not all hackathons

When I first started out doing hackathons (my first being in 2019), the goal was simple: go there, solve a problem, hopefully win something. The hopefully made hackathons really fun, learning with friends - in the beginning, at least.

Somehow, as time went on, the experience started to become less problem solving and more outcome-driven - padding resume, winning awards, building a portfolio. All projects had to contribute to a line in a CV, impression farm on LinkedIn or attract the attention of a potential recruiter.

The fun suddenly disappeared, and hackathons became an outlet for validating my existence, rather than a way to share in the joy of creation. Real question is - How did we get here?

I think the problem is exacerbated greatly by the whole idea that people flex the number of hackathons they've won, aiming to sell a hackathon "meta" as if there's a golden ticket to success - and people are falling for the bait, too.

Let's face it, most hackathon wins are pure luck. The idea that hackathons are competitions to be won, and not a showcase of innovation is a concerning outlook for a generation of developers that already are losing the ability to code. If you offload your thinking to LLMs in search of an AI-assisted "meta", and barely do the coding yourself, what autonomy do you have left?

As LLMs and the tools around them become increasingly advanced, it's clear to see that ideas and execution are starting to converge: if everyone uses the same commercial model, and pastes in the same problem statement, is it surprising that everyone ends up with some variation of the same underlying idea?

Anecdotally, the tighter the time pressure and the more generic the problem statement, the higher chance of this happening. Let me provide two examples of hackathons that I went to recently (at time of writing):

Note: This is not a commentary on them being good or bad, but just a discussion on how format heavily dictates outcome in an AI-assisted era.

The first example is JPM's Code For Good. The format is very short-form (less than 12 hours of coding) with a generic problem statement for social good. Out of all 7 groups, every single group came with the exact same solution.

The differentiator was simply an additional feature or two, suggested by someone with more domain knowledge than a generic GPT prompt. In the end, all the projects are equally forgettable - web apps with generic components, what I like to call the Vercel-style (shadcn & nextjs as the stack).

Does it work? Sure. Does this show any form of real innovation? Not really.

The second example is NUS Hackers' Hack&Roll. While the format remains relatively one-shot (24 hours), there is no problem statement (owing to the different context, social good in banking vs. a true "hacker" hackathon). In the end, teams had drastically different ideas, because the freedom of expression meant that you could approach problems in drastically different ways.

Yet, many solutions actually ended up solving similar problems with vastly different executions. This made reviewing projects (as a fellow participant) far more entertaining and joyful. Of course AI was used to produce code for majority of projects, but human ingenuiety makes these projects "fun".

Informally speaking, as a recruiter, if you are looking at such hackathons and attempting to recruit from there, which hackathon would be a better indicator of skill?. Of course, JPM has mentors on-site to observe how well participants communicate and problem-solve together, but in my opinion, there isn't much to observe because there isn't much to consider.

Events like Hack&Roll are important because they specifically encourage diversity, and that is a trait in thinking that forms the basis for Apple, for Cursor and so many other great products.

Hackathons are seem "meaningless" or "4fun" are exactly the kind of hackathons that unearth the gems in software dev. Maybe it's time to change our mindset from winning, and shift it towards creation for the sake of learning.

/endrant