[1]: https://www.visidata.org/blog/2026/ai/#self-assessed-ai-leve...
Why would you trust the bot to self-assess truthfully?
That's like trusting a spammer to tag his emails as spam.
AI agents are grounded on the user provided direction. If it doesn't work well then it naturally loses trust.
Too late. If it doesn't work well then it has already naturally destroyed trust.
I don't really see much value beyond the 3 levels of AI disclosure that have been floating around:
- fully human - ai assisted - ai generated
I do this exercise myself for code that I really care about: after AI has written the code and I think I have achieved a full understanding, I unapply the entire patch and make sure I can reproduce the patch with the same underlying idea perhaps with less verbosity in the comments. That’s what I call fully understood. If I “mostly understand” some AI code, it means during the reproduction I need to occasionally look back at the AI code to continue.