Consensus conversion: From weird to obvious

Thoughts on the age-old question in venture: go with the crowd or against it?

Thanks to Varadh Jain - this post is built on his idea. He’s among the sharpest people in venture (and also my brother).

Over the weekend, Martin Casado (GP at a16z) poked the bear: non-consensus bets aren’t where alpha lives in seed, because follow-on investors fund what looks obvious. VCs piled on (we can’t resist). Founders warned their peers about “non-consensus hipster VCs”.

The saying goes: alpha in venture comes from non-consensus bets. That sounds right: if everyone already agrees, the price is high. But there’s a catch. Startups don’t raise their next round from philosophers. They raise from follow-on investors. And follow-on investors fund what looks obvious to them. If your seed investment never becomes obvious, it dies. A lot of what gets branded as “contrarian” is theatre: performance without a plan to convert.

So the paradox at seed is this: the best companies usually start as non-consensus, but they have to become consensus to survive.

My job as a seed investor isn’t to fund weird ideas for their own sake. It’s to spot under-priced insights, truths about markets, tech, or behavior that look small, risky or strange, and back the teams turning them into the obvious (and help where I can). Some convert quickly. Others take years. Both can work.

Two kinds of consensus matter:

• Idea consensus: is the category already in the air?

• Team consensus: do the right people want work with them? join them? back them?

Team consensus is underrated. A non-consensus idea with a consensus-adjacent team converts faster: they can get design partners, close early hires, and borrow reputation. It’s also why hubs matter. Consensus-adjacent is just a fancy way of saying: people are willing to pick up the phone for you.Think Hayes Valley AI houses, El Segundo/“Space Beach”, ex-OpenAI circles, YC apartments, OSS Discords. The next non-consensus ideas usually start close to today’s “consensus” builders. They share offices, intros, design partners. Proximity speeds up proof. Distance slows it down. Venture is getting industrialized: bigger funds, more playbooks, assembly line demo days. But seed is still social: who is near whom and who trusts whom? It’s not factory work. It’s about proximity and proof.

You can even draw a 2×2 (yes, another VC 2X2). Idea consensus on one axis, team consensus on the other:

• High/High → the hot deal

• Low idea/High team → the sweet spot

• High idea/Low team → risky but fixable

• Low/Low → very hard but possible

The best seed bets don’t stay weird. They start non-consensus, but with the right networks, traction, and time, they become obvious.

Here are some examples (please send me counterexamples!):

• Airbnb (2007): “Obviously bad” idea - strangers renting air mattresses, but YC/PG network + obvious traction (by early 2011, they’d hit ~1M nights booked) → consensus.

• Coinbase (2012): Crypto was fringe. YC + top-tier backers and strong adoption (1M wallets by 2014) → mainstream.

• Palantir (2003): Truly weird idea, but consensus-adjacent founders (ex-PayPal Mafia). Took years of government contracts, but created its own network. Now a $380B+ public company.

• Anduril (2017): Non consensus idea - startup defense prime. Consensus-adjacent on team/network (Palantir/Founders Fund network). Now a $30B+ private company.

• Perplexity (2022): Non-default UX for search. Consensus-adjacent network (ex-OpenAI, Sequoia)+ insane usage (100M monthly visits in under two years) → mainstream. Raising at a $20B valuation.

• Figma (2012): Browser-based design sounded clunky, but Dylan (Thiel Fellow) attracted consensus backers (Sequoia, Index). Once multiplayer shipped, designers evangelized it → collaboration is now table stakes. $35B publicly traded company.

Not every great company converts quickly. Shopify in Ottawa, Atlassian in Sydney, UiPath in Bucharest weren’t near the hubs at all. It took years before the market “got” them. If seed investors only back what looks like it can flip by Series A, we systematically filter out the most radical opportunities. That’s expensive. It’s also antithetical to the point of (ad)venture capital.

YC is probably the best conversion machine ever built. It takes founders who might look like long shots and makes them legible. Demo Day compresses fundraising into hours. The SAFE made closing simple. The YC brand flips how people see you. Many investors who grumble about YC valuations wouldn’t have backed the same founder pre-YC. That’s what conversion looks like.

It’s easy to fall into traps if we don’t deal in specifics. Networks can compress time-to-proof, but they obviously can’t replace product-market fit.

Consensus vs non-consensus isn’t a moral stance. It’s a path. Some paths are short. Some are long. The job at seed is to see the path and back the teams that can walk it.

The best seed bets start weird and end obvious.

---

Which companies do you think stayed weird and still worked? I want to hear them. Email me at vedika@weekend.fund.