$ whoami
I run Revitt, a software business, and build agent tooling and local model experiments under Don Works. This is where I write up the things we built, what worked and what didn't.
How we kept Qwen3.8-27B going for 457,000 tokens without compaction by evicting and restoring its attention cache on MLX and llama.cpp, what that did and didn't prove, and where MDBX fits in.