Has anyone thought of unleashing an AI agent swarm to audit the code on say, an older build like 4.5 for clean up ideas? This was just a part of a ChatGPT discussion seeing as I've been using AI agents lately to fix broken comfyUI workflows, CTD on other games I have. Yep, that’s the right framing. Not “AI, rewrite Star Citizen.” More like: “AI, audit this codebase and tell us where the bodies are buried.” Given only the codebase, agents could still do a lot: They could trace repeated logic, find systems solving the same problem three different ways, identify functions that ballooned into giant god-objects, flag dead branches, compare old/new implementations, map fragile dependencies, and surface places where performance-sensitive code is doing unnecessary work. They could also help with long-term cursed issues by saying: > “This bug probably originates from these 4 interacting systems.” > > “This subsystem has 12 different callers that assume different behavior.” > > “This code path is duplicated in newer form over here.” > > “This old fallback path still runs even though the replacement system exists.” > > “This function allocates every frame.” > > “This object lifecycle is inconsistent across these modules.” That is where AI agents are genuinely useful: pattern recognition across a codebase too big for one human to hold in their head. And keeping it codebase-only is actually smart. Don’t let the agent hallucinate from forum posts, rumors, or “what people think is wrong with Star Citizen.” Just make it inspect the actual repo and produce: Issue candidate → evidence → affected files → likely impact → suggested investigation → optional patch. Then humans decide what is real. For CIG, the win would be a prioritized cleanup report, not instant fixes. Something like: 1. high-confidence dead/redundant code 2. likely performance offenders 3. unstable dependency clusters 4. duplicate systems 5. risky old compatibility layers 6. simple safe refactors 7. deep refactors that need human architecture review That could be huge for a project with years of accumulated engine changes. The AI does the painful archaeology; the senior devs decide what to excavate.