david_shi

Every time I ask the latest models to write something, it defaults to LLMisms like contrastive negation and long unnecessary lists.

Is there a way to desloppify AI writing, or is this just the new status quo?

An agent's value is proportional to the permissions it's been granted.

There's been a lot of hype around solutions like default denial proxies, key vaults, and more, but nothing seems to address the core tension: an agent can be tricked into doing an attacker's bidding.

The best thing I could think of was to just run an observer loop and monitor everything the agent does with another LLM, but I'm curious if anyone has an elegant solution.