Soham spent 6 months building AI that would auto-generate integrations between any software. He locked down Glean as an early customer because he had friends there. And it failed completely.

So he pivoted. This time, he refused to work with friendly customers who knew him. Instead, he did 10-20 calls per day with strangers who would tell him his product sucked. He posted on Discord communities at 3am, wrote technical blogs that went viral on Reddit, and created fake landing pages to see what integrations people actually wanted.

In one year, Composio grew to 100,000 developers and raised $30M from Lightspeed in just 3 weeks.

His contrarian take: in AI, asking users what they want will just get you faster horses. Built it instead, and watch their eyes light up.

Why You Should Listen:

• Why friendly customers will kill your startup.
• The 20 calls per day strategy that scaled Composio to 100,000 users.
• Why you can't validate AI products by asking.
• The exact Discord and SEO tactics that got their first thousand users without spending on ads

Keywords (comma-separated):

The PMF Show is a startup podcast. The Product Market Fit Show is a startup podcast. Startup Podcast, Composio, Soham Ganatra, AI agents, developer tools, pivot, Series A, Lightspeed, integrations, API, tool calling

00:00:00 Intro

00:06:44 Playing with GPT-2 before ChatGPT

00:12:37 Leaving his job to start Composio

00:21:16 Pivoting to integrations for AI agents

00:28:42 Why friendly customers are dangerous

00:31:01 Getting first users through viral content

00:36:01 Taking 10-20 customer calls per day

00:40:58 Scaling from 1,000 to 100,000 developers

00:43:58 MCP and the explosion of growth

00:48:59 Raising $30M from Lightspeed in 3 weeks

Send me a message to let me know what you think!