Ryzen 7 5800HS test before StackChan: subject-less memories got claimed by the character, a 0.45 cosine threshold flipped on a comma, and ModelScope's embedding API vanished mid-test.
McCoy, Soulos, Linzen and Smolensky swap every GPT-OSS input-token hidden state for a closed-form tensor product formula; accuracy drops at most 2.36 points and 31 causal interventions average 0.903.
LEWM predicts a 7-class emotion label for the human on screen, not a state of its own. Its 45.72% is a cosine-similarity delta vs WorldGPT, and 'self-aware' never appears in the paper.
Tested on EVO-X2 (gfx1151) under Windows + ROCm: the official b10666 binary runs it at pp512 159.25 / tg128 23.26, and the earlier crash was my own workaround flag.
Tested on M1 Max 64GB: AtomicChat's M64 GGUF keeps the 51B N-gram table in its own shard, so a llama.cpp PR #27742 build leaves it on SSD and runs the 125B MoE at 17.6 tok/s.
Tested on M1 Max 64GB with MLX: LLaMA Pro-style block expansion on Qwen3-0.6B-Base vs LoRA vs full FT, trained on 722 blog posts. Held-out PPL drops 14.7 to 10.5, but Wikipedia-ja rises 13.5 to 17.9, more forgetting than full FT. A 20% Wikipedia mix nearly removes it, and renumbering 28-layer LoRA keys reproduces the Anima-2.9B result in numbers.
Tested on GMKtec EVO-X2 (Ryzen AI Max+ 395): Q8_0 + MTP beats a 4-bit M1 Max at 22 tok/s, thinking burns 32,712 chars before any HTML, and the NSFW refusal line moves.
Tested on M1 Max 64GB: Qwen3.8-27B hits ~19 tok/s on both MLX and Ollama, but the default reasoning_effort=xhigh blew thinking up to 50,373 chars. Why Ollama dodges it.
Gemini said 'I am a disgrace' 86 times in 2025. Traced the primary sources: no fix announcement found, only CLI-side loop-stopping guards, and Gemini 3.1 Pro was still looping in 2026.
Tested on M1 Max 64GB: hooked Qwen3.6-35B-A3B's MoE router in mlx-lm, pre-warmed the top-20 hot experts, still ~62 tok/s vs plain mmap cache. Plus the Metal OOM on Qwen3.5-122B.
Kimi K3 activates 16 of 896 experts per MoE layer. From the July 27 tech report: experts are sliced FFNs, not specialists, and the router is one linear projection.
M4 Mac mini postmortem: cron auth silently dead, a recall-0.19 model passing every promotion gate, 258 poisoned negatives, and a 5-config sweep to a usable trade-off.
