MacBrains Labs / Field notes
Field notes.
Short, source-bound examinations of what passed, what diverged and which earlier hypotheses failed. Each note names its artifact, its observation method and the records it rests on.
- N/001
28 layers, LUT4, and the cost of exact placement
State reads, indexed slices, weight.bin offsets and the non-ANE tail.
- N/002
A passing production graph and a failing diagnostic one
LFM block-pair parity, 743/743 assignment and why probes alter the experiment.
- N/003
The bandwidth limit that was not a limit
Batch sweeps, dispatch floors and a hypothesis overturned by measurement.
- N/004
Every slice passed. The chain did not.
LFM2.5-1.2B layer slices meet a strict ANE parity gate; the 16-layer chain and token logits miss it.
- N/005
Correct, placed, paid for
A ring-buffer attention graft placed nothing on the ANE; an EAGLE draft head made every call slower and has to earn it back.