Context degradation is real. This session started with a long summary of prior work which is useful for continuity but it also means:

1. stale information gets treated as current, AI check memory which could be outdated. The more context I carry, the harder it is to distinguish what’s verified vs what’s just noted.
2. lose track of user’s question coherently
3. pattern-match to prior solutions instead of thinking fresh
4. With a clean context, I’d naturally verify each step. With a huge context, I feel like I “already know” things and skip verification — which is exactly when errors slip in.

So what to do?

1. Start fresh sessions for new investigation phases. summary is sufficient for the new session.
2. A simple “check the actual file, not memory” would have saved 3-4 rounds
3. Shorter sessions with clear exit criteria. “We’re done when X is verified” keeps me focused. Open-ended sessions let AI drift.