Ashley Stirrup

AI just killed the two biggest excuses for not experimenting. "We don't have enough engineers." "We don't have enough traffic." Kameron Tanseli and the Fyxer's growth team are just 4 people. Last year they ran 360 experiments — more than 2 per working day, per engineer. And their win rate? 25%. That means 75% of ideas failed — and got caught before shipping to everyone. On traffic: you don't need massive scale to experiment. You need bold enough ideas. A 20–30% effect — the kind that comes from testing pricing models, usage limits, or core flows — is detectable even with a small sample. Those are exactly the bets worth running at a startup. Here's how AI compressed the entire experimentation loop: → Cursor desktop previews experiments in a virtual environment — no pulling code, no manual QA → Shared Claude skills mean common workflows are reused, not rebuilt from scratch each time → An AI data analyst lives in Slack so anyone can query the warehouse without the data team → GrowthBook's API + Claude auto-diagnoses metric drops by cross-referencing recent experiment launches Research that took days: now hours. Dev that took a week: now an afternoon. Analysis that needed a data scientist: now available to everyone in Slack. The result isn't just speed. It's a fundamentally different learning curve. When you can run 2 experiments a day instead of 2 a month, you build customer intuition in weeks that used to take years. Counterintuitive wins get found instead of left on the table — like adding a credit card gate that jumped free-to-paid conversion from 5% to 35%. Fyxer went from $1M to $35M ARR in a year. Targeting $100M–$150M next. The multiplier wasn't headcount or traffic. It was learning faster than the competition. What's the excuse your team is still using? 📖 Full breakdown + podcast episode in the first comment #ABTesting #Experimentation #AIEngineering

Only one snapshot so far — the engagement-over-time curve appears once the daily scrape has captured this post at least twice.