Garry Tan on finding product-market fit and what early YC founders got right

Executive overview

Most startup success stories are told from the pedestal — after the fact, stripped of the years of grinding before anything worked. Garry Tan, YC partner 2010–2015, shares the unfiltered version through Soylent, Instacart, and Coinbase.

Each story turns on one founder ignoring conventional wisdom, demonstrating momentum instead of pitching, or walking away from guaranteed money to go after something fringe.

The pattern: great founders do the thing first, then ask for permission.

Soylent: why ignoring bad advice built a brand

• Rob Reinhart pivoted from networking hardware after building a Reddit community around meal-replacement powder.
• Tan told them not to call it Soylent. They ignored him.
• The controversial name generated tens of millions in free earned media — people who hated it still talked about it.
• Nine out of ten people hated it; the one in ten who loved it became subscribers and true believers.
• The lesson Tan drew: every business is a meme now — polarising beats forgettable.

Instacart: doing it beats pitching it

• Apoorva Mehta applied two months after YC had already started — near-zero odds of getting in.
• He sent Tan a six-pack of beer delivered via his own app, already live in Mountain View.
• Tan was the only partner to reply to his cold email. He brought Apoorva in the next day.
• What stood out: Mehta had built both the driver app and the retail app and hired two people off Craigslist to run it.
• Most people had the same idea (Lyft and Uber were launching at the same time). He was the only one who had built it.
• Instacart became the best company in that batch — a $17.5B company available to 80% of Americans within eight years.

Coinbase: walking away from guaranteed upside

• Brian Armstrong was head of anti-fraud at Airbnb in 2011, when Airbnb had just hit a $1B valuation — the event that put YC on the map.
• He read Satoshi Nakamoto's white paper and walked away from what was as close to a guaranteed payday as Silicon Valley offered.
• Crypto at the time was so fringe that people used pseudonyms; buying Bitcoin meant wiring money to Japan via a trading card site.
• Armstrong's insight: make it look like a bank. Put your name on the homepage. Accept a credit card.
• He removed all the friction, legitimised the category, and built the infrastructure crypto actually needed.

Shelf Engine: the slow-build pattern

• Stefan Kalb built a sandwich brand to $5M/year in Seattle, then had a Microsoft engineer friend write ordering software to cut food waste.
• The software added $1M in profit to the sandwich brand by reducing perishable waste.
• Tan pre-seeded it. It now runs optimised ordering for large grocery chains across hundreds of stores.
• 30% of perishable food is wasted; it's a problem that software solves cleanly.
• It took four years to get there — not an overnight success.

The 10-year overnight success

• Most startups follow a decade-long arc: years of nothing working, near-death moments, feeling like you'll lose everything.
• The press only covers the pedestal stage; founders watching from outside skip straight to wanting that and miss everything before it.
• Long duration wins: find a large, growing market and stay committed to it for years.
• Difficulty is the point — it has to be hard to be worth doing.