Notes on agent systems, evals, and self-hosted LLM infrastructure. Rust + Python.https://samlatino.dev/en-ushttps://samlatino.dev/writing/bm25-beat-my-vector-database/https://samlatino.dev/writing/bm25-beat-my-vector-database/A crossover framework for lexical versus vector retrieval on code — and the adversarial bench harness I built so my own argument can lose.Wed, 10 Jun 2026 00:00:00 GMThttps://samlatino.dev/writing/every-model-fails-tool-calling-differently/https://samlatino.dev/writing/every-model-fails-tool-calling-differently/Tool calling is the load-bearing primitive of every agent stack, and open models break it in at least eleven distinguishable ways. Naming the failure modes changes how you build the layer above.Sat, 06 Jun 2026 00:00:00 GMThttps://samlatino.dev/writing/red-teaming-my-own-agents/https://samlatino.dev/writing/red-teaming-my-own-agents/Turning "resists prompt injection" into a regression number: a deterministic harness, 146 probes across five OWASP agentic categories, and a hardening sweep that went 73% → 3% → 0%.Wed, 03 Jun 2026 00:00:00 GMT