The Cost of Instinct's 10% Daily Growth: Computing Power Demand Doubles Weekly, Founder Spends 40% of Time Figuring Out How to Buy

Beating AI News Flash: Noah Shinn, founder of personal AI Agent Instinct, said the company's user base has been growing about 10% to 11% per day in recent weeks, with corresponding compute demand nearly doubling every week. He now spends about 40% of his time thinking about how much compute to buy in advance.

The challenge is that compute cannot be purchased on demand in sync with user growth. Shinn did the math: if he buys only 2x the current compute, it will run out in about a week; at 5x, it will be maxed out in less than 3 weeks; buying 10x outright means betting a large upfront cost. Many resources also need to be reserved months in advance, and securing compute on short notice can cost 3 to 4 times more.

Even if Instinct's daily growth rate drops from 10% to 5% to 8%, Shinn estimates that after 3 to 4 months of continuous growth, it may need to prepare compute for a scale of 100 million users. The problem is that this is exactly the lead time required for new compute to come online. Buy too little and it cannot sustain growth; buy too much and, if growth slows, costs get amplified.

Shinn believes this is harder than scaling ordinary consumer internet products. When Instagram and Facebook doubled their users, they were mainly handling more requests; with personal Agents, each additional user also continuously consumes a large amount of model inference. He said the scaling speed Instinct is currently facing is even faster than similar problems Claude Code and Codex encountered earlier.