The worm remembers where it was fed; the route into behaviour is measured and waits - #216
Merged
Merged
Conversation
…ured and waits Thermotaxis memory (issue #198): the cultivation temperature stops being a declaration and becomes an experience. Senses.t_setpoint is an on-food-weighted slow average of experienced temperature -- dopamine-gated like the sleep homeostat's build, exact first-order so it does not depend on the step size, tau about two sim-minutes at the dense lawn's dopamine (the same clock compression the rest of the dish runs on). Measured: a feeding animal walked its setpoint 20.000 -> 19.729 C in three minutes, tracking what it actually experienced; a fasting animal keeps the declared value to the bit. The off-state test earned its keep the day it was written. The first cut learned on any positive dopamine, and a fasting animal quietly wrote its own memory -- the mechanosensory dopamine level idles at 0.029 on bare agar, never zero. "Fed" is now a threshold (0.08, measured above the off-food ceiling and below the sparsest lawn), and below it the setpoint is frozen exactly. The route into behaviour is refused for now, on evidence rather than caution. Four measurements (tools/thermo_memory.py): the signed tonic into AFD separates cultivation histories robustly and AWAY from home; its mirror and the rectified cold-side form are noise; and the primitive itself is flat -- a constant current held into AFD, -20 to +40 pA for 200 s, leaves end temperature unmoved on the gradient. AFD steers through its transient, phase-locked with the animal's own movement; a standing offset carries no phase information a klinotaxis can use. The recorded follow-up is the biology's own answer: the setpoint must CONDITION the differential -- warming is good below home and bad above it -- and that redesign wants its own paired gate. Until it wins one, thermo_setpoint_gain ships at 0.0, pinned in the runtime registry, and the memory ships live: learning, costing nothing, waiting for a route that works. Closes #198. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CBroMXcYdfkyb5buJPcKvL
YesterdaysLemon
marked this pull request as ready for review
August 27, 2026 22:09
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #198.
The memory
cultivation_tempstops being a declaration.Senses.t_setpointis an on-food-weighted slow average of experienced temperature — dopamine-gated like the sleep homeostat's build, exact first-order (step-size independent), tau ≈ 2 sim-minutes at the dense lawn's dopamine, the same clock compression the whole dish runs on (Hedgecock & Russell 1975; Mori & Ohshima 1995 — the existence of the memory is solid biology; the rule is modelled; the rate is the dish's).Measured: a feeding animal walked its setpoint 20.000 → 19.729 °C in three minutes, tracking what it actually experienced. A fasting animal keeps the declared value to the bit.
The off-state test earned its keep the day it was written
The first cut learned on any positive dopamine — and a fasting animal quietly wrote its own memory, because the mechanosensory dopamine level idles at 0.029 (max 0.045) on bare agar, never zero. "Fed" is now a threshold (0.08 — measured above the off-food ceiling, below the sparsest lawn's 0.10), and below it the setpoint is frozen exactly.
The routing: refused on evidence, not caution
Four measurements in
tools/thermo_memory.py:(T − setpoint)into AFD, gain +5: histories separate (−1.18 °C — the signature no fixed mechanism can fake!) but away from home — AFD excitation steers this connectome warm-ward.The recorded follow-up is the biology's own answer: the setpoint must condition the differential — warming is good below home, bad above — and that redesign wants its own paired gate. Until it wins one,
thermo_setpoint_gainships at 0.0, pinned inRUNTIME_UNSUPPORTED, and the memory ships live: learning, costing nothing, waiting for a route that works.Bookkeeping
New behaviour test (fed learns / fasting frozen-to-the-bit) registered in CI targets; every determinism pin held; NEXT.md roster updated. No runtime port needed — nothing behavioural changed at the shipped gain.
🤖 Generated with Claude Code
https://claude.ai/code/session_01CBroMXcYdfkyb5buJPcKvL
Generated by Claude Code