The Biggest Mistakes First-Time Founders Make — Michael Seibel

▶ Watch on YouTube · Y Combinator · 7:03

Y Combinator CEO and Partner Michael Seibel on the biggest mistakes first-time founders make.

TL;DR

A seven-minute talk on the Y Combinator channel from August 2019, when Michael Seibel was YC’s CEO. He lists eight mistakes he sees first-time founders make in their first year, and adds a caveat: “I bet you can name a startup that has made every one of these mistakes and has still been successful… but those tend to be exceptions.”

1. Solving a problem you don’t care about. Not fatal, since some founders learn to love their problem later. But “a lot of startups that fail basically fail because they lose motivation”, having picked a problem that seemed wanted, fast-growing or cool rather than one worth “five years plus of their life.”
2. Helping users you don’t care about. At Justin.tv the team liked the idea of democratising live video but was “not terribly in love with the people who chose to use the platform.” The turn came when co-founder Emmett Shear, a gamer, refocused the company on video-game streaming, which became Twitch.
3. Choosing co-founders you don’t know well. A pre-existing relationship (friend, co-worker, school project) tells you whether you can get through the hard times together.
4. Not having transparent conversations with co-founders. He names the usual sources of drama: performance (is my co-founder working as hard as I am?), goals and roles. Left unspoken, “resentment builds up”, and the eventual conversation “turns into a fight.”
5. Not launching. Founders picture launch as a press event and wait until they feel ready. Seibel asks whether you remember the day Snapchat, Instagram, WhatsApp or Uber launched: “launching is nowhere near a significant event to your users as it is to you.” Better “to build a crappier product, release it sooner.” Heavily regulated markets such as banking and lending are the exception. Otherwise “it’s actually possible to get some form of MVP built and launched in less than a month.”
6. Not using analytics, meaning not measuring what users do once they arrive.
7. Not knowing where your first users will come from. Finding the first 100 or 1,000 users may take digging. The first one to five should come from people you know: “you chose a problem where you don’t know anyone who has the problem, including yourself.”
8. Poor prioritisation, putting “sizzle over steak”: press, hiring, conferences and investors ahead of shipping product and talking to users. YC’s name for it: “you are cargo culting a startup… you are pretending to be a startup.”

How it connects

• First users. Point 7 is the same advice Kolysh’s Startup School talk gives seven years later (customers 1–3 from the warm network), and it feeds founder-led-sales.
• Launch timing. Point 5 recommends launching within a month, with a carve-out for regulated markets. Patrick Collison at Startup School 2026 describes Stripe’s two-year pre-launch build, kept grounded by a production customer from month two. He also asks whether the lean, launch-early default still holds in the AI era. Stripe is a payments company, so it sits in the domain Seibel exempted.
• Recurring YC mistakes. Dalton Caldwell on Lenny’s Podcast gives another YC partner’s list of early-stage mistakes. On analytics the two talk about different stages: Caldwell warns against A/B testing and feature flags “when you have no users”, while Seibel’s point 6 is about measuring behaviour once users arrive.
• Motivation and imitation. Points 1 and 8 cover the same ground as the same presenter’s Low Ambition episode with Dalton Caldwell seven years later: caring about the problem, and the “scenester” who plays at startups. Here that is “cargo culting.”

What was actually ingested

The full 7:03 talk from auto-generated English captions (176 segments), following nine YouTube chapter markers. The raw file is used as landed. Punctuation in quotes has been added for readability. “Cargo cult in the startup” is rendered “cargo culting a startup”, and the co-founder named only as “Emmett” is identified as Emmett Shear from public record. Provenance: YouTube D56QeyyQMLI.

Linked entities and concepts

• Entities: Y Combinator (channel)
• Concepts: founder-led-sales
• Dangling (not frontmatter authors, deferred): Michael Seibel (presenter; also co-host of every Dalton + Michael episode, which likewise list the channel as author), Emmett Shear, Justin.tv / Twitch

Scope and reliability

A 2019 talk, before generative AI. It is general founder advice with no AI content. The wiki keeps it as a pre-AI baseline for the YC founder-advice cluster, so that later AI-era claims (Collison on launch timing, Tan and Hu on tiny teams) can be read against what YC said before. It is one partner’s pattern-matching from office hours, with no data, and it says itself that every rule has counterexamples. No dynamic_capabilities: tags: founder practice outside the Warner & Wäger lens.