Why split up your code? Why have classes? "Design RIGGER". (AI doesn't stop the need for that!)

x.com/i/article/1848…

Couldn't access X.com on my local Dollar General wifi. (Was going to ask Groq about pain meds, will double check in time but, might not be going there anymore.)

I so can't wait until more people realize that design is a graph search problem, and an LLM is only a heuristic.

Imports would be so much easier for LLMs if they were written AFTER the code.

(Sitting in the hospital.... Again...)

I have LITERALLY told the fucking world how to, for free. Its not even hard now.

So? When are people finally going to realize that a prompt and some markdown docs handed to an LLM that amount essentially to "please oh please do what I say" are not sufficient for code generation systems to function at large scale?

New chapter in my RISC-V SoC blog series: this one introduces using C (instead of assembly) with the GCC cross-compiler to program my Femto CPU core to flash LEDs in classic “Blinky” style—using just five RISC-V instructions. blackmesalabs.wordpress.com/2025/09/14/bml…

Watch Starship's eleventh flight test → spacex.com/launches/stars… x.com/i/broadcasts/1…

new dev workflow unlocked 🤯 Use a realtime model to collect files/context and create issue. Assign to agent in github action iOS app makes it accessible from every where.

I think combining this concept and voice control could be useful for managing multiple agents at once

zenitsu_apprentice Good question, it's basically entirely hand-written (with tab autocomplete). I tried to use claude/codex agents a few times but they just didn't work well enough at all and net unhelpful, possibly the repo is too far off the data distribution.