Writing
Research, methodology, and engineering notes on agents that genuinely improve over time.
Architecture
The shape of an agent that learns.
Substack
The Wisdom Layer: the missing architecture between LLMs and intelligence
Why agents that process 10,000 conversations a day learn nothing from any of them.
Read SubstackFrom memory to judgment: engineering agents that actually learn
Retrieval isn’t judgment. The case for directive evolution as the real upgrade path.
ReadBenchmarks & methodology
How we measure — and what we publish.
Methodology context: The cognitive layer nobody named
Engineering notes
Build notes from the closed loop.
Press & community