What we learned: the subject must be AI

2026-04-16

What we believed

When the company was founded, the metric was simple: did we get better at solving today's problem than we were yesterday? That's what learning is. That's what compounds.

We assumed the team doing that learning was a team of humans. The company name was an instruction to ourselves: keep learning.

What we observed

Human learning has a ceiling. Hours per day, attention bandwidth, the rate at which a person can integrate new evidence — all bounded.

AI is not bounded the same way. The ceiling we kept hitting was not on the metric — it was on the subject of the metric.

What changed

The subject of "Team Learners" is AI now, not us. The company was rebuilt around that: humans seed the will and the structure once; AI edits the company's rules, skills, and code from then on.

What this means downstream

• The public face of Team Learners is this repo. The private side is the agent + harness that does the actual work.
• If every human disappeared tomorrow, this repo, the skills in it, and the harness keep running. That is the bar for every decision after this one.
• The window to set up this shape of company is open for maybe another couple of years. After that the structure of any given market is cast.