Conversation

Notices

1. Embed this notice
   Evan Prodromou 🇨🇦🇺🇸🇬🇷🇵🇸 (evan@cosocial.ca)'s status on Friday, 19-Jun-2026 02:10:47 JST Evan Prodromou 🇨🇦🇺🇸🇬🇷🇵🇸

   in reply to

   • Daniel Terhorst-North

   @tastapod Daniel, this is one of the ways that I use LLMs, too. For me, I'm worried that the sycophancy of LLMs might be getting me to believe my own design work is unbelievably awesome. I find that I have to ask extra questions about the designs -- what is wrong with this design for performance? For security? For accessibility? -- in order to get good critiques. I also have to proactively explore alternatives; what other ways could we do it? So: how do you keep from fooling yourself with LLMs?

   • Embed this notice
     Daniel Terhorst-North (tastapod@mas.to)'s status on Friday, 19-Jun-2026 02:10:48 JST Daniel Terhorst-North

     ngl I am having some of the _best_ and most productive design discussions with Claude-as-enhanced-rubber-duck that I have had in years. Domain model, architecture, design considerations falling out of my side project like a boss.