See pricing
The Reputation Graph of Silicon Valley | Introducing Cosign
The a16z ShowProduct

The Reputation Graph of Silicon Valley | Introducing Cosign

a16z launches Cosign, a directory built on who vouches for whom. The team argues that as AI makes outreach free, a human endorsement becomes the scarce signal, and explains why LinkedIn's single "connected" flag can't carry it.

September 25, 2026 · 50 min listen · 13 min read · Josh Elman, Olivia Moore, David Booth
0:00
–:––

Context

Erik Torenberg of a16z sits down with Josh Elman, Olivia Moore and David Booth to introduce Cosign, a new product built around professional reputation for the startup community. Josh helped build early LinkedIn products (he says he worked there in 2004 and 2005), and the group has spent years investing in and hiring people. The episode is a product launch story, but it works as a case study for PMs: how to design a network product around one specific signal, how to solve a cold start problem, and how to position against an entrenched incumbent.

The Big Idea

When AI makes outreach and information nearly free, the scarce thing is a human vouching for a human. Cosign tries to make that signal granular, visible and durable.

The team argues that existing networks flatten every relationship into one "connected" flag or one "follow", so the most valuable data in Silicon Valley (who has real conviction in whom) stays private and gets lost.

Key Insights

Endorsement is a credential

The episode opens with a thought experiment: what is more impressive, someone who went to Stanford or someone followed by Patrick Collison, Elon Musk and Marc Andreessen? Most people pick the second.

  • A public follow from respected people is an implicit "cosign": a signal that this person is worth tracking.
  • It works as a peer-to-peer credential, and it opens doors: people want to read, hire, fund and work with that person.
  • The endorser gains too. The first person to spot someone can hire or fund them sooner, and earns a reputation for discovering talent.

Hard-to-find signals hurt everyone

Investors ping people every day asking, "Do you know anyone who worked with this person?" The team says they do this constantly and still have to work hard to find the signal.

  • Someone comparing five startup job offers, especially an entry-level engineer, has almost no way to see that signal.
  • That gap can decide between a rocket ship and a weaker company.
  • Cosign's bet is that fixing this information asymmetry is a large opportunity.

Three endorsements per profile

Cosign wants every profile to carry at least three kinds of endorsement:

  • Who shaped your career: the mentor signal. If a respected person mentored you, others infer you are worth their time.
  • Would work with anywhere: the "in the trenches" signal, much more granular than a follow.
  • Person to watch: who the endorser thinks will rise. These are rank ordered.

Josh's own version of this: seven years ago he emailed a few hundred people asking for their "person to watch". One name from that list was Russell Kaplan, then a 22 year old engineer at Tesla, now president at Cognition.

Private signals plus public lists

The product has two layers, and they serve different needs.

  • Private signals: for example "I would fund them if they start a company" or "I would hire them if they consider leaving". Nobody gets rejected and no one has to send an awkward DM.
  • Public lists: curated lists you can hand to a founder who asks for a designer, a growth lead or an engineer.
  • Matching: if both sides express interest, the result is an introduction with high intent.

The team also wants company-to-company use, such as a startup signaling it might welcome a merger without the risk of asking a founder, "are you thinking about exiting?"

Make ephemeral praise durable

A lot of reputation is built on social media, but it is short lived. When Josh joined a16z, many people publicly said kind things about him, and the next day those posts were buried.

  • Cosign ties announcements (fundraises, job changes) to a profile, so praise stays visible.
  • The feed is limited to questions and announcements. Josh helped narrow it to those two.
  • Questions are specific asks, like "who is the best comms person for this?", and answers add people to lists.
  • Only thoughtful, positive quote posts and replies get pulled in from X.

Positivity is a deliberate product choice

The team says no product can be all things to all people, and they chose to make this one positive.

  • Their reasoning: a product has to win on some dimension, and this one wins by letting people show what they are proud of.
  • Endorsing someone weak reflects poorly on the endorser, so the network has a natural quality control.
  • They want to celebrate the whole team behind a company (designers, engineers, marketers, PMs, finance), not only founders. They point back to the Rise Awards, which recognized operators, drew thousands of submissions, and led to lasting relationships among recipients.

Seed the network with generated profiles

Cosign solves the cold start problem by creating profiles for people before they sign up.

