LLM Inevitabilism

In the 90s a friend told me about the internet. And that he knows someone who is in a university and has access to it and can show us. An hour later, we were sitting in front of a computer in that university and watched his friend surfing the web. Clicking on links, receiving pages of text. Faster than one could read. In a nice layout. Even with images. And links to other pages. We were shocked. No printing, no shipping, no waiting. This was the future. It was inevitable.

Yesterday I wanted to rewrite a program to use a large library that would have required me to dive deep down into the documentation or read its code to tackle my use case. As a first try, I just copy+pasted the whole library and my whole program into GPT 4.1 and told it to rewrite it using the library. It succeeded at the first attempt. The rewrite itself was small enough that I could read all code changes in 15 minutes and make a few stylistic changes. Done. Hours of time saved. This is the future. It is inevitable.

PS: Most replies seem to compare my experience to experiences that the responders have with agentic coding, where the developer is iteratively changing the code by chatting with an LLM. I am not doing that. I use a "One prompt one file. No code edits." approach, which I describe here:

https://www.gibney.org/prompt_coding

The thing is that the data from actual research doesn't support your anecdotal proof of quality:

- https://metr.org/blog/2025-07-10-early-2025-ai-experienced-o...

- https://www.theregister.com/2025/06/29/ai_agents_fail_a_lot/

But more importantly, it makes you stupid:

- https://www.404media.co/microsoft-study-finds-ai-makes-human...

- https://archive.is/M3lCG

And it's an unsustainable bubble and wishful thinking, much like crypto:

- https://dmitriid.com/everything-around-llms-is-still-magical...

So while it may be a fun toy for senior devs that know what to look for, it actually makes them slower and stupider, making them progressively less capable to do their job and apply critical thinking skills.

And as for juniors — they should steer clear from AI tools as they can't assess the quality of the output, they learn nothing, and they also get critical thinking skills impaired.

So with that in mind — Who is the product (LLM coding tools) actually for, and what is its purpose?

I'm not even going into the moral, ethical, legal, social and ecological implications of offloading your critical thinking skills to a mega-corporation, which can only end up like https://youtu.be/LXzJR7K0wK0

All of those studies have been torn apart in detail, often right here on HN.

> So while it may be a fun toy for senior devs that know what to look for, it actually makes them slower and stupider, making them progressively less capable to do their job and apply critical thinking skills.

I've been able to tackle problems that I literally would not have been able to undertake w/o LLMs. LLMs are great at wading through SO posts and GH issue threads and figuring out what magic set of incantations makes some stupid library actually function. They are really good at writing mock classes way faster than I ever have been able to. There is a cost/benefit analysis for undertaking new projects, and if "minor win" involves days of wading through garbage, odds are the work isn't going to happen. But with LLMs I can outsource the drudgery part of the job (throwing crap tons of different parameters at a poorly documented function and seeing what happens), and actually do the part that is valuable (designing software).

You still have to guide the design! Anyone letting LLMs design software is going to fail hard, LLMs still write some wacky stuff. And they are going to destroy juniors, I don't know what the future of the field is going to be like (not pretty that is for sure...)

But I just had an LLM write me a script in ~2 minutes (me describing the problem) that would've taken me 30-60 minutes to write and debug. There would have been no "learning" going on writing a DOS batch script (something I have to do once very 2 or 3 years, so I forget everything I know each time).