a16z Turned a 7-Year-Old Email Into a Product Because AI Made Cold Outreach Worthless

Source: a16z | Published: 2026-09-25T14:00:23Z

When AI drives outreach costs to zero, genuine human-to-human vouching becomes the scarcest signal. a16z formalized that logic into a new product called Cosign.

The experiment never became a product — they got busy with other things. Now, a16z has picked up that same idea and built it into Cosign.

One Tweet Worth More Than Four Years of College

Writer David Perell once told people that his Twitter account was more valuable than his college degree — he spent hundreds of thousands of dollars on school, but Twitter delivered far more career returns on that investment.

This isn't just a throwaway observation. At its core, college offers three things: education, a network, and credentials. At its peak, Twitter replicated all three. You learned through conversations, you met people through follows, and you earned a kind of peer-endorsed credential — being followed by Patrick Collison, Elon Musk, or Marc Andreessen opened more doors than a Stanford diploma.

The problem is that this kind of endorsement is implicit, scattered, and ephemeral. What Cosign wants to do is make it explicit, durable, and searchable.

Three Types of Endorsement, Each With Its Own Granularity

Cosign's product is built around three categories of endorsement.

The first: who shaped your career. This is mentorship made explicit. If a senior partner has invested serious time in you, that itself is a signal — they believed you were worth their time, and others should take note.

The second: who you'd go to war with. This is far more precise than "I follow them." On existing platforms, you can't tell whether two people fought side by side in the same trench or just bumped into each other at a conference. Every connection on LinkedIn gets flattened into a single dimension.

The third: who you believe is a rising star worth watching. This is the highest-risk endorsement — you're staking your judgment on someone the market hasn't yet validated, putting your reputation behind theirs. This is exactly the logic behind that email experiment seven years ago, and why Russell Kaplan's story became the product's founding myth.

These endorsements can also be ranked. Over time, a traceable record emerges: who believed in whom, when, and how strongly.

Expressing Intent Privately: The Mutual Match

There's one design decision in the product worth discussing on its own: the private signal.

Many of these intentions can't be expressed publicly. You might really want to hire someone, but you can't message everyone one by one. You're interested in investing in a founder, but the timing isn't right yet. You think a company might be worth acquiring, but asking openly would be too forward. On existing platforms, these intentions either go unexpressed or, when they do surface, make the other person uncomfortable.

Cosign lets you flag privately: I would invest when they raise, I would hire them when they're considering a move. Only when both parties have expressed mutual intent does the system facilitate an introduction. A high-intent, bilateral introduction is far more efficient than a cold DM sent at random.

One partner on the podcast shared a story from 2018. SV Angel hosted an event and sent all attendees a list of 150 people beforehand, with one rule: both sides select who they want to meet, and mutual selections trigger an automatic introduction. The people she met at that event became her most important collaborators and friends. She noted that there was no way to have a deep conversation with 40 people at a venue, but the mutual selection mechanism provided initial context — enough for real relationships to form afterward.

The problem was, that system only existed for that one afternoon.

Making Endorsements Last

There's a real example from inside a16z. When new partner Josh joined the firm, people who had worked closely with him for years posted on social media: "Josh helped my career — I genuinely recommend him." Those words lived on Twitter for a day, then got buried under new content.

Endorsements on existing social media are real-time but short-lived. You retweet someone's post, signal your trust in them, and three months later that retweet has vanished into the feed as if it never existed. The most critical moments in someone's career — when someone was the first to stand up and say "this person is worth watching" during their most obscure years — those words shouldn't be at the mercy of the day's algorithm.

Cosign's approach is to anchor these moments to a person's profile. Every time someone changes jobs or a company raises a round, there's a window for endorsement. Rather than letting social recognition dissolve into the feed, the idea is to deposit it into a durable reputation record — so the first time you look someone up, you can see who said what, and when.

LinkedIn's Legacy and What It Can't Do

Josh spent his early career at LinkedIn, around 2004 to 2005, when the company was just beginning to build its network.

