Feature request: Compact long ChatGPT conversations into a clean new thread

Existing compaction in AI tools is usually optimized for token/context management: summarize older context, keep critical state, discard less relevant details. The requested feature is slightly different: validity-aware compaction. It should distinguish: * active facts and decisions * superseded decisions * disproven hypotheses * invalidated assistant claims / hallucinations * user corrections that should become the new source of truth This is especially important because ordinary summaries can accidentally preserve hallucinated claims as “things discussed.” A clean handoff should not preserve everything discussed; it should preserve only what remains valid.