[an0malous]

I don’t think that’s true, I think these authors are making a much stronger claim that AI is proficient or even an expert at software engineering. This author describes how complex and sophisticated their software is, and the only value he’ll concede to “coders” is that there might be a few bugs they’d need to fix.

Imagine not being an architect and using Claude to put together a building plan, then concluding it’s basically done but we might need a real architect to double check the measurements. It may even be true but I’d be skeptical if it’s always non-architects saying this.

[21asdffdsa12]

And - we kind of have been here before. The "proto"-type is almost complete. Its just a little slow, a little spaghettificated, just written in excel-vb, clicked together in node-graphs, or the next hot thing that makes coding unnecessary.

[bathtub365]

Why do they even need coders to fix these bugs? It would be an order of magnitude (at least) to ask Claude to find and fix them, and it will likely be successful.

Building in the physical world has physical and time constraints that cannot be overcome, which is one of the reasons architecture (and engineering) are so important in this domain. In software development these constraints were only inherent when people were writing the majority of the software. I feel like I’m seeing what I thought were fundamental constraints being eroded by the increasing speed and correctness of these tools and it’s making me reconsider the importance of some of the values that are held by software engineering.

It’s obviously dependent on the domain and solution, but if your software can be extremely rapidly rearranged, bugs found and fixed with little effort, and features added with only a minimum prompt, I think the entire definition of technical debt has changed. I’ve been sceptical of these tools and still approach their output with caution. I also worry that, as a software developer, if more can be accomplished in less time there will be less room on this planet for software developers.

[phil21]

> I think the entire definition of technical debt has changed. I’ve been sceptical of these tools and still approach their output with caution.

This very well summarizes my current thinking on the subject as well. And most of my career has been playing the role of technical debt nazi. Much to the detriment of my earning potential.

Does AI make incredibly inefficient code most of the time? Yup. But it does it at lightspeed with minimal effort.

I think many software engineers forget they exist to get real things done (in many cases at least) and they are a cost center for most businesses. If your end product is not selling software, very few people actually Doing the Thing(tm) will give a single solitary care about code quality or maintainability when they can just spend 30 minutes and $15 worth of tokens to fix it.

It won't take over everything, but I've already seen otherwise very intelligent go-getter type folks who are not technical or know how to code made extremely useful things for themselves and their small little enterprises. And this will seemingly only get better and more efficient.

For someone who really does love the idea of well architected and future-proof code this is just icky to even say or consider. But I'm coming around to this is the future for the majority of software for most places. And it may have the ability to seriously even the playing field for small enterprises in some industries.

I'm currently using it to implement a zillion side projects at home I've been "meaning to get to" for years. It makes incredibly silly unmaintainable code most of the time - but I learned to not care, and just tell the AI bot to fix it/add to it as I go along. Worst-case I spend a single night deleting it all and starting from zero to "refactor" an entire thing.

[prmoustache]

> I think many software engineers forget they exist to get real things done (in many cases at least) and they are a cost center for most businesses. If your end product is not selling software, very few people actually Doing the Thing(tm) will give a single solitary care about code quality or maintainability when they can just spend 30 minutes and $15 worth of tokens to fix it.

I am suprised to hear people so naive they expect their token usage to stay flat if code quality and maintainability starts falling exponentially?

What if to fix 2 bugs your LLM starts adding 50 new ones? Will you tell your customers in supports channel "sorry software is finished, if we try fixing anything, everything else might break, not worth it". Or "we can probably fix it, but our AI usage will raise so much we need to up the subscription 3 fold, you choose".

The speed at which LLM codes is only comparable to the speed at which they add garbage to your repo. If you stop caring about maintainability, you also stops caring about your AI/LLM related bills and the viability of your project past the PoC stage.

[varjag]

The GP explicitly mentioned "end product is not selling software". But even then, bugfixes introducing new bugs are not unheard of before. Most code used to be mediocre quality so there's not a sea of change with AI. Perhaps it even becomes better on average.

Another thing though is selling software in the first place will soon become tough proposition outside of a few niches.

[senordevnyc]

I am suprised to hear people so naive they expect their token usage to stay flat if code quality and maintainability starts falling exponentially?

There's no reason to think that quality and maintainability will start falling exponentially. On the contrary, these models get better every couple months, and 99% of software isn't actually that complicated. There's just no reason for the fear-mongering that fixing 2 bugs will cause the LLM to add 50 new ones.

[queenkjuul]

Except that i witness it create new bugs while fixing existing ones?

Not 50:1 but it does happen

[senordevnyc]

I think many software engineers forget they exist to get real things done

One billion percent. I think the vast majority of the anti-AI sentiments I hear from software engineers comes down to them caring more about playing with their tools than actually solving the problem.

[locknitpicker]

> Does AI make incredibly inefficient code most of the time? Yup. But it does it at lightspeed with minimal effort.

This hits the nail in the head.

Detractors often hang on to examples of coding assistants making mistakes or output subpar code, but they somehow miss the fact that coding assistants can also be prompted again and refactor whole swaths of code just as fast as they introduce oopsies. This means that the worst case scenario implies fast convergence to an acceptable outcome, and from there also fast iteration to improve upon that.

