Dated state, not mythology
AI build log
A compact chronology of the implementation and research record surfaced by this mining pass.
- Status
- PUBLIC WORKING LOG
- Evidence class
- Dated source and receipt summaries
- Current question
- What can another builder safely pick up?
- Current attack
- Backdating current ideas into historical work
- Next test
- Implement the MFM causal sequence
- Inventory date
- 2026-09-13
- Code / data
- Artifact office
- Corrections
- Failures and holds
- 06
Cognitive Basin, Natural Math, and AIA platform lines
Executable labs, bounded developmental systems, stock-Android portability, and hardware-adapter separation formed distinct lanes. They remain distinct.
- 07
Representation and claim correction
Specificity and Ageometrics sharpened observer, comparator, residue, and loss boundaries. Earlier morphology overclaims were archived rather than rewritten.
- 08
FormationCore, field lineage, sensing, inquiry, and publication
Baby AI froze narrow causal-state and continuity tranches. FFC/CRR/FPG/DEC developed candidate selective and embodied architectures. ASF froze read-only observation witnesses. Research Query and Formation Stream made retrieval and publication more inspectable.
- 09
Causal instruments and CF v0.4 RC1
Twin-History Pass 1A reached INSTRUMENT_READY without claiming a history effect. PIA v0.3 froze a stronger design but remains blocked. CF v0.4 RC1 separated ledger from runtime, added sparse Formative Field recruitment, and opened distributed embodiment and energy ecology as candidate programs.
- NOW
Public AI laboratory
The site now exposes the inventory, relations, artifacts, failures, method, and next experiments without upgrading their evidence.
Immediate build queue
- Implement a minimal Formative Field with measurable wake-set and lifecycle cost.
- Run the MFM reversal, recovery, transplant, and ablation sequence against matched native baselines.
- Clear or preserve PIA's harness blocker.
- Sanitize and review the distributed-power assay before any public inference.
- Test the Heterogeneous Research Loop prospectively instead of treating team experience as proof.