “by an order of magnitude more,” the people who believed in them first — as illustrated by Brian Chesky crediting YC and Paul Graham in a post when Airbnb went public despite much larger later checks.

Deep dive

1. Cosign’s premise: the endorsement can be a credential

• Erik Torenberg opens with his friend David Perell’s line — “My Twitter account is more valuable to me than my college education,” despite “a quarter mil” in tuition — and unpacks college as a bundle of education, network, credential, and, Josh adds, social life. Twitter can supply parts of the education and network, but the credential piece is where Cosign aims.
• David Booth’s framing of the mechanism is that a follow from someone credible can be an endorsement given off as “exhaust” — “this person’s not even trying to give them reputational benefit… and the exhaust is this endorsement.” His test: Stanford degree versus being followed by Patrick Collison, Elon Musk, and Marc Andreessen — “most people would say the latter.”
• The giver benefits too: Silicon Valley is “all in a little bit of a race to discover talent,” and being early means hiring and funding people first. Angel investing already expresses a cosign, “but a lot of people aren’t founders, and a lot of people aren’t investors.” Erik frames the company-endorsement version as “almost like fantasy investing or social capital investing.”

2. Three endorsement types — and proof the person-to-watch signal works

• David wants every profile to have at least three rank-ordered endorsements: who shaped your career — the “mentor economy… what Silicon Valley runs on”; who you’d “go to war with,” or work with in the trenches, which is much more granular than a follow; and a person to watch, the potentially ascendant pick.
• David’s receipt for the third signal: “I did this experiment seven years ago where I emailed a few hundred people… for their person to watch.” One answer was Russell Kaplan, now president at Cognition, who was then a 22-year-old engineer at Tesla.
• David says the mentorship signal compounds both ways: if Josh Elman shaped him, Josh is putting his reputation behind him; “if you respect Josh, you respect me by association.” The mentor also benefits by becoming known as someone who discovers and helps talent.

3. Olivia’s thesis: conviction data is the Valley’s most valuable — and least accessible — dataset

• After investing for almost a decade, Olivia Moore believes “more and more every week” that who has conviction in whom drives investing, hiring, and job choice. Yet “there’s such information asymmetry” that even a16z investors work daily to extract it by asking people whether they know a candidate. Someone evaluating five startup offers usually cannot access that signal, even though it could mean the difference between joining a rocket ship and a company that is not doing as well.
• Her prior art: she and her twin sister Justine, who works on a16z’s infrastructure team, ran a Sunday newsletter for six and a half or seven years curating startup job postings. They had gone to Stanford and run a startup incubator there, but still had no way to distinguish which founders and startups were legitimate or exciting. Campus scouts read the newsletter, and some later became founders of companies Andreessen invested in. Cosign is the more scalable version of that network.
• The team’s product examples include Packy McCormick’s public list of 37 companies in deep tech and vertical integration — “he’s cosigning some of his reputation against that company” — and the decision to keep the product free, unlike Crunchbase and PitchBook, which Olivia calls great products that have to be businesses.

4. Private signals: matching intent without the social cost

• Erik’s model event is an SV Angel conference around 2018 that sent attendees a list of 150 people to swipe on. Mutual swipes triggered automatic introductions. “There are so many people in the world who we wanna meet, who also wanna meet us, and it’s just an information problem.” Cosign extends that model to granular professional intent: “I would fund them if they start a company,” or “I would hire them if they consider leaving,” with an introduction triggered when interest is mutual.
• Olivia adds the timing-discrepancy problem. Asking a founder “are you thinking about exiting your company?” can be offensive or insulting, but a two-sided network could match acquisitions, hires, and job changes “when the time is right.” Erik jokes, “We’ll get the product features shipped tomorrow,” as the roadmap expands.
• Josh agrees that matching needs and skills at exactly the right time is hard. Erik’s follow-on is that codified lists — the people who are great designers, crisis-ready PR operators, or trusted career mentors — let trust travel without depending on who is “fresh in your mind.” The introduction can still route through a trusted person rather than becoming an open directory.

