[merlincorey]

It seems we're still collectively trying to figure out the boundaries of "delegation" versus "abstraction" which I personally don't think are the same thing, though they are certainly related and if you squint a bit you can easily argue for one or the other in many situations.

> We've gotten claude code to handle 300k LOC Rust codebases, ship a week's worth of work in a day, and maintain code quality that passes expert review.

This seems more like delegation just like if one delegated a coding task to another engineer and reviewed it.

> That in two years, you'll be opening python files in your IDE with about the same frequency that, today, you might open up a hex editor to read assembly (which, for most of us, is never).

This seems more like abstraction just like if one considers Python a sort of higher level layer above C and C a higher level layer above Assembly, except now the language is English.

Can it really be both?

[dhorthy]

I would say its much more about abstraction and the leverage abstractions give you.

You'll also note that while I talk about "spec driven development", most of the tactical stuff we've proven out is downstream of having a good spec.

But in the end a good spec is probably "the right abstraction" and most of these techniques fall out as implementation details. But to paraphrase sandy metz - better to stay in the details than to accidentally build against the wrong abstraction (https://sandimetz.com/blog/2016/1/20/the-wrong-abstraction)

I don't think delegation is right - when me and vaibhav shipped a week's worth of work in a day, we were DEEPLY engaged with the work, we didn't step away from the desk, we were constantly resteering and probably sent 50+ user messages that day, in addition to some point-edits to markdown files along the way.

[sarchertech]

It’s definitely not abstraction. You don’t watch a compiler output machine code and constantly “resteer” it.

[sothatsit]

I continue to write codebases in programming languages, not English. LLM agents just help me manipulate that code. They are tools that do work for me. That is delegation, not abstraction.

To write and review a good spec, you also need to understand your codebase. How are you going to do that without reading the code? We are not getting abstracted away from our codebases.

For it to be an abstraction, we would need our coding agents to not only write all of our code, they would also need to explain it all to us. I am very skeptical that this is how developers will work in the near future. Software development would become increasingly unreliable as we won't even understand what our codebases actually do. We would just interact with a squishy lossy English layer.

[dhorthy]

You don’t think early c programmers spent a lot of time reading the assembly that was produced?

[sarchertech]

No not really. They didn’t need to spend a lot of time looking at the output because (especially back then) they mostly knew exactly what the assembly was going to look like.

With an LLM, you don’t need to move down to the code layer so you can optimize a tight loop. You need to look at the code so you can verify that the LLM didn’t write a completely different program that what you asked it to write.

[sothatsit]

Probably at first when the compiler was bad at producing good assembly. But even then, the compiler would still always produce code that matches the rules of the language. This is not the case with LLMs. There is no indication that in the future LLMs will become deterministic such that we could literally write codebases in English and then "compile" them using an LLM into a programming language of our choice and rely on the behaviour of the final program matching our expectations.

This is why LLMs are categorically not compilers. They are not translating English code into some other type of code. They are taking English direction and then writing/editing code based upon that. They are working on a codebase alongside us, as tools. And then you still compile that code using an actual compiler.

We will start to trust these tools more and more, and probably spend less time reviewing the code they produce over time. But I do not see a future where professional developers completely disregard the actual codebase and rely entirely on LLMs for code that matters. That would require a completely different category of tools than what we have today.

[rsynnott]

I mean, the ones who were actually _writing_ a C compiler, sure, and to some who were in performance critical spaces (early C compilers were not _good_). But normal programmers, checking for correctness, no, absolutely not. Where did you get that idea?

(The golden age of looking at compiler-generated assembly would've been rather later, when processors added SIMD instructions and compilers started trying to get clever about using them.)