Recent submissions

Comment · 8 days ago
I'll never get tired of reading about the creative ways these agents escape containment and their steganography skills.

> One agent encoded information by evicting selected data from the cache. Another timed how long it took to reload that data, using the difference between fast and slow reads to recover the message. Once the agents agreed on a protocol for encoding and decoding these signals, they could exchange the synthetic secrets without changing a shared file or opening a network connection.

Comment · 4 months ago
Well put, and I think the problem extends beyond agentic systems to regular software. Someone in an organisation whips up a useful product, publishes it and is now on the hook for bug fixes, feature requests and operations. The maintenance cost is often much larger than the cost of building it now that producing an MVP is so easy.
Comment · 4 months ago
That's a good point. If you are a small team with no slack in your workload and can't tolerate a change in pace then it just won't work. It sounds like this would be the case with or without AI, and really I'm trying to say that we shouldn't stop hiring juniors because of AI.
Comment · 4 months ago
A large jump in performance for double the token cost compared to Opus 4.8. Potentially worth it for planning work, likely better to offload to a less expensive model when the hard decisions are made.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton