Since writing software features has become close to free sometime in early 2026, everyone building an AI Service is worried about defending their business in the long run. And rightly so - building a 100B company over >10 years clearly requires market power; even the most beautiful castle will fall to the enemy if the gates are open and the surrounding lands flat and clear.

The next wave of software companies sell AI services, not software. The unit of work is not a session, where AI makes a knowledge worker more productive, but a transaction or workflow: AI services software runs 24/7, is endlessly scalable and carries out tasks as they are queued up, with human intervention reserved for edge cases.

Smart people join bad start-ups all the time. Competent people work for a stressed-out founder in a start-up without real growth for a few years. Eventually they get frustrated and leave, and often go on to have very successful careers in other companies. I have done this myself. How could competent people get it so wrong? It happens because many instincts we have about finding a job come from a world before start-ups. At best, these instincts do not apply to a start-up job search. At worst, they hinder it.

I write when I am unsure about something and need to properly think it through. I find that writing and the process of editing helps me understand how things relate to each other.