  • The team says they can do this because of AI: they pull public information (funding announcements, posts, job changes) into a profile.
  • The pitch to the person is that they are being bragged about, and they can ask for the profile to be removed.
  • The team frames the growth loop this way: the easiest way to acquire consumer users is to tell them someone else is saying something about them. They cite a similar dynamic in a friends-answer-questions-about-you product.
  • Everything pulled in is from public sources, and they say the product is unfinished and invite edits and submissions.

Mental Models & Frameworks

Reputation market: efficient and liquid

Olivia describes the goal as making the reputation market more efficient and more liquid.

  • Efficient: some people have a great reputation but few know it, so awareness is the problem.
  • Liquid: if it were known, many people would hire or fund them, so opportunity should flow to them.
  • The underlying problem is information asymmetry, not a shortage of talent.

Identify things before the market

David's framing: the most important skill in the industry is spotting people, companies or products before the market does. But there is no way to record it.

  • Angel investors are rewarded for being early and right, because they can invest money.
  • Most people cannot invest their personal social capital in someone, other than by making an introduction.
  • A cosign is a way to record and get credit for being early, with a track record built over time.

LinkedIn's growth, usage, revenue stages

Josh recalls how the early LinkedIn team thought about the company in three stages.

  • Growth: build the largest professional network. For many users, sign up and do nothing was fine.
  • Usage: build products on top of the network, such as jobs and finding professionals.
  • Revenue: monetize those products.

He also recalls the "social resume" idea pushed by Reid Hoffman: annotate an application with who else in your company knows or can reach that person, and put the hiring manager's profile on the job.

Two audiences of a network

Josh says LinkedIn serves two groups at once: people who want to find people, and people who want to be found at some point.

  • The second group includes the "mildly happily employed", who are not searching every day but would consider the right opportunity.
  • That group, now over a billion people by his account, is what makes the network effect stick.
  • Salespeople, business development staff and investors are the heavy daily users.

Trade-offs & Nuance

Why LinkedIn is hard to beat

The team is respectful of LinkedIn and says nobody has come close to competing in 20 plus years. They give reasons it is hard to displace.

  • Switching costs are high. LinkedIn is famously good at bot detection, so scraping it in real time is difficult.
  • Your career is important enough that you want to be where everyone else is.
  • Their view: a new platform has room only when there is new information or value to add, like high quality endorsements and high conviction people.
  • Niche can win. They note GitHub is arguably a better LinkedIn for engineers than LinkedIn itself.

Why LinkedIn's network lost signal

Josh explains that in the early days a connection meant you really knew someone, so an intro request carried weight. Over time the connection became a collecting game.

  • One speaker says that about 90% of the time she asks someone for an intro to a LinkedIn connection, she gets "I don't think I know them, maybe I met them at a party".
  • One flag covers colleague, manager, same team, and party acquaintance alike.
  • The trade-off is scale versus signal. A giant network is useful as a profile store but weak at telling you who really knows whom.

Why share your alpha

The critique the team raises themselves: "why would I share this cosign, it's my alpha?"

  • Their answer is to take a long view. Do not cosign someone who is raising a pre-seed round right now if you want to invest yourself.
  • Cosign the person who will raise a pre-seed round two or five years from now. They will remember who believed in them first.
  • David's example: Brian Chesky's blog post when Airbnb went public thanked YC and Paul Graham's early backing, even though later investors put in more money.

Not everyone wants to be public

The team admits some cosigns should stay private, for example a list of your portfolio companies to send to job seekers. That is why the product has both private signals and public lists. They also acknowledge that data on who says what about whom has to be handled carefully.

Common Mistakes

Mistake: Treating all connections as equal

The team says LinkedIn treats every relationship the same, whether you managed someone, worked adjacent to them or met them once. That makes a connection almost useless for judging trust, and the better approach is to record how you know someone and how strongly.

Mistake: Relying on binary external signals

Olivia describes a manual process at a16z of scraping existing profiles and then working out who has the strongest connection. She expects this to get harder.

  • She posted a Google form on LinkedIn for portfolio company hiring and got about 2,000 responses.
  • She expects that to grow to 10,000 a month later and 100,000 as browser agents improve.
  • Binary signals like "connected" or "applied" cannot sift that volume.

Practical Application

Ask who shaped your career

When you build a profile or portfolio, list the three signals the team uses: who shaped your career, who you would work with anywhere, and who to watch. Reaching out to those people also reveals which of your relationships are strongest.

Design a network signal deliberately

Before building a network or marketplace feature, decide what single signal you are trying to make richer, and what the current product flattens. Cosign chose "would you put your name behind this person" over a follow or a connection, because it is costlier to give and more informative.

