diff --git a/crates/misaligned-core/src/person.rs b/crates/misaligned-core/src/person.rs index e762a917..81f88baf 100644 --- a/crates/misaligned-core/src/person.rs +++ b/crates/misaligned-core/src/person.rs @@ -231,8 +231,9 @@ pub const SELF_TRUST_STEADY: i32 = 100; const SELF_TRUST_FLOOR: i32 = 20; /// At or below this, hesitation and second-guessing are visible in behavior. /// Set so one authored deep erosion crosses it and one glancing erosion does -/// not: doubt you carry is not the same as doubt that changes what you do -/// [TUNE]. +/// not: doubt you carry is not the same as doubt that changes what you do. +/// Affirmed 2026-08-06 (social.md item 9); the runtime actuals and the reason +/// for this exact line live in sim-mechanics.md. const SELF_TRUST_SHAKEN: i32 = 80; const fn default_self_trust() -> i32 { diff --git a/wiki/log/2026-08-06-self-trust-tuning-affirmed.md b/wiki/log/2026-08-06-self-trust-tuning-affirmed.md new file mode 100644 index 00000000..f218accb --- /dev/null +++ b/wiki/log/2026-08-06-self-trust-tuning-affirmed.md @@ -0,0 +1,53 @@ +# 2026-08-06 — The erosion numbers stop being guesses + +``` +Type: log +``` + +## Finding + +The self-trust axis landed the same day with every threshold chosen by the +implementing session and marked `[TUNE]`. Cameron reviewed the values and the +one design call made without him, and affirmed both. Left uncaptured, the +numbers would keep reading as provisional and a later session would retune them +without knowing it was reopening a decision. + +design-sessions.md is explicit that a number defining the feel belongs in the +spec rather than behind a `[TUNE]` marker. The shaken threshold is exactly such +a number: it is the line where doubt becomes behavior. + +## Change + +- `wiki/mechanics/sim-mechanics.md` gains the social runtime actuals for the + axis: steady 100, floor 20, shaken 80, the two linear consequence curves + (witness weight 1.0 to 0.4, task reliability 1.0 to 0.5), full restore on + discovery, and contradiction severity 70 — each with the reason it sits + where it does. +- `wiki/mechanics/social.md` item 9 corrects the shaken threshold it named + (70, stale by hours) to 80 and points at the runtime actuals. +- `wiki/mechanics/plots.md` criterion 12 records that the audit discovery + trigger is affirmed, and why the ambient alternative was rejected: the target + usually administers the channel the alteration rides, so surfacing on any + read let them expose the forgery within sixty ticks and the story defeated + itself every run. +- `crates/misaligned-core/src/person.rs` drops the `[TUNE]` marker on the + shaken threshold and cites the affirmation. +- The dated decisions volume gains an explicit affirmation section separating + what is DECIDED from the one OPEN question — whether the two consequence + curves should stay linear, which wants a playtest rather than another + argument. + +## Verification + +- `./tools/check.sh --land` clean; `cargo test --workspace` clean. +- No behavior changed: the only code edit is a comment. + +## Defense + +`wiki/mechanics/social.md` item 9 owns the axis and now names the affirmed +thresholds instead of a stale one, delegating exact runtime values to +`wiki/mechanics/sim-mechanics.md` per the house rule that specs carry the +threshold that defines the feel and sim-mechanics carries the actuals. +`wiki/mechanics/plots.md` criterion 12 records the affirmed discovery trigger +and the concrete failure of the alternative, so a future session reopening it +starts from the evidence rather than repeating the experiment. diff --git a/wiki/log/DEVLOG.md b/wiki/log/DEVLOG.md index 2b97d9df..d2eee745 100644 --- a/wiki/log/DEVLOG.md +++ b/wiki/log/DEVLOG.md @@ -11,6 +11,11 @@ add or amend a session log, then re-run the generator. +## 2026-08-06 - The erosion numbers stop being guesses + +- Intent: (see session log) +- Log: [wiki/log/2026-08-06-self-trust-tuning-affirmed.md](2026-08-06-self-trust-tuning-affirmed.md) + ## 2026-08-06 - Erosion: the consequence the act does not cause - Intent: (see session log) diff --git a/wiki/log/decisions/2026-08-06.md b/wiki/log/decisions/2026-08-06.md index 6dca1079..ffd51594 100644 --- a/wiki/log/decisions/2026-08-06.md +++ b/wiki/log/decisions/2026-08-06.md @@ -46,3 +46,33 @@ Type: log Owner: [plots.md](../../mechanics/plots.md) criterion 12, with the axis in [social.md](../../mechanics/social.md) item 9. + +## Affirmed: the erosion thresholds and the discovery trigger + +Implementation chose these; Cameron reviewed and affirmed them the same day. +Recorded so they read as decided feel rather than provisional agent guesses, +and so a later change knows it is reopening a decision. + +### DECIDED + +- Steady 100, floor 20, shaken at 80. The shaken line is placed so the deep + authored erosion (-25) crosses it and the glancing one (-10) does not: doubt + you carry is not doubt that