llm
- llama.cpp's MoE cache wants gigabytes and long prompts
On a 24 GB RTX 3090, llama.cpp's MoE expert cache read long prompts up to 85% faster, did nothing for a Q8 model, and left a 2-4x gap to Strata.
- Flint mistyped its path and blamed the tools
Flint typed one wrong token in its workspace path. OpenCode refused 81 calls without naming the path; the model blamed a tool glitch and called its code done.
- MTP acceptance is a side effect, not a target
A friend said MTP acceptance near 0.8 meant an idle GPU and I should draft more. I swept draft length on a 27B. Acceptance fell; speed peaked early.
- Claude Code read my home folder as a project
The bench's Claude Code cell moved its config dir and excluded the repo's CLAUDE.md. My global ~/.claude still loaded, labelled as project instructions.
- The model card said think harder. The audits got worse
Qwen3.8's card defaults to xhigh and warns lower effort can mean more failures. xhigh passed 1 of 24 audits and 24 of 36 coding runs, against 4 and 27.
- It fits in 16 GB. That was the easy question
Qwen3.8-27B with its KV cache in system RAM fits a 16 GB card. Generation drops from 42 to 15 tokens/s at once, and to 1.5 at 200k context.
- The hard tier sorted by a rollback, not by speed
Five fast MoEs passed 1 of 45 hard-tier runs and skipped the same rollback in 33 of 45 ledger runs. A fast 2-bit model passed 7 of 9.
- The hidden test asked for more than the prompt
39 agent runs on a Java ledger task were marked down by one hidden check. The prompt never asked for it, and the project's README steered them into failing it.
- The harness decided how long the model thought
One 27B model, one llama.cpp server: 250 s per agent run in OpenCode, 116 s in Pi. Under OpenCode the model reasoned 3.6 times as much.
- 29 of 34 passing agent runs had a blocker the hidden tests never saw
Blind review packets, a merge rubric and probes that must run: 34 agent-written money transfers passed the hidden tests, and review found a blocker in 29.