[ishaansathaye]

What changes

Adds a deterministic, Rust-owned spike-timing experiment as the scientific foundation for doke's next phase. Three cue cells connect to a target. Eight close cue-before-target pairings strengthen the selected connection enough to change a subsequent cue-only probe; reversed or distant pairings produce different calculated outcomes, while untouched cues remain controls.

Training explicitly prescribes both spike trains and clamps the target between imposed spikes. Probes remove that clamp, freeze learning, clear transient state and retain learned weights. This avoids reporting requested timing when an already-strong cue would actually trigger an extra target spike.

The model uses a 1 ms event clock, exact passive leak, 10 ms transmission delay, bounded additive all-to-all STDP traces, and an explicit same-tick convention. Inputs and queue sizes are bounded; invalid requests leave state unchanged.

Research and next phases

docs/NEXT_PHASE.md documents the original interaction design, equations, authored parameters, source-to-claim mapping, limitations, acceptance criteria, and later experiments in short-term recovery, delayed modulation, homeostasis and controlled comparisons. Phloem inspired the input/history/consequence interaction principle; no code or assets were copied.

This PR contains the engine, research contract and GitHub verification workflow. The dependent visual-experience PR #2 connects it to the website. Review this PR first, then retarget #2 to main after approval and merge of this foundation. Existing atlas and population-model behavior is preserved.

Validation and review

• cargo test -p doke-core --locked: 15 passed, including 11 new numerical/behavioral tests.
• cargo clippy -p doke-core --locked -- -D warnings: passed.
• Tests cover analytic timing windows, arrival-clock semantics, zero-lag ordering, passive leak/refractory behavior, trained versus untouched probes, reversed pairing, saturated induction, frame batching, invalid-input atomicity, cancellation/reset and long-run bounds.
• Parent review checked the equations/event ordering and their agreement with the research contract; a separate neuroscience agent reviewed the final specification.
• CI tests the core, checks the Wasm target, validates source records and checks browser-script syntax.

The model is pedagogical and uncalibrated, not a human-brain or chemical-concentration prediction. Awaiting owner review; do not merge automatically.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.