But at least for friends and family it should be possible to create some flow where every member has a key-combo and you trust them to only sign stuff they wrote etc. and have local mini-keysign parties.

The most frustrating issues that pop up are usually library/API conflicts. I work with Gymnasium or PettingZoo and Rlib or stablebaselines3. The APIs are constantly out of sync so it helps to have a working environment were libraries and APIs are in sync beforehand.

I do use the Feynman Technique if I come across something interesting and try to explain it on paper. So if I was using it just for work, I'd probably do that. Something like "Spec driven development (Github Spec Kit and similar toolkits) is essentially a bunch of md files that provide more context for agents. There are some scripts that provide scaffolding, having agents write the md uses a lot of tokens so writing them manually after the scaffold is generated makes more sense. Try with a small project."

Love the movie, got a spray can and sprayed my whole keyboard army green after watching it then realized I can't 10 finger type. What a golden age of interesting young people in computer security. Roughly one year later (iirc), I read "Smashing the Stack for Fun and Profit" which might have been my most influential IT related read. It's probably tied with "Man-Computer Symbiosis" :)

I am a user, I like to tinker, I'm fairly confident there's more than 1% of people who care about these things. If you live in a country that is threatened by export embargos and the like it also makes a lot of sense to prioritize open.

Claude Sonnet 4.5 was able to figure out a way to resolve it eventually (around 7 fixes) and I let it create an rllib.md with all the fixes and pitfalls and am curious if feeding this file to the next experiment will lead to a one-shot. GPT-5 struggled more but haven't tried Codex on this yet so it's not exactly fair.

All done with Copilot in agent mode, just prompting, no specs or anything.

""" To implement this filter, we begin by ranking URL domains according to the volume of texts they contribute to the FineWeb (Penedo et al., 2024a) and FineWeb-2 (Penedo et al., 2025) corpus, as an approximation of web-level English and multilingual data. From this ranking, we select the top one million English domains and the top one million non-English domains. Due to domain overlap and the fact that some sites are now offline, the total number of accessible robots.txt files is smaller than two million. For each domain that remains reachable, we retrieve its robots.txt file as of January 2025 and examine the directives relevant to AI training. In particular, we focus on those targeting the AI-specific user agents listed in Appendix A. Any contents blocked by the current robots.txt is removed retroactively from the entire 2013-2024 range of the training dataset. We follow an opt-out policy, that is, if the corresponding robots.txt files are not available, we consider the data usable for training. The filtering process results in an estimated token loss of approximately 8% in English data and 4% in multilingual data. """