
Generative AI.
Sigh. I’ve written multiple drafts of this one, hesitated for a year to write it. But let’s talk about Generative AI, my (non) usage of it, and why.
TLDR if don’t want to read everything: Most people would categorize me as Anti-GenAI, even though this lack of nuance. There’s no generative AI output in any of my projects (games, websites, open-source libraries, …). I’ve tried the tool multiple times and decided I won’t use it in my daily life.
Full transparency: there has been one genai translation experiment (on Manufactur’inc steam store page) that has been removed when I remembered it while writing this.
Disclaimer: If you’re going to:
- Tell me to test that model because it’ll prove me wrong and that actually AI is great
- Tell me that I’m not pure enough to be your kind of Anti-GenAI
Please, refrain. I’ve had my fair share of debates online 😅
Sooo, now that’s done, we can talk in detail about gesture hands all of that.
As a context, I’m passionate about development and lots of tech-y things for more than a decade, this is my hobby becoming a job and my favorite activity. Also, because this is important here, I think: I love automating stuff. And a final point, software quality and maintainability has become two very important things in the last few years for me.
So, with all that intro done. What are my issues with generative AI?
The Law, the Morals, and Ecology
They are 3 big topics that are the usual issues with this technology. I think it could be solved, at the price of more expensive models and lower quality output, but it won’t without a strong regulation.
- Law: Well, models are trained on stolen stuff. Copyright hasn’t been respected. While it’s okay for research, it shouldn’t be a thing in commercial applications.
- Morals: Where to start? Slavery used to sort data, destroying very old books, military usage, forcing agendas/opinion into the models, … I could go on for hours
- Ecology: This actually probably the easier problem to solve: stop building cheap datacenters that are using polluting sources of energy, cheap cooling systems using water instead of lots of better (ecologically-speaking) solutions, …
Issues in my daily life
- Hardware cost: I’m scared of the day I’ll have to change my RAM or GPU. Even if I’ll probably be able to afford it because of my job, it’s going to be very expensive for something probably worse than what I have right now
- Forcing genai everywhere: Just, give people the choice. If your technology is so great, it’ll grow (spoiler: it won’t).
My own experimentations
As I said in the TLDR, I experimented with generative AI (and I still do once or twice a year), why? Because I want to have a technical point of view. If the result is good, I still won’t use it, but I will be able to know what’s working, what’s not.
Again, spoiler: I tried and it’s underwhelming. Two rules I’m giving myself when I try this tech: no output goes in my projects and I don’t pay anything, ever (so I tried: ChatGPT 2022-2025, Microsoft Copilot 2024-2026, GitHub Copilot 2025, Proton Lumo 2025-2026, and a bit of Claude in 2026).
- Code generation: I easily found out that, if I want to keep a good code quality and maintainability level, it’s going to be expensive, both in money or time. So it’s not worth it.
- Code reviewing: I especially tried that at work (because, well, I kinda had to), lots of false positive (either totally wrong stuff but it’s rare, or just contextually wrong). Eventually, my position for now is: I keep using Sonarqube for my projects and it’s going to be plenty enough, and again, genai is not worth the time and/or money spent.
- Translations: It’s not bad, that’s probably the best result in all my tests. It could work probably, under human review, and you have a ROI. But for the previous ideological reasons, I won’t use it.
- Marketing proof-check/review: This was the biggest and most recent tests, I wrote multiple marketing texts and tried making the model review them, compare, tell me the best of it. It was validating me, too much. When I asked humans about the same texts, the review were a lot more interesting and challenging.
- Search Engine: Ehm, well, I skip google AI overview every time. Good (?) ol’ trio is still saving me: StackOverflow, GitHub Issues and Reddit. And it goes straight to the point, not like the very verbose AI replies.
Overall, I’m not convinced that even without the ideological issues from the previous chapters, I would use it.
What research is telling us?
Research tend to agree with me, you only go faster at the expense of quality. If you try to compensate quality, you go back to your starting point (or worse). See the 2026 Harvard study about that, for example.
Also, research tells us it has a huge negative impact on people capacities. Basically: people become dumber when they use it.
But machine learning is still great
Yep, I still think (despite all of that) that machine learning is a great technology. It helped on a lots of topics in the 40+ years it existed, and probably will still help. Think of protein folding, materials engineering, OCR, and lots of stuff like that.
And at the same time, Major LLM that are sold as “generative AI”, are a pain. A pain for me, a pain for the society, a pain for everyone except Nvidia and a few people’s pockets which are filling with money right now, until the bubble pops (hopefully).
I am an artist somehow
It feels like a great title I don’t deserve, but I think it’s not that wrong, and a good reason not to use it. I love creating, I love learning, and I’m passionate about all of that. And that, can be the reason sole not to use it.
Conclusion
I won’t use it in the near future.
And honestly, if I write that devblog that’s because I start to feel threatened in my job. Not that I think that genAI can create what I create, it can’t. But I’m scared that too many people think that, and eventually, good software will be replaced by slop.
In the meantime, try supporting the kind humans around you. People creating stuff are amazing, and if you create stuff, anything, even small, you are amazing too.
Hero image: Picture by neelam279