Independent Research
Experiments with real numbers. Agentic systems, strategy, and engineering craft. Honest findings, full limitations — for engineers and leaders shipping AI.
Research
10 posts
Real experiments with numbers. Benchmarks, security tests, and RAG evaluations on open-source models.
Strategy
8 posts
How to think about AI at the org level. Architecture decisions, build vs buy, adoption patterns.
Craft
3 posts
Concrete engineering. Patterns, code, and the non-obvious decisions that separate good AI from great.
Recent
At Build 2026, the Microsoft AI team led by Mustafa Suleyman announced the MAI model family: 7 models covering reasoning, coding, images, voice, and transcription — every one built entirely in-house.
CraftDumping your notes into an LLM context is easy. Giving an AI agent structured, safe, scoped access to a living knowledge graph is not. Here's the architecture I built.
ExperimentI wanted AI that runs with no internet and keeps everything private. No API keys, no cloud, no data leaving the machine. So I built it in Rust using llama.cpp. Here's what that actually looks like.
ExperimentGemma 3 1B has 1152 dimensions. Gemma 3 4B has 2560. They speak different sizes. I trained a single layer to translate between them and logged every number.
Agentic AINine links in, Claude Code stopped summarizing and called out the avoidance. Notes on what happened, and what training keeps this behavior intact.
BenchmarksGemma 4 E2B scored 80.4% across 9 enterprise suites — 0.4 points behind the previous-gen 4B model. But the real surprise: 70% multi-turn, the highest score in the entire Gemma family.
Stay current
One email per post. Real findings, honest limitations. No roundups, no sponsorships.
Suggest something to test. If it’s interesting, I’ll run it and publish the results.