Full ZASS

Current version: v0.3.7 ZASS SYSTEM: v0.1.0

Full ZASS is the deeper decision-control method for projects where decisions, evidence, risks and architecture interact.

v0.3.7 aligns the Full ZASS surface with the ZASS SYSTEM v0.1.0 contract: progressive disclosure, next-meaningful-action UX, factual SAVE/sync receipts, and shared local/Web core semantics. Core decision semantics remain unchanged.

Core principles:

Don't shortcut thinking; eliminate repeated thinking.

AI produces possibilities. Evidence tests them. Humans decide. Architecture follows the decisions.

Capture broadly, converge deliberately. Form candidates first; research only what can change the choice.

ZASS Convergence Loop

CAPTURE → MATCH → SYNTHESIZE → RESEARCH → CROSS-CHECK → LOCK → ARCHITECTURE

Research follows candidate formation, not idea capture. Capture ideas, constraints, risks, evidence and existing LOCKs without filtering too early; match complementary ideas; synthesize coherent candidates; shortlist against must-haves; research only decision-changing uncertainties; cross-check against real capabilities and evidence; update the matrix/candidate state; then let the owner LOCK decisions before architecture.

See ZASS Convergence Loop for the expanded workflow.

Mental model

RAW IDEA
    ↓
EXPLORE
    ↓
QUESTIONS / RISKS / OPTIONS
    ↓
TEST when needed
    ↓
HUMAN DECISION
    ↓
LOCKED
    ↓
DRAFT ARCHITECTURE
    ↓
REVIEW
    ↓
YA, CONFIRM ARCHITECTURE
    ↓
CONFIRMED ARCHITECTURE

The authority boundary

AI suggestion
      ≠
Owner decision
      ≠
Git change

AI may propose.

Only the owner approves decisions.

A Git change is real only after it is actually persisted.

Main project records

Full ZASS can track:

• RAW IDEA
• GOALS
• NON-GOALS
• CONSTRAINTS
• IDEA candidates
• QUESTIONS
• RISKS
• METHOD REVIEWS
• OPTIONS
• ARCHITECTURE CANDIDATES
• DECISION LEDGER
• LOCKED DECISIONS
• REJECTED ideas
• DEFERRED items
• OPEN LOOPS
• EXPERIMENTS / EVIDENCE
• ARCHITECTURE READINESS
• EVIDENCE CONFIDENCE

The owner should not have to memorize these IDs during normal conversation.

Main commands

ZASS / ZASS!!

Map and explore the relevant discussion without silently changing LOCKED decisions.

PROCEED

Full ZASS v0.3.5 uses an explicit approval set.

Before offering PROCEED, AI must show:

PROPOSED FOR PROCEED
- [ID / action / change]
- [ID / action / change]

PROCEED approves exactly that set.

Unlisted suggestions are not approved.

If the set changes, conflicts or becomes ambiguous, the AI must show the revised set and wait for a new PROCEED.

PROCEED does not commit or push.

PIVOT

Explore a different direction while preserving history.

COMMIT

Persist already-approved changes as an atomic Git commit and push to the GitHub Source of Truth.

Do not report success before the real commit/push succeeds.

DRAFT ARCH

Prepare or revise a working architecture draft from the authoritative state.

BUILD ARCHITECTURE

Show the draft, relevant LOCKED decisions, readiness, critical assumptions and blockers.

If ready, request the exact owner confirmation:

YA, CONFIRM ARCHITECTURE

PARK

PARK is no longer a Full-ZASS command.

PARKED remains a valid internal/execution state for deferred work.

IDs

Common Full-ZASS IDs include:

I-xxx   Idea
Q-xxx   Question
R-xxx   Risk
MR-xxx  Method Review
AC-xxx  Architecture Candidate
D-xxx   Decision
L-xxx   Locked Decision
E-xxx   Experiment

Source of Truth

For Git-backed work:

GitHub is authoritative.

Do not silently replace the latest repository state with an older chat, memory or handoff.

Authoritative files:

• ZASS.md
• ZASS_EN.md