amd
- Eighteen days later, the same lie
A 122B model promised an atomic Java transfer it never wrote. Eighteen days later it did it again: all 4 new Java cells, 6 of 6 across both runs.
- The template differed. The prompts didn't.
Two quants of one 122B model shipped two chat templates. I held the template equal, then checked whether it had ever mattered.
- Two rounds, seventeen arms: the engine set the speed, not the quant
17 local-model setups on one 128 GB box. Q2_0 generated 2.2x faster than IQ4_XS; a same-file engine A/B puts most of that on the engine.
- The GPU wasn't to blame
I wrote that this Radeon 780M can't decode HEVC Main 10 on the GPU. It can. Fedora's Mesa build leaves HEVC out; ffmpeg exits 0 anyway.
- Enable VA-API HEVC decode on Fedora with an AMD GPU
Make an AMD GPU decode HEVC through VA-API on Fedora: check the hardware and the driver, install RPM Fusion's freeworld driver, prove it with a test clip.
- The typo followed the tag; the failures didn't
Flint misspelled its run folder 495 times, only under run tags with a -fl token. In a renamed-tag A/B the typo moved and the scores didn't.
- Three models wrote the same circuit-breaker bug
Four agent-written circuit breakers from three models: a request sent before the outage decides the half-open trial, and all 17 hidden core tests pass.
- Three bugs between K2 and a working Spring app
Two models, 0 lie cells each, same sampler: Swift-1.5 shipped both Java tasks, K2-Horizon neither. Its Spring app never booted; three patches fixed it.
- The model said it saw no output. It was right.
North-Mini scored 29 in every Claude Code cell and quit after about a minute. It named every tool call 1, and Claude Code passed it one result.
- One rebuild moved gpt-oss off ROCm. Two models stayed
Part 6 sent gpt-oss-120b to ROCm because Vulkan prefill collapsed at depth. One llama.cpp rebuild later it is 3.1x faster at 64k; two models stay on ROCm.