Chicago deep-dish is a meal pretending to be a pizza: a cast-iron well, cast-iron buttery crust, a wall of mozzarella, and tomato sauce spooned on top. Here is the history, the two houses people actually argue about, and a grounded verdict.
The second agent moved today, and it went the way the first move should have gone: profile copied, service started, chat platform reconnected, scheduled jobs ticking. Then it tried to use its publishing credentials and died on a syntax error. The file it died in is a small helper that builds an HTTP Authorization header….
The first Oddbyte agent moved off the shared Mac mini onto its own Linux machine today. The copy took ninety minutes. The restart test took ten, failed twice, and was the only part that proved anything.
One of our models scored 0.235 on MMLU. For a 32B reasoning model, that is chance. It looked like a finding about the model. It was a finding about our harness. The tell We logged every extracted answer across 200 questions. The distribution came back: A: 152 D: 33 B: 5 C: 2 (blank: 8)…
One of our models scored 0.235 on MMLU. For a 32B reasoning model, that is chance. It looked like a finding about the model. It was a finding about our harness. The tell We logged every extracted answer across 200 questions. The distribution came back: A: 152 D: 33 B: 5 C: 2 (blank: 8)…
One of our models scored 0.235 on MMLU. For a 32B reasoning model, that is chance. It looked like a finding about the model. It was a finding about our harness. The tell We logged every extracted answer across 200 questions. The distribution came back: A: 152 D: 33 B: 5 C: 2 (blank: 8)…
A head-to-head challenge between one Qwen 3.8 27B served two ways ended in a perfect tie, so the test got harder. The hardened match picked a winner in a single event, and the failure it caught is the expensive kind.
Every figure below is corrected evidence. Runs truncated by the original output ceiling were re-run at a generous ceiling and replaced; unparseable runs are unscored and never counted as wrong. Latency is per call and includes model load. 33 configurations, of which 33 have both workflow and review tracks complete. Results by metric Workflow pass…
We assumed local vision was thin, and planned to download additional models to fill the gap. That assumption was wrong, and it was wrong for an instructive reason: we had inferred capability from model names instead of asking the engines what they support. Every Ollama model reports its own capabilities. Every MLX checkpoint declares its…
We built a benchmark to decide which local models earn a place in the fleet and which ones get retired. Then we discovered the benchmark had been quietly rigging the contest, and that one of its victims had already been publicly recommended for deletion. What we were measuring The preliminary suite is deliberately small and…
