On Hackathons
(Epistemic Status: Rough and Rambly. I don't want to rewrite it since I haven't been to a hackathon recently.)
(06/09 - Someone else had a more optimistic post on the state of the hackathon. Maybe you should read that too!)
(a.k.a. why I want to go to Waterloo)
Like every well sought-after school, every year, UCSB, UCLA, and Waterloo each host their own hackathon. And these days, besides Waterloo's hackathon, I think it's fair to say all the rest suck.
When I was at SBHacks this spring, everything looked, sounded, and quacked like an AI-generated project. I used AI to write code. The other people at my table did the same. And so did pretty much everyone in that auditorium. Considering that most hackathon projects are nothing more than a repository and a website, it makes sense that AI was used extensively — the 'classic' bottleneck of a hackathon was the amount of code you could write rather than the idea you had given that you had a limited amount of time (usually 24h) to hand-write the code and build the features you wanted to demo.
Obviously, that dynamic has shifted. Code is now cheap, and in a sense, the idea matters more than your ability to code or even how well-executed the idea is; rather than being a coding/hacking contest, SB Hacks felt more like an idea contest (of course, it was a mandatory requirement to realize your idea via vibe-prompting).
And beyond the programming part of the hackathon, the other core part of the hackathon — the part where a bunch of cool people get together and talk and have fun and exchange ideas — felt like it was gone. There was no meeting other people. There was no bonding over programming (but you could "bond" over tetris/clash royale/poker I guess). In essence, at this hackathon, by and large, most people didn't seem to be having any "fun". Rather than focus on the people, the ideas, and the moment itself, the whole event to me and others seemed more like a competition where you "won" by getting some LLM to generate an implementation for a marketable idea.
In contrast to all these idea-generating competitions, Waterloo's hackathon (TerribleHack) is explicitly anti-productivity/anti-AI. It's not that AI is 'evil' or something, it's just that the whole point of the hackathon is to make something no one would use that's in mockery of the whole thing. With this premise, the facade of a competition is abandoned altogether as there is no "best" idea anymore; by any reasonable metric, how can you compare a 20-factor authentication system to a smoke detector detector (that works by starting a fire and seeing if any smoke detectors start blaring). They're equally as funny and real world applicable, which is in the end all that matters. You win if you make something, and even if you don't make anything, you still win when you see all the silly creations made by your fellow peers. In contrast to one of these zombified/brain-dead hackathons where a few groups win and everyone else doesn't, at TerribleHacks, everyone wins and has fun.