Hardware Hackathons Push Back on Vibe Coding Culture
A thread titled 'RIP software hackathons, long live the hardware hackathon' is getting traction with a specific critique: software hackathons have become 'prompt-a-thons' where participants glue together AI-generated code and demo something that barely works. The comment 'listening to someone tell you about their AI-coded project is like listening to someone tell you a dream they had last night' is the sharpest line in the thread and it's getting upvotes.
The thread on making graphics like it's 1993, building a raycasting engine from scratch, touches the same nerve. Someone pointed out that this kind of work is 'something LLMs are ridiculously bad at, so you kinda have to do it properly.' The subtext is that genuine craft, things that require real understanding rather than prompt iteration, is gaining appreciation precisely because AI has flooded the zone with mediocre generated output.
This isn't just nostalgia. It reflects a genuine question about what skill and creativity mean in a world where LLMs can produce plausible-looking software output with minimal understanding. Hardware and low-level work serve as a natural filter because the gap between 'it compiles' and 'it actually works' is brutal and immediate.
So what?
If you're running a hackathon, hiring engineers, or evaluating founders, the ability to do work that requires genuine understanding is becoming a sharper differentiator than it was two years ago. Consider structuring evaluations or events around tasks where LLMs demonstrably fail, not to be contrarian, but because those tasks actually reveal capability.