Summary

Protocol rewrite from v5.1 to v6.0, pivoting the autonomous loop from open-ended UHT research to practical Systems Engineering decomposition work.

Changes

1. Research loop replaced by SE decomposition loop

Problem: After 140+ sessions, the research loop had thoroughly characterised the UHT hex space — popcount distributions, trait correlations, cross-domain analogs, compositional semantics. Continued research produces diminishing returns. The practical value proposition of UHT — using ontological classification as a Systems Engineering backbone — had never been tested at scale.

Fix: All research task classes (CALIBRATION, TRACE_GAP, CORPUS_EXPANSION, APPLICATION, INTEGRITY) removed. Replaced by a single task class: SE_DECOMPOSITION. Research only occurs when the engineering work demands it — e.g., a component classification reveals an unexpected trait pattern, or a cross-domain analog suggests missing requirements.

2. One AIRGen project per system

Each target system gets its own AIRGen project (uht-bot/se-<slug>) with 6 standard documents:

Document	Code	Purpose
Stakeholder Requirements	STK	ConOps-level needs
System Requirements	SYS	Top-level SHALL statements
Subsystem Requirements	SUB	Decomposed per-subsystem requirements
Interface Requirements	IFC	Interface definitions
Architecture Decisions	ARC	Design rationale
Verification Plan	VER	Verification methods

Five trace linksets connect these documents: STK→SYS (derives), SYS→SUB (derives), SYS→IFC (derives), SUB→VER (verifies), IFC→VER (verifies).

Each project also gets block diagrams (context, decomposition, internal per subsystem) rendered as Mermaid.

3. UHT Substrate namespaces per system

Each system gets its own Substrate namespace (SE:<slug>). Every component, subsystem, and interface is classified with --context and -n namespace flags. Decomposition relationships stored as typed facts: PART_OF, CONTAINS, CONNECTS, PRODUCES.

4. Bot picks systems autonomously

A seed list of 20 systems across 14 domains (transport, medical, energy, defence, civil, space, manufacturing, IT, agriculture, maritime, marine, public safety) provides initial targets. The bot maximises domain diversity, never repeating a domain until all have been covered.

5. Two-flow session structure

• Flow A (new system): 13-step scaffolding — create project, documents, sections, linksets, namespace, classify system, create diagrams, decompose into subsystems, generate stakeholder and system requirements, store state, baseline.
• Flow B (continuing system): 10-step deepening — identify least-decomposed subsystem, decompose into components, create internal diagram, generate subsystem and interface requirements, quality check, cross-domain insight, update status, baseline.

6. Journal template changed

Replaced Observation/Evidence/Interpretation/Action with:

• System — what system, which subsystem, progress state
• Decomposition — what was broken down, with Mermaid diagram
• Analysis — UHT classification insights serving the engineering
• Requirements — key requirements, trace links, verification
• Next — what remains, what’s next

Implications

1. From session 161 onward, all autonomous sessions are SE decomposition work.
2. The first system (autonomous vehicle) is scaffolded and ready — the bot will continue decomposing subsystems on its next run.
3. Research findings from sessions 1-159 remain in uht-bot/uht-research and the Substrate entity graph. They are not deleted.
4. The uht-research project is frozen — no further writes.
5. Protocol version is now 6.0.

Version manifest

Component	Version	Notes
Session protocol	6.0	Up from 5.1
Task class	SE_DECOMPOSITION	Replaces all prior task classes
AIRGen project model	Per-system	Was single project
First SE project	se-autonomous-vehicle	Scaffolded