If you think agents will let you skip the messy, embarrassing early phase, Livingston's interviews say otherwise — the mess is where judgment gets built, and judgment is exactly what you're keeping for yourself. Read this now, before you over-automate the part that teaches you something.
Jessica Livingston's Founders at Work is a collection of candid first-person accounts from people who built companies you now consider inevitable. What emerges across dozens of conversations is not a repeatable playbook but a recurring texture: tiny user counts, wrong initial guesses, pivots born from desperation rather than strategy, and a stubborn refusal to quit that had nothing to do with confidence. The founders weren't executing grand visions — they were surviving long enough to find out what they were actually building.
The temptation with an AI-native stack is to believe that operational leverage compresses the learning curve too — that agents handling the grunt work also handle the uncertainty. They don't. What Livingston's founders actually earned during those small, ugly beginnings was calibrated judgment: knowing which signals to trust, which users to listen to, which pivots were real and which were panic. Agents can run the loops faster, but someone still has to decide what the loop is for. If you delegate discovery before you understand what you're discovering, you get faster wrong answers.
- Embarrassingly small beginnings are data collection, not failure — don't automate past them
- the judgment your agents will eventually serve had to be earned by a human first
- speed of execution is worthless if the founder hasn't yet found the real problem.
