@denohawari LLM
I need to apply this to Skiddee because I think it’s a great product for generating illustrated explainer videos but just needs to get the word out there
@mattpocockuk This.
I made this to-design skill designed to fit in between to-prd and to-issues. It ensures the architecture is optimal (per improve-codebase-architecture) first before writing the new feature:
https://t.co/6dPa3MIz4y
@nicbstme@henrytdowling What if it was a self improving workflow? One that teaches itself to use better thinking methods, build custom tools, verification methods, etc.?
@theo@shadcn I tried Ghostty and it was a big step down. Didn’t even have Cmd-F search, not sure if that changed.
I like iTerm's tmux over SSH integration with the GUI. Do any of these options have it?
Why hasn’t Gmail added AI prioritisation? Most emails, I never want to see. For some, I want to be notified with an unsilenceable pager when they arrive.
Has someone made this app or am I going to have to make it myself?
@mattpocockuk I think all skills for optimising anything, such as software design, need to be framed as opposing forces.
You can't just say "make this as simple as possible", lest the agent reduce it to an empty file.
@mattpocockuk Yay for more critics of the unit test but I think any blanket statements like this need a balancing counter-force, otherwise it collapses to only system tests of the entire thing and that can't be right. What would be that equal and opposing force be?
"Read this paper. Turn it into a general skill that we could use to optimise all sorts of things in relation to my codebase. Then use that skill on itself to make it optimally simple and effective while maintaining the essence of the technique."
introducing Autoreason, a reasoning method inspired by @karpathy's AutoResearch which extends the strategy for subjective domains
the paper was co-written with Hermes Agent by @NousResearch, using a research-paper-writing skill developed while writing it
paper + results below