Posts with this tag: 4
A PreToolUse gate and a capability marker made the preferred discovery tool the first move, with 14/14 checks passing. A later wrapper fix restored byte-for-byte passthrough for repeated payloads.
ReadA LoRA patch scored 89.06% on its worst prompt against a frozen 90% floor and looked FAILED. The unpatched base, run against the same floor for the first time, also failed at 89.06%, on a different prompt. The bar was unreachable by construction. Llama.cpp's own f16-KV configuration misses its own f32 reference the same way, at 5 of 7 tested lengths.
ReadA cross-node state-transfer gate scored corrupted states as correct because the ids downstream of a bad K/V happened to survive. Comparing raw bytes instead found save_state exporting one prompt's conversation and SSM state under a different prompt's checkpoint, fixed the same day.
ReadAn audit of this project's own optimization loop built four candidates designed to cheat its performance gate. One of them, a one-line deletion of a GPU synchronization call, passed at a fabricated plus sixty-three percent. The fix moved the stopwatch out of the file the candidate is allowed to touch.
Read
