[armchairhacker]

The author notes that vibecoding has entirely replaced coding in hackathons (where speed is essential, bugs are tolerated, and only the demo is judged). I agree.

But then says this means software is “solved” so only hardware hackathons matter. Why?

If anything, I think software hackathons have become more useful, because ideas have become more useful. Even if ideas are cheap, not everyone has 24-72 hours for a prototype, in a creativity-inducing space that may inspire better details.

And software isn’t solved: some ideas still require low-level knowledge and skill to translate into prototypes, especially if the hackathon judges require some functionality.

Whether your purpose of a hackathon is:

- Make a prototype, then if it seems useful afterwards rewrite it into a full product

- Make a prototype that seems useful to attract investors (whether you start a company that may not launch or apply to a company that wants your creativity)

- As an organizer, find ideas related to your company

- Have fun, enjoy free food and good company

[dvh]

>But then says this means software is “solved” so only hardware hackathons matter. Why?

Because ai cannot do hardware, it cannot solder wire, it cannot replace red led with blue one and find current limiting resistor for optimal brightness. It cannot see what part of enclosure needs to be cut. It cannot see the startup transient on the ldo.

[ssl-3]

> Because ai cannot do hardware

It can.

> it cannot solder wire

Plenty of placement and soldering machines exist that are far faster than humans. They just aren't yet integrated with the bot.

For the stuff that is unusual or particularly difficult, just add a human. Same as with code.

> it cannot replace red led with blue one and find current limiting resistor for optimal brightness

Sure it can. Just add a camera.

> It cannot see what part of enclosure needs to be cut.

Challenge accepted.

> It cannot see the startup transient on the ldo.

It can see that using the same tools we use to observe such things. https://github.com/aimoda/rigol-dho824-mcp

[necovek]

And one needs to think from the outcome perspective too: maybe it is too hard to cut the part of the enclosure, but can it 3D print a new one hollowed out exactly where needed?

[ssl-3]

Or, I mean: This is a [hypothetical] human hackathon that is tainted by bots, not a bot hackathon that is devoid of humans [although I'm sure we'll have those some day as well].

Instructions that deliberately lack specificity, like "Drill a hole for the switch and then mount it" work fine in meatspace whether those instructions are issued by a bot or a human teammate.

[aiisjustanif]

Not at the entry price point of a hackathon, which is the point. Using AI in software hackathons are magnitudes cheaper than AI doing manual labor in hardware hackathons.