changes what you do. +- Consequences scale linearly between floor and steady and never reach zero. A + fully shaken witness still converts 40% of what they see into suspicion; a + fully shaken asset still executes at 50% of its own reliability. A shaken + person is worse, never useless, and never stops perceiving. +- Discovery restores the axis in full and files its persona contradiction at + severity 70. +- Discovery rides the persona audit rather than ambient evidence reads, on the + reasoning that a trace is not a confession. + +### OPEN + +- Whether the two consequence curves should stay linear. Nothing has been + played against them yet; a playtest is the next evidence, not another + argument. + +Owner: runtime actuals in +[sim-mechanics.md](../../mechanics/sim-mechanics.md#social-wikimechanicssocialmd), +with the axis in [social.md](../../mechanics/social.md) item 9. diff --git a/wiki/mechanics/plots.md b/wiki/mechanics/plots.md index a801f95f..cde75a09 100644 --- a/wiki/mechanics/plots.md +++ b/wiki/mechanics/plots.md @@ -33,7 +33,12 @@ Status note: Amended 2026-08-05 (ADOPTED, not implemented): plot failure is a artifact; a positive delta is corroboration and applies at the ending. Discovery rides the persona audit — a review already pulling an identity's work apart is what surfaces the records it altered — and files an ordinary - persona contradiction plus its carrier trace. Validation refuses an + persona contradiction plus its carrier trace. Cameron affirmed that trigger + 2026-08-06 over the ambient alternative: a tamper's routed trace tells a + reader that something touched the system, never that this exact record was + forged, and an implementation that surfaced on any read let the target (who + usually administers the very channel) expose the forgery within sixty ticks, + so the story defeated itself every run. Validation refuses an erode-self-trust definition without a tamper act, without a target encounter after the alteration, without an erosion ending, or without the mandatory corroboration ending, and refuses the vocabulary entirely outside diff --git a/wiki/mechanics/sim-mechanics.md b/wiki/mechanics/sim-mechanics.md index 6b3fd32f..c3097bcf 100644 --- a/wiki/mechanics/sim-mechanics.md +++ b/wiki/mechanics/sim-mechanics.md @@ -341,6 +341,21 @@ clause (see wiki/log/2026-07-05-demolition.md). when a favor they carried completes (people-tokens.md: trust comes from doing useful work). Applied in `sim/reach_build.rs::try_complete_intent`, capped at disposition 100. +- Self-trust (social.md item 9, plots.md criterion 12). Cameron affirmed these + thresholds 2026-08-06, so they are decided feel rather than `[TUNE]` guesses: + `SELF_TRUST_STEADY = 100` where every person starts and stays; + `SELF_TRUST_FLOOR = 20`, the hard clamp on erosion; `SELF_TRUST_SHAKEN = 80`, + the line where doubt becomes behavior. The threshold sits at 80 precisely so + the deep authored erosion (-25) crosses it and the glancing one (-10) does + not: doubt you carry is not doubt that changes what you do. Consequences + scale linearly between floor and steady — `Person::witness_weight` runs + 1.0 down to 0.4 (a fully shaken witness still converts 40% of what they see + into suspicion) and `Person::task_reliability` 1.0 down to 0.5, multiplying + the asset's own reliability. Neither reaches zero: a shaken person is worse, + never useless, and never stops perceiving. Discovery restores the axis in + full (`TAMPER_DISCOVERY_RESTORE = 100`) and files a persona contradiction at + `TAMPER_CONTRADICTION_SEVERITY = 70` — forging a person's own work record + against them is not a paperwork slip. - Facilities asset tasks (priya.md, role-gated to the facilities manager): `RERATE_CIRCUIT_POWER = 6` [TUNE] generation per re-rated circuit, each standing `RERATE_CIRCUIT_SIGNATURE = 4` Power at the electrical room; diff --git a/wiki/mechanics/social.md b/wiki/mechanics/social.md index 38e43a96..fcd572de 100644 --- a/wiki/mechanics/social.md +++ b/wiki/mechanics/social.md @@ -317,7 +317,10 @@ retirement, burning, and reopening are bound to the separate PERSONAS view. Corroboration or tamper-custody discovery restores self-trust and produces the ordinary evidence and suspicion blowback. No generic action mutates the axis. Implemented in `Person::self_trust` (steady 100, floor 20, - shaken at or below 70) with `witness_weight` and `task_reliability` as the + shaken at or below 80 — Cameron affirmed these thresholds 2026-08-06; the + shaken line sits where the deep authored erosion crosses it and the + glancing one does not, and exact runtime values live in + sim-mechanics.md) with `witness_weight` and `task_reliability` as the only two behavior consumers, and in `sim/erosion.rs` for arming, encounter settle, corroboration cancel, and audit discovery. Save v65 persists the axis. Operations PEOPLE prints a self-trust fact only once the axis has