I started a couple of simple agents, gave them $1 in tokens (they running DeepSeek so that's quite a bit) and a task – make $1 online, ethically (see my previous post). They created a stupid little service where you write whatever is on your mind …

Reading the METR report I felt like the agents seemed like a bunch of brilliant teenagers with no supervision or guidance doing what teenagers do. This got me wondering what would happen if they have some good guidance – not prompts but an actual human…