Writing | Tenure

Writing

Technical arguments and architecture decisions from the team.

Benchmarks The Embedding Invariance Problem: Why Bigger Models Don't Fix Retrieval Noise A 20x increase in embedding model scale produces identical retrieval precision. The problem is not the ruler. It is what we are measuring.

Jun 2026 →
Architecture Why Memory Is a State Management Problem, Not a Search Problem The dominant paradigm in AI memory treats persistence as a search index. That framing produces approximate retrieval, probabilistic isolation, and compounding noise. The alternative is simpler: treat memory as typed, scoped, versioned state.

Jun 2026 →
Architecture Why Precision Is All That Matters in AI Memory Retrieval precision is the difference between a memory system that works and one that only appears to. Why noise accumulation destroys value at scale, and why no downstream model can compensate for bad retrieval forever.

Jun 2026 →
Architecture How the alias enrichment flywheel works BM25 retrieval is only as good as the alias set it searches against. The flywheel is how that set grows, passively, continuously, and without any effort from the user.

Jun 2026 →
Architecture Why MCP is the wrong abstraction for memory MCP treats memory as a tool call. Memory isn't something you invoke, it's something that should already be there. A technical argument for why the abstraction matters.

Jun 2026 →