5. Durable reputation, positivity-only, and recognition for everyone who is not the founder

• David says social-media reputation is ephemeral. When Josh joined the firm, longtime collaborators surfaced endorsements that were valuable but quickly disappeared into the feed. David’s own example is Marc Andreessen’s blog post about David joining the firm: he wants that durable association to be the first thing people see when they look him up. Erik adds that the product’s feed pulls thoughtful positive replies and quote-tweets into the relevant profiles.
• The positivity constraint is strategic, not squeamish: David says “products have to win on some dimension, and this product is gonna win on people using it to show off.” Olivia notes the self-policing economics — “if you endorse someone who’s not great, that reflects poorly on you” — while Josh frames launches as vulnerable moments deserving celebration; “there’s plenty of places for critique.”
• David’s RISE Awards, around 2015–16, were awards “for everybody else” — engineers, designers, and product managers. He says there were thousands of submissions and believes Josh was a judge; the program is intended to return in some form. Recipients also built lasting relationships through retreats and groups.
• Olivia’s sharpest version is that reference calls often reveal “this person who was engineer number four actually was a huge determinant in whether this company was successful or not,” yet that accomplishment is not portable or publicly legible. David’s summary is that Cosign makes the reputation market more efficient and liquid.

6. Josh’s LinkedIn anatomy: why the incumbent is iconic, degraded, and difficult to displace head-on

• From inside LinkedIn in 2004 and 2005, Josh recounts Reid Hoffman’s sequencing — “GRR: growth, usage, revenue” — and the “social résumé” epiphany: annotate an application with who at the company knows the candidate, who worked at similar companies, and who can help reach those people. Put the hiring manager’s profile on the job listing so it is no longer “throwing your résumé over the wall.”
• The trusted graph — “Olivia says David’s really good; will you take this intro?” — faded into collecting: “we met once at a party, we high-fived.” Olivia says that 90% of the time, an introduction request through LinkedIn produces “Hmm, I don’t think I know them,” because managing someone, working at the same company, working there at the same time, sharing a team, and meeting at a party are all flattened into one connection.
• Olivia says switching costs and network density explain why no LinkedIn killer has landed despite a16z having invested in some attempts. LinkedIn is also “famously good” at bot detection, and whole businesses are built around scraping its data. Josh says he thinks there are over a billion people on the network now, including many “mildly happily employed” people who are not actively searching but might be open to the right opportunity. That long-term network effect makes moving the masses difficult.
• Erik’s caveat is that disruption can come through generational shifts and hyper-niches — “GitHub is arguably a better LinkedIn for engineers than LinkedIn itself.” Josh insists Cosign need not beat LinkedIn: “if Cosign works great and it still only serves the tech community, that’s a huge win.”

7. The AI timing: infinite outreach makes human conviction the scarce resource

• Olivia posted one LinkedIn Google Form to help portfolio companies hire and got 2,000 responses — “it’s gonna be 10,000 a month from now as browser usage gets better… it’s gonna be 100,000.” Even using Astra at maximum strength, the problem becomes how to sift and sort. “In an age of endless outreach and interest with AI, the scarce thing is the human endorsement or the human conviction.”
• AI also helps address the cold start. Inspired by Wikipedia, whose profiles people find aspirational, Cosign creates profiles from public sources and aggregates information people have posted elsewhere — “we’re bragging for them on their behalf.” David says only because of AI can the team generate this information at scale and recreate a professional graph. Olivia connects the approach to Nikita Bier’s “your friends just answered this question about you” strategy, which worked three times.
• Erik places the product in the AngelList lineage. Ten years earlier, AngelList had been a public startup-community utility organized around companies, investors, and talent before moving away from its free public social network. David also invokes Eric from Benchmark’s comment on Patrick O’Shaughnessy’s podcast: “It’s all going to work. Everything is going to be huge” at the category level. LinkedIn can keep thriving, Stanford can remain fine, and another great product can add value rather than eliminate the incumbent.
• David stresses that Cosign is an invitation rather than a finished product: people can edit profiles, submit updates, cosign claims, and help build the network together.

8. Cosigns make careers — and the long-game answer to “why give away my alpha?”

• Erik’s canon includes Paul Graham publicly naming Sam Altman among the five best entrepreneurs around 2009, before Altman had a major exit; Ben Horowitz and Ali from Databricks; and Ryan Hoover bringing Erik into Product Hunt to help build its community. Erik’s current examples include Dwarkesh, whom he says was 25 and had never had a job as far as he was aware, plus Michael Truell, in his mid-to-late 20s, and Leopold Aschenbrenner, whom Erik says was 24 or 25.
• Erik supplies the long-game answer to “why would I tell you they’re good?” If someone is raising a pre-seed round now, perhaps do not cosign them for immediate advantage; instead, cosign the person who may raise two or five years from now. When that founder eventually raises, Erik’s example is that they may remember Olivia’s early help and call her first.
• Erik’s close is that people remember, “by an order of magnitude more,” the people who believed in them first. When Airbnb went public, he says, Brian Chesky wrote a blog post crediting YC and Paul Graham, his first investor, despite much larger later checks. “If you’re talented, if you’re smart in this industry, someone will discover you, cosign you, and make your career.”