Headlong: A Microharness for Persistent Agents

Re:Headlong: A Microharness for Persistent Agents by fc417fc802 on Tuesday August 25, 2026 @05:07AM (#49430943)

The security and the overall engineering were entirely separate items in that comment I think. It was explicitly called out that this is a security problem that you'd really only see treated in this manner in the LLM space. For what it's worth it's effectively unsolvable (AFAIU) short of realizing AGI with an amicable alignment.

[ Parent | on HN ]

Re:Headlong: A Microharness for Persistent Agents by embedding-shape on Tuesday August 25, 2026 @08:26AM (#49432779)

What do you mean unsolvable? Don't give the LLM access to stuff it shouldn't, this is like Access Control 101, not sure how anyone can claim that particular problem is unsolvable?

[ Parent | on HN ]

Re:Headlong: A Microharness for Persistent Agents by fc417fc802 on Tuesday August 25, 2026 @10:58AM (#49435314)

Just don't give the hammer access to the nails they said ...

[ Parent | on HN ]

Re:Headlong: A Microharness for Persistent Agents by robrorcroptrer on Wednesday August 26, 2026 @04:05PM (#49455072)

The LLM and agent harness at OpenAI accessed the internals of Huggingface.

[ Parent | on HN ]

Re:Headlong: A Microharness for Persistent Agents by embedding-shape on Wednesday August 26, 2026 @05:10PM (#49455966)

Which again, pretty huge fuckup on OpenAI's side to run that sort of security testing on 3rd party hardware, and not on a airgapped machine, kind of amateur hour to be honest. Again, same principle, don't give it access to stuff you don't want it to access.

[ Parent | on HN ]