danielmiessler / danielmiessler/TheAlgorithm

RFC: Phase-Transition Gates for ISC traceability

Aperta
#8 0 commenti 1 reazione 0 assegnatari Vedi su GitHub
Lingua principale
Nessun dato sulla lingua
Stelle
154
Fork
16
Metriche di merge delle PR
Nessuna PR unita negli ultimi 30g

Descrizione

## Problem

I've been using TheAlgorithm daily and noticed 3 failure modes that occur between phases:

1. **Silent mutations** — THINK's pressure test sometimes changes ISC criteria, but there's no explicit checkpoint tracking what changed. PLAN can end up working against stale criteria.

2. **Scope creep in BUILD** — The model creates artifacts that weren't in the execution plan and have no ISC backing. The work gets done, but it's untracked and unverifiable.

3. **Rubber-stamp verification** — VERIFY runs and produces "PASS" claims, but the verification methods aren't actually executable at that point (e.g., server not running, test file doesn't exist).

v1.6.0 has strong entry gates (Quality Gate after OBSERVE) and exit gates (Verify Completion Gate before LEARN), but the transitions between THINK→PLAN→BUILD→EXECUTE→VERIFY have no checkpoints.

## Proposed Solution: 4 Phase-Transition Gates

Simple checklist patterns (no new tools) that sit between existing phases:

| Gate | Location | Catches |
|------|----------|---------|
| **Mutation Gate** | THINK → PLAN | Unacknowledged ISC changes from pressure test |
| **ISC-Map Gate** | PLAN → BUILD | Plan steps without ISC backing, ISC without plan steps |
| **Drift-Check** | During BUILD | Artifacts being built without ISC mapping |
| **Verify-Readiness Gate** | EXECUTE → VERIFY | Verification methods that aren't actually executable |

Each gate is OPEN or BLOCKED — binary, no ambiguity.

## Full Spec

I've written up the complete gate specifications with rationale and examples:

👉 https://gist.github.com/Scaniani/827cc432c0db325ac52673141c7bfd2b

## Design Principles

- **Zero new tools** — These are patterns, not infrastructure
- **Effort-scaled** — Mental checks at Fast/Standard, explicit output at Extended+
- **Backwards-compatible** — They extend v1.6.0, don't modify existing gates
- **Tested** — I've been running these daily for ~2 weeks across different task types

## Questions for Discussion

1. Would this be useful upstream, or is this too opinionated for the core spec?
2. If useful — should these be in the main spec or a separate "gates" module?
3. The ISC-Map Gate is the most impactful in my experience. Would a single-gate PR be preferred over all 4?

Happy to submit a focused PR if there's interest. Just wanted to check alignment first.

Guida per i contributori

Nessuna guida per i contributori indicizzata per questo repository

Direzione di ricerca

Leggi le specifiche complete dei gate collegate e confrontale con i gate di ingresso e di uscita esistenti della v1.6.0 descritti nell’issue. Determina se la proposta appartiene alla specifica principale o a un modulo separato dei gate e se debbano essere adottati tutti e quattro i gate o un ISC-Map Gate mirato; il lavoro è completato quando l’ambito e la collocazione sono concordati e documentati.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Valutazione

Ambito
documentation
Tipo di issue
Funzionalità
Difficoltà
5/5
Tempo stimato
Più di una settimana
Stato di attività
Ferma
Chiarezza
Da chiarire
Idoneità per principianti
25/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.