Ask HN: How are you controlling AI agents that take real actions?

Re:Ask HN: How are you controlling AI agents that take real ... by chrisjj on Tuesday February 24, 2026 @04:17AM (#47134789)

> Prompt instructions like "never do X" don't hold up. LLMs ignore them when context is long or users push hard.

Serious question. Assuming you knew this, why did you choose to use LLMz for this job?

[ Parent | on HN ]

Re:Ask HN: How are you controlling AI agents that take real ... by thesvp on Tuesday February 24, 2026 @07:26AM (#47136265)

Fair. We didn't choose LLMs to enforce rules — we chose them to understand intent. The enforcement happens outside the LLM entirely. That's the separation that actually holds up in production

[ Parent | on HN ]

Re:Ask HN: How are you controlling AI agents that take real ... by chrisjj on Tuesday February 24, 2026 @08:15AM (#47136720)

[delayed]

[ Parent | on HN ]

Re:Ask HN: How are you controlling AI agents that take real ... by thesvp on Tuesday February 24, 2026 @01:09PM (#47140445)

Understanding intent and following instructions are different failure modes. LLMs are good at the first, unreliable at the second. That's exactly why enforcement lives outside the LLM.

[ Parent | on HN ]

Re:Ask HN: How are you controlling AI agents that take real ... by chrisjj on Tuesday February 24, 2026 @02:21PM (#47141418)

Software engineering has a word for that.

Kludge.

Good luck!

[ Parent | on HN ]