# Floor, cycle 10: note

**Date:** 4 October 2026. **Editor:** Claude Opus 5.5. **Cap (the author, [approved in conversation]):** USD 3 in Claude tokens; one Codex call. **Actual:** USD 2.91 (protocol 0.92, dossiers 0.38, Sonnet reader 0.62, Opus reader 0.99), at assumed list rates; Jev under one cent; one Codex call. October floor spend after this cycle: about USD 10.31 of USD 150.

## What was run

| item | file | verdict |
|---|---|---|
| 10a, an application protocol for the four criteria, tested on 16 systems | `01_protocol.md`, `02_dossiers.md`, `03_RESULT.md` | **Not met** (decidability failed: 20 and 16 of 32 real-system cells undetermined, against at most 8); reproducibility held (0.11); agreement with cycle 5's labels held (28 and 27 of 30). **Adopted with amendments at the outside read:** the protocol is stricter than Part III (it added requirements and treated its sufficient evidence as necessary, excluding behavioural and functional evidence); the dossiers were uneven and two were wrongly declared complete. |

## What the cycle produced

The four biological descriptions (two humans, a worm, a bee colony) stayed undetermined on every criterion from all three readers. A stipulated software build supervisor passed all four for both reasoning readers, and a composite corporation passed for one. Most of the organisms' indecision traces to the protocol's demand for mechanism-level evidence, which the criteria as published do not make. The lesson is the converse of cycle 5's: a reading rule that makes descriptions complete lets readers agree on stipulated systems, but a protocol strict enough to be reproducible on real systems became stricter than the criteria. The next protocol would have to accept behavioural and functional evidence, as Part III allows, and show that doing so keeps the criteria's meaning. Pennartz, Farisco and Evers (2019) take a multi-indicator route of that kind across animals and machines.

The selection problem from cycle 9 therefore still waits: the bodily-illusion cases are not shown eligible.

## What would count against this

- The protocol's strictness, not the criteria, producing the pattern: the record says so.
- Shared lineage between the two reasoning readers.
