A software designer and engineer. Makes modest websites. Member of the Anti JavaScript JavaScript Club.
Lives in Taipei, Taiwan, and sometimes in Berlin. Previous home bases include New York, London, and Birmingham, UK.
As recent as Feb 6, 2026, self-employed. Sleeps, codes, crafts, learns German, bakes pretzels, and contributes accessibility and civic causes through advocacy and tech.
Stories
RSS What are these?Photos
All photos RSSFilm photos
All film photos RSSNotes
Recent notes RSSA friend told me to use LLM for something… and said it’s often correct in a particular area. Idk. I just asked one to rank some classic designs of flip clocks and it got things wrong and misattributed designers multiple times, then double down when I pointed it out.
I think LLMs are most correct in the things we don’t know the answer to. Sure, whatever you outputted must be true.
I also met up with a friend who works at one of the big LLM companies and he told me their product is getting so much better at coding at an incredible speed. So I asked him, is it better at x, y, and z? The things that made it basically unusable to me? The answer is no, and they are known issues people are still working on them. This really makes me wonder what the amazing developments were… if it continues to be incapable of doing the most basic things I can use help with?