After months of coding with LLMs, I'm going back to using my brain

in which Windsurf is forging ahead with an agentic LLM product that endeavors to not only replace software engineers but actually take over the entire software engineering PROCESS.

We're at a very interesting point, where investors and corporate interests are crystal clear in their intent to use LLMs to replace as many expensive humans as possible, while the technology available to do so is not there yet. And depending on your own perspective, it's not clear it ever will be, or perhaps it'll eventually be "good enough" for them to downsize us anyway.

I keep thinking about compilers. The old timers had been writing assembly by hand for years. The early C compilers were producing defective garbage output that was so incredibly poor that it was faster to keep coding by hand. Then the compilers got better. And better, and better, and now pretty much nobody inspects the output, let alone questions it.

LLM is a tool. I think we still need to learn how to use it, and in different areas it can do either more or less stuff for you. Personally I don't it for most of everyday coding, but if I have something tedious to write, the LLM is the first place I go for a code draft. That works for me and for the LLM I use.

comparing it to no longer doing the long division portion of a math problem isnt a great 1 to 1 here. long division would be a great metaphor if the user is TRULY only using llms for auto complete of tasks that add 0 complexity to the overall project. if you use it to implement something and dont fully grasp it, you are just creating a weird gap in your overall understanding of the code base.

maybe we are in full agreement and the brunt of your argument is just that if it doesnt fit ur current usecase then dont use it.

i dont think i agree with the conclusion of the article that it is making the non coding population dumber, but i also AGREE that we should not create these gaps in knowledge within our own codebase by just trusting ai, its certainly NOT a calculator and is wrong a lot and regardless if it IS right, that gap is a gap for the coder, and thats an issue.

Much of the time I spend writing code, not thinking about the general overview etc but about the code I am about to write itself, and if I actually care about the actual code (eg I am not gonna throw it away anyway by the end of the day) it is about how to make it as concise and understandable to others (incl future me) as possible, what cases to care about, what choices to make so that my code remain maintainable after a few days. It may be about refactoring previous code and all the decisions that go with that. LLM generated code, imo, is too bloated; them putting stuff like asserts is always a hit or miss about what they will think is important or not. Their comments tend to be completely trivial, instead of stating the intention of stuff, and though I have put some effort in getting them use a coding style similar to mine, they often fail there too. In such cases, I only use them if the code they write can be isolated enough, eg write a straightforward, auxiliary function here and there that will be called in some places but does not matter as much what happens in there. There are just too many decisions at each step that LLMs are not great at resolving ime.

I depend more on LLMs if I care less about maintenability of the code itself and more about getting it done as fast as possible, or if I am just exploring and do not actually care about the code at all. For example, it can be I am in a rush to get sth done and care about the rest later (granted they can actually do the task, else I am losing time). But when I tried this for my main work, it soon became a mess that would take more time to fix even if they seem like speeding me up initially. Granted, if my field was different and the languages I was using more popular/represented in training data, I may have found more uses for them, but I still think that after some point it becomes unsustainable to leave decisions to them.