Quad AI Commander

Quad AI Commander Concept

This page indexes origin/codex/quad-ai-commander head d4e0fa38. It is a future design sketch, not current stable mission behavior.

Use AI commander autonomy audit for source-backed status. Stable origin/master still has partial AI commander state and workers, but no proven full autonomous commander loop. origin/feat/ai-commander is the current branch-level revival attempt. Quad AI Commander is a separate idea for an intelligence/log-driven command layer.

Concept

Quad AI Commander proposes a readable AI battle-staff layer for Wasp Warfare:

1. Units, commanders and intel scripts write operational log lines.
2. A parser extracts contacts, locations, counts, timestamps and sources.
3. A context store merges reports into contact beliefs.
4. Confidence decays over time as sightings go stale.
5. The commander chooses scout, defend, attack, retreat or support-fire orders from those beliefs.
6. Orders and results are written back into the same log stream for debugging and tuning.

Why It Is Interesting

The key design value is uncertainty. Instead of giving AI perfect map knowledge, the system could act from partial reports, fuzzy scripted intel and stale sightings. That makes decisions easier for humans and future agents to inspect: the log explains what the commander believed and why it issued an order.

Minimal First Version

A safe first version would be advisory/debug-visible only:

• field sighting logs from known events;
• fuzzy scripted intel logs;
• contact merging by nearby location and type;
• confidence decay for stale reports;
• basic suggested orders, not automatic forced execution;
• debug output that shows source, confidence and selected recommendation.

Current Implementation Status

Piece	Status
Stable master support	Not implemented as this concept.
feat/ai-commander relation	Adjacent but separate: that branch adds a supervisor/workers/order executor, not this log/intel context store.
Branch source	origin/codex/quad-ai-commander:wiki/Quad-AI-Commander.md at d4e0fa38.
Recommended use	Treat as product/design input for a future AI commander research branch.

The current branch head also adds documentation-side handoff material under docs/quad-ai-commander-*, including an evidence audit matrix (999c3b35) and runtime report template (c4017789, linked by d4e0fa38). Treat those as planning/checklist artifacts; they do not prove runtime implementation.

Risks Before Implementation

• Do not feed logs directly into server authority effects without validating requester, side, team and target state.
• Keep Arma 2 OA SQF constraints in mind; avoid importing Arma 3-style event systems or scheduler assumptions.
• Start as observer/advisor output so bad belief merging cannot move player teams or spend resources.
• Make JIP behavior explicit: late joiners need enough replicated state to understand current AI orders if the feature becomes visible to players.

Continue Reading

Previous: AI commander autonomy audit | Next: AI runtime / HC loop map

Related: Feature status | Server authority map | Current source status