I just completely vibecoded with exe.dev and Opus 4.5 a backoffice for our FIPS 140 validation, with a separate view for the lab (where they can also upload test vectors), public links for clients, guided scripts for testing, and APIs to upload test results to S3.

I have not looked at the code once. It works great.

I am... processing this.

I care deeply about the open source ecosystem, so I feel the argument that disposable personalized software is by definition not shared or reused.

However, let's be honest here: I would not have built this without LLMs, and no one was going to open source a FIPS 140 backoffice manager.

This only displaced a spreadsheet, a local script, a Google Drive folder, a bunch of emails, and error-prone manual processes.

@filippo its hit or miss, and it seems to do really well with either

a - things that have been documented over and over and over i.e., stack exchange and forums and all that, and things that have countless tutorials available like how to set up dnsmasq on digitalocean lol.

when youre trying to have it make things that it doesnt really have to do much more than copy to complete the sentence... they seem to actually do pretty well at producing working code.

theyre also seemingly quite good at first drafts of html/css but will fall apart pretty quickly thereafter.

@filippo other than search engine though, the big use i find them actually quite great for is helping me ID chips on boards and that kind of thing, or helping find concepts in datasheets that arent easily grepped for and such

@filippo none of these uses need a datacenter though, a video card at home works just fine for all this

@filippo
1. Is it really better than a spreadsheet and a bunch of local scripts? How do you know if you haven't seen the code?

2. Are you still maintaining skills needed to make spreadsheets and write shell one-liners, in case the LLM service ceases to exist, jacks up the prices, or deliberately degrades in quality?

3. If this software is disposable, how do you dispose of it in a way that nobody in the future will accidentally use it?

(don't answer if you dont want to, just food for thought)

@filippo unfortunately the source material that allowed you to do this is unlikely to be restricted to that obtained with consent.
I still can't bring myself to go near these tools, though I accept that will likely kill my career before I've paid off my mortgage.

@tmcfarlane I know it became unpopular recently even in my circles, but I still believe copying is not stealing, information wants to be free, and using *public* data doesn't require consent beyond the act of making it public.

@wolf480pl

1. yes. by using it.

2. yes, I trust I will survive

3. the code is not public???

@filippo
by 3 I mean someone else in your organization

@filippo I'd argue that you didn't vibecode because that pretends the agency was wholly yours. While you say you wouldn't have built this without LLMs, other people would've. And you could have asked them to do the same thing – but never would you have said that "you" had contract-coded this piece of software. It would've somebody else who did this.

By by using an LLM, it wasn't you who wrote the code. It was generated by an LLM based on previously written code. It's removing the factor of human work and collaboration. It enables an individual to not pay or enslave other people to do work for oneself, but to get rid of such an enslavement altogether, but without freeing up the resources for those. It's the ultimate dream of the capitalists. Subjugating other human workers was never the goal, it was the means. LLMs provide the goal without the messy means, but it's not less terrible for those involved on the other wise.

What I'm saying is: This may feel amazing, empowering – independent of other human's willingness to do that work for you. But it's not you who did the work, neither is it someone else. It's a machine that can't be held responsible, but also won't ever not do what it's told. I find this scary more than anything.

Regular tools also displace work. But they need operators.

@ljrk

> other people would've

Nope, wrong. I address this in the post!

@filippo it works great, until it doesn't...

@rafaelmartins unlike spreadsheets, manual processes, and scripts, which always work forever!

@wolf480pl I am the only one that can access the exe.dev account. Thank you for your concern.

@filippo personally I was always against copyright infringement (if you ignore me under the age of about 18), I went all in on CopyLeft and open source because, for me, it's about owning that decision to give stuff away.
I don't think art counts as "information", it's output of human creativity, and has likely been a commercial since day 1.
Is code art? Maybe not, but I don't think I can make that decision for other people.

@tmcfarlane That's a coherent position and I appreciate it even if I don't agree. Also, FWIW, I have much weaker opinions on art and GenAI art, also because I am a programmer and not an artist, so I am more entitled to an opinion on how code, rather than art, is used and produced.

@filippo *would've been able to. It's irrelevant to my point whether they'd open-source it or give it to you for free.

@ljrk did you read the thread?

https://abyssdomain.expert/@filippo/115826635670699524

@filippo I've wasted more time with opus 4.5 and similar than with manual processes since I started using it... it is great at first glance, but as soon as it gets lost, you waste more time trying to make it work than doing the actual work

@filippo time AND MONEY, FWIW

@rafaelmartins you are replying to a post where I share my experience building a functioning system that's a lot better than what I had before! I'm sorry it didn't work for you, but I know how it worked for me.

@filippo
oh, so that's how it works...

Also maybe me thinking of it like it's a faulty power cable is a bit excessive...

Anyway, it looks like there will be less need for average programmers writing boring CRUD apps. I hoped it'd result from programming becoming more accessible to people, similar to writing.

If instead it happens through a centralized service, that's kinda sad.

But maybe it means those of us who keep their jobs can focus on the difficult/interesting problems?

@wolf480pl @filippo This is the use-case what I see among my non-programmer friends: simple little apps that they can't, others won't write (for a reasonable price). Also, usually 0 risk (think: "will I get a notification about a show").

One thing to keep in mind is that all these services are currently provided at a loss for the providers. It will be interesting to watch what happens when prices start to approach actual costs and expected margins.

@wolf480pl @filippo
> I hoped it'd result from programming becoming more accessible to people, similar to writing.

Depends on your definition of accessibility. If you do not live on a western world developer salary, vibe coding an app for $20 that may or may not even work is really not an option.
One of the good things about programming and FOSS in particular was that as long as you could get your hand on a computer, you could participate and learn. LLMs are turning it into pay to win.

@ck
@filippo
Yeah, by accessible I meant some form of programming language that is easy for more people to learn and be productive in - sth between spreadsheets and AutoHotKey, maybe like HyperCard.

Outsourcing programming to a centralized service does not count as "programming being accessible".

Also, good point about pay-to-win. And I think it's only gonna get more expensive.

@filippo @tmcfarlane What about free software licences? I mean, the code is public, but it is not public domain. Copying is authorized, but attribution is still required by most of them. Maybe it is not "stealing" legally speaking, but it seems to me like this is a licence violation anyway. Am I missing something?