Reid Hoffman described the company as having three phases: growth, usage, revenue. LinkedIn's early network value rested on a premise — that connections were real. You knew the person, you were willing to vouch for them, you'd introduce them to people you respected. The entire trust graph was transitive: David asks Olivia to introduce him to Eric, Olivia tells Josh "David is a close friend of mine, would you help with this introduction?" — and that chain of trust meant something.

But as the network expanded, "connection" got diluted. A high-five at a conference, an email exchange — both became connections. Eventually, when you asked a friend "do you know this person on LinkedIn?", the answer was probably "hmm... I don't really remember."

LinkedIn's connections exist on a single dimension — connected, or not connected. Whether you were on the same team, whether you weathered a crisis together, whether you were someone's direct manager — all of it compressed into a single line.

This isn't to say LinkedIn did anything wrong. Over a billion users, decades of network effects — no one has genuinely threatened it. But there is an obvious gap between what it can do and what this space actually needs: high-quality, granular, emotionally weighted endorsements have never really worked on LinkedIn.

RISE Awards: The People Outside the Spotlight

About ten years ago, a16z launched an initiative called the RISE Awards, specifically for engineers, designers, and product managers — the people who built the companies but rarely appeared in press coverage. Josh Elman was one of the judges and received thousands of nominations.

Erik Torenberg, an a16z partner, brought it up on the podcast because Cosign wants to carry forward the same logic: startup stories are usually founder stories, but the people who actually make a company succeed are usually far more numerous. That fourth engineer might have been critical to whether the company survived, but when they leave to start something new or join another company, they can't take that chapter with them — no media outlet will put them in the headline.

This is a structural gap in the information market. Individual contributions are non-portable: it doesn't count when you say what you did yourself. It only counts when someone publicly endorses your contribution.

They noted that the most valuable byproduct of the RISE Awards was that winners got to know each other, attended follow-on events, and built lasting relationships. The award itself was a one-time thing, but the connections endured.

AI Solves the Cold Start

Every reputation platform faces the same cold-start paradox: no one's here, so no one comes; no one comes, so no one's here.

Cosign's solution is to use AI to generate profiles for people and companies from public information first — funding announcements, job changes, positive mentions across social media — aggregating and structuring this data into a "people are already praising you" homepage, then sending it to the person in question. Look, here are endorsements that already exist. Come and complete your profile.

This also explains why this product is appearing now. One partner noted that she posted on LinkedIn saying she wanted to help portfolio companies hire, and received two thousand applications — a number that could be ten thousand in the future, and a hundred thousand after that. In an era where AI is driving the cost of outreach toward zero, she argued, what becomes scarce is genuine human endorsement — the human vouching, the human judgment, the human conviction.

When anyone can express interest effortlessly, the signals that actually carry weight are those where someone is willing to put their own reputation on the line.

The First Person Who Believed in You

Around 2009, Paul Graham wrote an essay listing the entrepreneurs he considered the best, and included Sam Altman. At the time, Sam didn't have a major exit on record. But that endorsement directly shaped his trajectory — running YC, then OpenAI.

There's a counterintuitive design choice baked into the product: why publicly endorse the talent you've discovered? Aren't you giving away your alpha? Josh Elman addressed this on the podcast by citing a line from a novel: be so good that people would do business through you. If you cosign two people, they might meet, collaborate, invest in each other — and all of that can happen without you even knowing.

The answer is: the person you want to invest in today is raising a pre-seed round, but the person you should really be endorsing is the one who'll come to you two years from now. They'll remember you as the first person who publicly believed in them. When Brian Chesky wrote his letter at Airbnb's IPO, the first people he mentioned were Paul Graham and YC — not the investors who came later with more money and more time.

Dwarkesh Patel was 25, had never held a formal job, and moved to San Francisco to start a podcast. A few early people chose to appear on his show and vouch for him — then Zuckerberg came on, and other founders followed. Leopold Aschenbrenner was 24 when he wrote Situational Awareness. This industry is always searching for the person who hasn't been discovered yet. And the person who finds them first will be remembered for a long time.