I've never worked in web development, where it seems to me the majority of LLM coding assistants are deployed.
I work on safety critical and life sustaining software and hardware. That's the perspective I have on the world. One question that comes up is "why does it take so long to design and build these systems?" For me, the answer is: that's how long it takes humans to reach a sufficient level of understanding of what they're doing. That's when we ship: when we can provide objective evidence that the systems we've built are safe and effective. These systems we build, which are complex, have to interact with the real world, which is messy and far more complicated.
Writing more code means that's more complexity for humans (note the plurality) to understand. Hiring more people means that's more people who need to understand how the systems work. Want to pull in the schedule? That means humans have to understand in less time. Want to use Agile or this coding tool or that editor or this framework? Fine, these tools might make certain tasks a little easier, but none of that is going to remove the requirement that humans need to understand complex systems before they will work in the real world.
So then we come to LLMs. It's another episode of "finally, we can get these pesky engineers and their time wasting out of the loop". Maybe one day. But we are far from that today. What matters today is still how well do human engineers understand what they're doing. Are you using LLMs to help engineers better understand what they are building? Good. If that's the case you'll probably build more robust systems, and you _might_ even ship faster.
Are you trying to use LLMs to fool yourself into thinking this still isn't the game of humans needing to understand what's going on? "Let's offload some of the understanding of how these systems work onto the AI so we can save time and money". Then I think we're in trouble.
> Are you trying to use LLMs to fool yourself into thinking this still isn't the game of humans needing to understand what's going on?
This is a key question. If you look at all the anti-AI stuff around software engineering, the pervading sentiment is “this will never be a senior engineer”. Setting aside the possibility of future models actually bridging this gap (this would be AGI), let’s accept this as true.
You don’t need an LLM to be a senior engineer to be an effective tool, though. If an LLM can turn your design into concrete code more quickly than you could, that gives you more time to reason over the design, the potential side effects, etc. If you use the LLM well, it allows you to give more time to the things the LLM can’t do well.
Fully agree. In my own usage of AI (which I came to a bit late but have tried to fully embrace so I know what it can and can't do) I've noticed a very unusual side effect: I spend way more of my time documenting and reviewing designs than I used to, and that has been a big positive. I've always been very (maybe too) thoughtful about design and architecture, but I usually focused on high-level design and then would get to some coding as a way of evaluating/testing my designs. I could then throw away v0 using lessons learned and start a v1 on a solid track. Now however, I find myself able to get a lot further in nailing down the design to the point I don't have to build and throw away v0. The prototype is often highly salvageable with the help of the LLM doing the refactoring/iterating that used to make "starting over" a more optimal path. That in turn allows me to maintain the context and velocity of the design much better since there aren't days, or weeks, or even months between the "lessons learned" that then have to go back and revise the design.
The caveat here though, is if I didn't have the decades of experience writing/designing software by hand, I don't think I'd have the skills needed to reap the above benefit.
" They make it easier to explore ideas, to set things up, to translate intent into code across many specialized languages. But the real capability—our ability to respond to change—comes not from how fast we can produce code, but from how deeply we understand the system we are shaping. Tools keep getting smarter. The nature of learning loop stays the same."
Learning happens when your ideas break, when code fails, unexpected things happen. And in order to have that in a coding agent you need to provide a sensitive skin, which is made of tests, they provide pain feedback to the agent. Inside a good test harness the agent can't break things, it moves in a safe space with greater efficiency than before. So it was the environment providing us with understanding all alone, and we should make an environment where AI can understand what are the effects of its actions.
maybe. I think we're really just starting this, and I suspect that trying to fuse neural networks with symbolic logic is a really interesting direction to try to explore.
that's kind of not what we're talking about. a pretty large fraction of the community thinks programming is stone cold over because we can talk to an LLM
and have it spit out some code that eventually compiles.
personally I think there will be a huge shift in the way things are done. it just won't look like Claude.
I don't know why you're being downvoted, I think you're right.
I think LLMs need different coding languages, ones that emphasise correctness and formal methods. I think we'll develop specific languages for using LLMs with that work better for this task.
Of course, training an LLM to use it then becomes a chicken/egg problem, but I don't think that's insurmountable.
I don't think "understanding" should be the criteria, you can't commit your eyes in the PR. What you can commit is a test that enforces that understanding programatically. And we can do many many more tests now than before. You just need to ensure testing is deep and well designed.
I suspect that we are going to have a wave of gurus who show up soon to teach us how to code with LLMs. There’s so much doom and gloom in these sorts of threads about the death of quality code that someone is going to make money telling people how to avoid that problem.
The scenario you describe is a legitimate concern if you’re checking in AI generated code with minimal oversight. In fact I’d say it’s inevitable if you don’t maintain strict quality control. But that’s always the case, which is why code review is a thing. Likewise you can use LLMs without just checking in garbage.
The way I’ve used LLMs for coding so far is to give instructions and then iterate on the result (manually or with further instructions) until it meets my quality standards. It’s definitely slower than just checking in the first working thing the LLM churns out, but it’s sill been faster than doing it myself, I understand it exactly as well because I have to in order to give instructions (design) and iterate.
