Lab Notes
Investigations that sit between formal releases. Each note follows a question through an experiment, diagnosis, or negative result and records the boundary of what the evidence supports.
Assistant-drafted lab notes
- 26–27 March 2026Infrastructure investigation
Moving a proof-search model to Tenstorrent
Moving DeepSeek from MLX to one Tenstorrent QuietBox with four dual-Wormhole N300 cards, and finding the attention-sharding assumption that initially prevented it.
- 2 April 2026Experimental diagnosis
An unused tactic changed the proof result
Why an apparently controlled theorem-proving experiment could not yet support a causal interpretation.
- 4–8 May 2026Negative method result
Can a brain-response model measure lifelikeness?
Why a failed biological-motion control made us abandon an appealing interpretation before running the full comparison.
- 19–27 May 2026Cross-system analysis
What a shape map leaves out
What a shared map of Lenia, embryo and fish forms preserved, what it erased, and why the strongest fish-like lead remained unconfirmed.
- 15–20 June 2026Research prototype
Running morphogenesis in reverse
Searching for rules that produce a requested body, using recorded developmental trajectories.
- 19 July–1 August 2026Technical investigation
A faster database query was not enough to retire the archive
Separating missing lineage from empty data, then making the replacement database fast, checkable, and incapable of authorizing deletion by accident.
- 5–8 August 2026Negative pilot
Stripes, dust, and false fission
How stripes, dust, and mutation artifacts forced a change in experimental design.
- 16–29 August 2026Developmental result
Older Lenia organisms resist the same disturbance
Separating simple growth from a measured change in developmental susceptibility.