Session Discipline

What it measures

In your last complex session, did output quality stay consistent from start to finish?

The four diagnostic questions

1. 01You start a new session at inflection points: scope changes, new modules, post-failure rebuilds, or after ~20–30 substantial prompts.
2. 02You carry updated briefing and current-state context into new sessions — not the full prior conversation history.
3. 03You recognise degradation symptoms (contradictions, ignored constraints, repetition) and stop prompting before sending more fix requests.
4. 04You extract or document build state before switching modules or ending a session.

What good looks like

Strong session discipline. Ensure inflection points are identified upfront, not retrospectively after drift occurs.

Common failure mode

Long sessions with degrading output signal missing re-grounding. The cost of a new session is always lower than the cost of drift compounding.

Critique & references

Editor's notes — to be expanded

Long-context studies report ~40% accuracy degradation past certain history thresholds. The cost of starting a fresh session with a tight briefing is reliably lower than the cost of compounding drift inside a bloated one.

• → Lost in the Middle (arXiv 2307.03172)
• → Anthropic — Claude Code best practices