Parma John

The star wars stuff is fun

Dark Crystal

Ooh fun

That’s definitely a huge part of it!

“Everyone else treats me like an idiot when I say dumb shit, but AI really gets it.”

“AI is getting really good, I really feel like I can trust it now”

THIS IS THE PROBLEM. It isn’t getting good, it’s getting better at looking good to people who can’t measure quality or accuracy.

Oh hell yeah, I love it when useless decor melts my cpu

And they become hostile when an engineer tells them it must be rebuilt, which causes strained departmental relationships and those can be extremely harmful.

Okay but she’s a complete lunatic and believing anything she says is a mistake, including this

We’ve just been feeding their bad code to another LLM for a vicious code review and sending them a link to the conversation. Thus far it has worked quite well to discourage that shit with very little time wasted on our part. No one has attempted to fix their code, they just ask us to rewrite

100% with you here

You can’t create code so obviously bad that a non-engineer will know it’s bad, all code looks bad if you don’t know how to read code, and most won’t even look at it at all if the end result appears functional.

Because the *code* it generates is treated as production ready in many cases. I’ve been handed many prototypes and been told I “only need to fix the styling, the functionality is already good to go,” when that was obviously not the case at all.

The incentive structures the industry has set up do not work toward AI prototyping being a functional and beneficial solution, is what I’m getting at.

1,000,000%

But it’s impossible to enforce in most professional settings because the concept of designers handling dev work and making developers unnecessary is too enticing.

The artifact of the prototype itself becomes intentional passive miscommunication. The folks who are using it this way completely ignore the other reasons that wireframing is useful - it is immediately obvious what it is and that it is *not* ready for production.

This is not hypothetical, to be clear. This is happening. The failure to define its purpose and *limit* its scope via the aesthetic it must use (which is rough, to preserve speed) makes its purpose unclear when communicating outside of the team and within it.

Because that real speed increase causes delays down the chain. It fills in gaps incorrectly that weren’t in the prompt, and now designers believe they have working production-ready code and become belligerent to the folks trying to implement their half-thought garbage.

Yes, but the problem that the AI prototyping tools solve is allowing people who cannot code (and would likely work faster manually) to produce code they can’t fix or meaningfully understand.

It likely will, but I can’t imagine why we’d want to introduce another cost barrier to a situation that has mostly reduced it to the cost of hardware and an internet connection.

I’m referring to prototyping tools

Do not do this. Mr. Why is wrong, but trying to get him fired is insane

It could, but if you know your existing design tools well enough, in most cases building a wireframe by hand using some preexisting components will be faster than trying to explain your vision to an LLM anyway

Yeah, which I suspect will go up as soon as it becomes good enough to be a genuine Production Ready tool.

They also seem to take it *more* personally when you provide less-than-laudatory feedback, which is weird because they didn’t make the choices I’m criticizing

This hasn’t been my experience. In my experience the design function is passing off shitty prototypes as final specifications even when they’re missing huge chunks of the flow, like failure states, loading, and parallel accessibility flows like screen-reading.

This is true for both the person creating it (who often desperately wants the fidelity to be indicative of quality) and those being asked to implement them.

They’re trying it at my day job and I’ve never seen slower turnarounds, more wheel-reinvention, or designers who’re so incapable of justifying “their” choices.

It’s offering us nothing but delays and frustrated designers (who brought it on themselves).

It’s fun that we’re trying to make software engineering cost-prohibitive through, I’m sure that’s not going to bite us.

He’s just a big old dipshit, it seems

Unlikely, but thanks for the empty guess