And a path to autonomous harness development
What and Why as the new "Source Code"
Teaching agents per-user decision policies
Agentic evaluations that test tool use over memorization
From YOLO Loops to Permissioned Platforms
Proving Specifications, and Commoditizing Correctness
"Politically Sensitive Prompts", a dataset for telling the difference
Faster, cheaper, and more accurate memory with LightMem
+How "GraphSEARCH" Addresses Them
Improving on "LLM Summarization" for Context Compaction
Receive the latest and subscriber only articles
