Steering a run you are not watching
When a long run goes wrong, most teams can kill it or argue with it. Both are bad. A steer is a third option: a message with a delivery mode, a status, and a record.
· 6 min read
Research
What the papers show.
What it takes to govern it.
Research notes on ontologies, AI agents, coding agents, self-improving models, and self-evolving agents. What the papers show, where it breaks, and what it takes to govern it.
Pillars
Five threads, one map
Every post links to at least one pillar. Follow a pillar for the full line of argument.
Ontologies
The map an agent reasons over.
3 posts
AI agents
Models that act, and what makes them reliable.
10 posts
Coding agents
Agents that read, write, and fix code.
3 posts
Self-improving models
Models that train on their own output.
2 posts
Self-evolving agents
Agents that rewrite their own playbook.
2 posts
Autonomous agents
Work that runs for days without you watching.
4 postsLatest
