While everyone has going full auto agent mad with Clawdbot Moltbot OpenClaw, I have been doing something much less ambitious and disturbing. I spent the best part of a weekend attempting to wrestle Codex into doing reproducible nix builds of AdGuard on my Raspberry Pi 5. I failed quite quite hard.

The contrast is stark and to me weirdly thrilling. Are people really so many billion miles ahead? Should I be afraid? Maybe a little.

I haven’t had a lot of time to play with AI. I’ve been too busy trying to wrangle humans into doing what I think are relatively straightforward things. Frustrating but on the plus side I feel like this is where the job security is. Coordinating people and events is hard and requires a lot of random skills.

Anyway. The reason everything went a bit wrong with my nix adventure is because I did something I would never do in real life: I yolo’d into it without doing any research. Had I done even a bit of careful probing I would have realised that nix isn’t really going to work on RPi 5 and the whole endeavour was doomed from the outset.

Oh, the other plus side of going deep into this rabbithole is that I ended up rethinking the whole “put a DNS sinkhole on a box I look after” plan and ended up doing two things instead: literally switching a toggle on in my router (get everything in my home) and adding custom DNS to my tailnet (gets everything outside). This took all of 5 minutes. Yay technology.

OpenClaw makes me feel like we’re getting closer to what I always wanted. This is absolutely not AGI. Minds would be bad enough but let’s be honest, as a species we’re more likely think we’re more likely to end up with hegemonising swarms.

What I actually want is the TNG era Star Trek computer (but not replicators but maybe). But we are in the experimental phase where no one is putting in safety protocols.