My favorite definition of “legacy code” is “code that is not tested” because no matter who writes code, it turns into a minefield quickly if it doesn’t have tests.
How do you know that it's actually faster than if you'd just written it yourself? I think the review and iteration part _is_ the work, and the fact that you started from something generated by an LLM doesn't actually speed things up. The research that I've seen also generally backs this idea up -- LLMs _feel_ very fast because code is being generated quickly, but they haven't actually done any of the work.
Because I’ve been a software engineer for over 20 years. If I look at a feature and feel like it will take me a day and an LLM churns it out in a hour including the iterating, I’m confident that using the LLM was meaningfully faster. Especially since engineers (including me) are notoriously bad at accurate estimation and things usually take at least twice as long as they estimate.
I have tested throwing several features at an LLM lately and I have no doubt that I’m significantly faster when using an LLM. My experience matches what Antirez describes. This doesn’t make me 10x faster, mostly because so much of my job is not coding. But in term of raw coding, I can believe it’s close to 10x.
> I know exactly what the result should be, the LLM is just typing it for me.
This is the mental model people should be working with. The LLM is there to tighten the loop from thought to code. You doing need to test it like an engineer. You just need to use it to make you more efficient.
It so happens that you *can^ give an LLM half-baked thoughts and it will sometimes still do a good job because the right thing is so straightforward. But in general the more vague and unclear your own thoughts, the lower quality the results, necessitating more iterations to refine.
> My favorite definition of “legacy code” is “code that is not tested” because no matter who writes code, it turns into a minefield quickly if it doesn’t have tests.
Unfortunately, "tests" don't do it, they have to be "good tests". I know, because I work on a codebase that has a lot of tests and some modules have good tests and some might as well not have tests because the tests just tell you that you changed something.
> My favorite definition of “legacy code” is “code that is not tested” because no matter who writes code, it turns into a minefield quickly if it doesn’t have tests.
On the contrary, legacy code has, by definition, been battle tested in production. I would amend the definition slightly to:
“Legacy code is code that is difficult to change.”
Lacking tests is one common reason why this could be, but not the only possible reason.
It’s from Working Effectively with Legacy Code. I don’t recall the exact definition but it’s something to that effect. Legacy = lack of automated tests.
The biggest barrier to changing code is usually insufficient automated testing. People are terrified of changing code when they can’t verify the results before breaking production.
More glibly legacy code is “any code I don’t want to deal with”. I’ve seen code written 1 year prior officially declared “legacy” because new coding standards were being put in place and no one wanted to update the old code to match.
I think it was Cory Doctorow who compared AI-generated code to asbestos.
Back in its day,
asbestos was in everything,
because of how useful it seemed.
Fast forward decades and now asbestos abatement is a hugely expensive and time-consuming requirement for any remodeling or teardown project.
Lead paint has some of the same history.
I see where you're coming from, and I agree with the implication that this is more of an issue for inexperienced devs. Having said that, I'd push back a bit on the "legacy" characterization.
For me, if I check in LLM-generated code, it means I've signed off on the final revision and feel comfortable maintaining it to a similar degree as though it were fully hand-written. I may not know every character as intimately as that of code I'd finished writing by hand a day ago, but it shouldn't be any more "legacy" to me than code I wrote by hand a year ago.
It's a bit of a meme that AI code is somehow an incomprehensible black box, but if that is ever the case, it's a failure of the user, not the tool. At the end of the day, a human needs to take responsibility for any code that ends up in a product. You can't just ship something that people will depend on not to harm them without any human ever having had the slightest idea of what it does under the hood.
Some of those words appear in my comment, but not in the way you're implying I used them.
My argument was that 1) LLM output isn't inherently "legacy" unless vibe coded, and 2) one should not vibe code software that others depend on to remain stable and secure. Your response about "abandonware" is a non sequitur.
I presume that through some process one can exorcise the legacy/vibe-codiness away. Perhaps code review of every line? (This would imply that the bottleneck to LLM output is human code review.) Or would having the LLM demonstrate correctness via generated tests be sufficient?
Just to clarify, you're inferring several things that I didn't say:
* I was agreeing with you that all vibe code is effectively legacy, but obviously not all legacy code is vibe code. Part of my point is also that not all LLM code is vibe code.
* I didn't comment on the dependability of legacy code, but I don't believe that strict vibe code should ever be depended on in principle.
As far as non-vibe coding with LLMs, I'd definitely suggest some level of human review and participation in the overall structure/organization. Even if the developer hasn't pored through it line by line, they should have signed off on the tech stack/dependencies/architecture and have some idea of what the file layout and internal modules/interfaces look like. If a major bug is ever discovered, the developer should know enough to confidently code review the fix or implement it by hand if necessary.
Take responsibility by leaving a good documentation of your code and a beefy set of tests, future agents and humans will have a point to bootstrap from, not just plain code.
For those who are less experienced with the constant surprises that legacy code bases can provide, LLMs are deeply unsettling.