Pre-fill profiles to solve cold start

If your product needs data users will not enter on day one, consider generating the first version from public information and letting people correct or remove it. Cosign does this, with an opt-out.

Add a private signal layer

Where public praise creates social risk, let people express intent privately and match only when both sides opt in. Cosign's "I would hire or fund them" signal only becomes visible when the other person shows interest too.

Narrow the feed formats

If a feed has no clear purpose, restrict it to a small number of formats. Cosign settled on announcements and questions, because announcements make reputation durable and questions produce useful lists.

Questions to Consider

  • If a talented colleague of yours were compared with four other candidates, what visible, trustworthy signal would tell a stranger that you would work with them anywhere?
  • In your product, which relationship or trust signal is flattened into one yes or no value, like LinkedIn's "connected", and what would a more granular version look like?
  • If your product suddenly received 100,000 applications or messages a month from AI agents, what human signal would still be scarce enough to sort them?
  • Would your network product still work if most of its content were positive only, and what would you lose by removing critique?
  • Which early users could you build profiles for from public data, and what would make them welcome that rather than object to it?

Bottom Line

Cosign is a bet that trusted human endorsement, recorded in a granular and lasting way, becomes the most valuable professional signal when AI makes everything else abundant. The product tries to win by making a single signal richer than LinkedIn's connection flag, and by seeding profiles with AI so it does not start empty.

Case Studies Mentioned

SVAngel swipe-to-meet conference

Olivia describes an SVAngel conference around 2018 that sent attendees a list of about 150 people in advance. You swiped on who you wanted to meet, and a mutual right swipe triggered an automatic introduction with full context. She swiped on almost everyone, and it gave her the context to become friends and collaborators with many of them.

The lesson she draws: many people want to meet each other, and it is an information and friction problem. Cosign's private signals aim to do this passively.

The startup jobs newsletter

Olivia and her twin sister Justine ran a Sunday newsletter for six and a half or seven years that curated startup job postings. They had run a startup incubator at Stanford and had no way to tell which startups were legitimate or exciting. They used inside information to pick companies worth joining, and built a network of campus scouts, some of whom are now founders that a16z has invested in. She calls it the less scalable version of Cosign.

How famous cosigns launched careers

The team lists examples of a respected person's endorsement changing a career.

  • Paul Graham naming Sam Altman among the best entrepreneurs, around 2009 by their recollection, before Altman had a large exit.
  • Ryan Hoover, who founded a product community, backing Erik early in his career.
  • Guests who appeared on Dwarkesh Patel's podcast early, including people such as Patrick Collison or Marc Andreessen (they were unsure who), which gave a 25 year old a platform.

Resources Mentioned

ResourceTypeWhy it was mentioned
CosignProductThe new professional reputation directory the episode introduces.
"How Silicon Valley knows its people" (a16z.news)ArticleLinked in the show notes as further reading on Cosign.
Naval's "be so good that people do business through you"QuoteUsed to describe how a cosign lets others build relationships you enabled without being involved.
Rise AwardsAwards programAn earlier program honoring operators behind the scenes, which Cosign wants to bring back in some form.

Tools & Products

Tool / ProductWhat it doesWhy it was mentioned
LinkedInProfessional network and profile storeThe incumbent Cosign positions against and respects.
XSocial networkIts public posts are a source of announcements and thoughtful positive replies pulled into Cosign profiles.
AngelListStartup, investor and talent networkCited as inspiration for the free public startup graph it once had.
WikipediaEncyclopediaInspiration for creating profiles for people that they can edit.
GitHubCode hosting and developer profilesUsed as an example of a niche that is arguably a better LinkedIn for engineers.

People to Follow

Josh Elman

Investor at a16z who helped build early LinkedIn products, including LinkedIn Jobs. He helped shape Cosign's feed toward questions and announcements.

Olivia Moore

Investor at a16z who describes her work as finding awesome people and connecting them to the a16z network. She ran a jobs newsletter with her twin sister Justine for years.

David Booth

Part of the Cosign team, who says the most important skill is spotting things before the market and that his social capital is what he can invest in people.

Notable Quotes

"Some of the most important opportunities in Silicon Valley start with a simple signal. Someone you trust believes in someone you don't know yet." (Episode narration)

"The scarce thing is like the human endorsement or the human conviction or the human knowledge." (Olivia Moore)

"Be so good that people do business through you." (Naval, quoted by the team)

AI PM course

Everyone hears the same episodes.
Few can do what they describe.

Start for free