Open Source Barware: free, local-first bar inventory software (GPLv3)

it’s easy to agree with your point on the surface, but then we’d be handwaving a lot of negative subtle signaling. I think because HN is mostly developers on the inside we have to assume that a non-Technical audience might see it differently. That is of course true but one very realistic problem I see is that a vibe coded app with made up testimonials, figures, facts pretty much made-up everything is indistinguishable from an actual company. And we put this entirely on the customer to discern. That’s a terrible precedent, no different from posting falsehood on Facebook.

edit: OK I wrote this before I even opened the page and now that I seen i the entire persona is made up. Or it’s seeded in reality and then AI fills in the 80% to round it out, except it’s on the audience to trust. Sad.

I've been chewing on this for a while, but I don't think the thought is fully formed. I'll rough it out here anyways, it certainly isn't a novel or unique one.

I think my main issue with vibe coded apps is that the signal to noise ratio got flipped upside down.

Things that may have been positive signals in the past aren't really great indicators anymore. Maybe they shouldn't have been positive in the past, but they certainly aren't now. Broadly, things like polished website design, punchy language, clean branding, slick landing page, etc were all indicators. At a minimum, they at least meant that time and effort was put into it. It didn't really mean anything about the product, but it would display at least some level of time and effort.

Now, the outward appearance of a project no longer conveys anything about effort or durability, I can easily spin up a home page that would have taken me weeks pre-AI, then never touch it again.

The site states that it took roughly a hundred hours to develop. I have no real way of verifying that and I don't even think that it matters. The most evident signal that I feel I can still get a pulse on is when something is clearly, heavily AI written. All other signals that would supersede it require additional time to evaluate. Things like reading the docs, skimming the codebase, testing the product etc.

With a project like this, the ask made to the end user is to spend the time to, at a minimum, read the docs to see if it's a good fit. If they think it's a good fit, the next ask is to spend the time to integrate it with their current inventory. Then learn the workflow. Then learn the idiosyncrasies. All of these things involve, real human, time.

When an end user is asked to invest this time, there are only so many signals available to them to help decide if it's a good use of their time, energy and resources. For better or for worse, it seems like that list of signals is shrinking and right now, something that appears vibecoded is still one of the clearest visible signals.