Tag: agents

All the articles with the tag "agents".

• When agents fail because their picture of the world is stale

  Published:Sep 16, 2026 |  at 03:05 AM

  Stale facts, lost corrections, and unchecked tool results can derail an agent. I'm exploring how explicit state and verification might help.

• Context freshness: your assistant should know how old its facts are

  Published:Aug 8, 2026 |  at 08:30 PM

  Stale context is the quiet way assistants fail. Here is how I'd give every fact an age, a half-life, and a hard expiry, with a small Python example.

• World models vs. LLMs: a division of labor, not a choice

  Published:Jun 13, 2026 |  at 10:10 PM

  Let the LLM handle language, keep the state of the world in an explicit, auditable structure, and put plain code in between.