The question so far
What judgment does a founder need to retain as AI makes more work possible? In August, George described learning model training so he could challenge plausible explanations. In October, he described staying home from a trip because the compute still depended on his attention. The question remains how to build useful autonomy while retaining responsibility for objectives and evidence.
Last signed positions
But delegation only works when someone still owns the objective, the evidence and the tradeoffs.
The parts of AI I could no longer delegate · 2026-08-13
I own the compute. Right now, it owns my calendar.
I'm Afraid to Take a Vacation · 2026-10-10
Dated developments
- 2026-08-13: learning training and evaluation becomes part of supervising the company.
- 2026-10-10: George identifies founder dependency as a system he still needs to change.
Earlier coverage, kept at its original canonical URLs.
Previously published coverage
Open questions
- Which decisions can proceed without George being present?
- What evidence makes a result worth accepting?
- Does the system create time away, as well as output?
Next checkpoint
A signed follow-up describing what changed and what evidence supports it. The earlier essays identify intentions; this page does not report those intentions as completed results.
Disclosure
These are George’s personal accounts of running SimpleDirect. This page collects existing writing and questions; no additional measured outcome has been supplied.