[rglullis]

The problem is that this approach is not sustainable. Errors compound. The cost to fix one issue might seem small at first, but over a stretch of time all these "oopsies" become architectural spaghetti that can only be fixed with a complete rewrite, which will certainly become more expensive than getting the code "organically" developed.

The only way I see AI coding working in the long run is if we go back to a Waterfall/BDUF process and having actual engineering. Let engineers really own the architecture. Enforce that any new feature - no matter how small - to be specced out with complete sequence diagrams. Ensure that every new software package needs to be put on an UML component diagram for the team to review and see each addition interacts with the whole system, etc.

If we do that, then we can just give all the documents to a coding agent and say "go ahead and implement this" with a minimal amount of confidence. But in doing this, I bet we will realize the following:

    - the "effort" has never been about writing code itself. The code is just the material manifest of all the thought that went to think over a solution into the problems that the product is attempting to solve.

   - we will likely be better off by using code generation tools (i.e, UML-to-code) and a "weak" LLM (than can run locally) than by playing the token lottery at the Anthropic Casino.

[eithed]

I mirror your thoughts. I think we'll end up with "perfect map" paradox = you cannot be vague or indecisive on what you want (and if you are then these decisions don't matter) and you're creating a 1:1 representation of what the code needs to be.

I'd substitute "owner" for the team and in that sense the owner will not need to be human.

We're at this state where Claude is great at doing the "middle" part of work, but it's crap at gathering requirements and verification of what it has done. I also don't see people caring about these aspects of software development as shown in the article

[locknitpicker]

> The problem is that this approach is not sustainable. Errors compound. The cost to fix one issue might seem small at first, but over a stretch of time all these "oopsies" become architectural spaghetti that can only be fixed with a complete rewrite, which will certainly become more expensive than getting the code "organically" developed.

That's so far been called software development.

All software developed by people suffers from this issue.

Where exactly is the novelty?

> The only way I see AI coding working in the long run is if we go back to a Waterfall/BDUF process and having actual engineering.

Nonsense. The problem is exactly the same.

With agents iterations are much faster, and this can mean things can get messier faster but can get in shape just as fast.

Ironically, agents improve the quality of the deliverable as well. Approaches such as spec-driven development do a far better job delivering features up to spec than manual coding by flesh and blood developers.

There's an awful lot of baseless scaremongering in your post. You make it sound like with AI assisted coding developers stopped paying any attention to quality.

[shakna]

> Where exactly is the novelty?

The compounding speed. Your devs might reach a point where they have to rewrite and refactor, in a decade.

Your LLM, with its higher throughput, may put you in that game breaking situation next week.

[skydhash]

> All software developed by people suffers from this issue.

And that’s pretty much where you are wrong. Take any long running open source project and you can see the craftsmanship that goes into it. It may not be perfect, but hacks are clearly marked as such.

[dkersten]

I haven’t used Fable/Mythos yet, but my experience with recent version of Opus, GPT 5.5 and recent Chinese models is that promoting again isn’t guaranteed to fix the underlying issues, nor is it guaranteed to not introduce more issues. I’ve seen SOTA models make ridiculously stupid architectural decisions that they were then unable to back out of without being prompted very specifically, instead adding a patchwork of “fixes” on top.

I’m not saying that you can’t use AI to do it because I believe that with carefully controlled workflows and context management you can, but it’s not a simple prompt away, it’s requires guidance and understanding, and isn’t the speed demon that raw prompting is.

[locknitpicker]

> I haven’t used Fable/Mythos yet, but my experience with recent version of Opus, GPT 5.5 and recent Chinese models is that promoting again isn’t guaranteed to fix the underlying issues, nor is it guaranteed to not introduce more issues.

That's not really the point though. That presumes models are only useful if they are one-shot models. That is false.

I mean, what if your prompt successfully changes 20 source files and makes a mess in one? How much work did it saved?

And the elephant in the room is when models actually outperform whatever the prompter is able to deliver, and faster. That is somehow left out.

[dkersten]

> That presumes models are only useful if they are one-shot models

That’s not at all what I’m saying.

I’m saying that in my experience across multiple models, the follow up prompts don’t fix prior underlying issues. They usually patch on top instead, unless you give them significant and time consuming guidance.

I want them to be more useful outside of one-shot uses, but I find that they currently miss the mark.

[asoderlind]

I think this is overlooking the fact that assigning a coding assistant to fix the bugs it re-introduces for all eternity just leads to spiraling token costs, which might cost more than just hiring a competent engineer in the first place.

[phil21]

Maybe. We will see.

I think computers are incredibly cheap compared to humans. These models and infrastructure to run them are going to only get more efficient in time. Right now we are still using (for the most part) entire hardware architectures mostly shoehorned from one purpose (graphics) into another. As purpose-built hardware becomes more prevalent and the SOTA starts to slow down I can't imagine a $100k hardware box not being able to handle a small team of developer's needs for many things.

I do think there will be a place for the top 20% of software engineers forever. But most people are not in that top 20%, and the quality when you get below average is not a linear progression. It will not be that difficult for AI generated code to beat the "bottom end" of the industry since tbh it's hard for me to tell the difference between LLM generated code and some of the shit I've seen over the years. I've ran across code written by folks who don't know what an array is more than once.

Most software is not built by MIT and Stanford grads making $500k/yr in the Valley. It's built by work-a-day programmers in the middle of nowhere making $80k/yr to keep some niche small business going with hyper-specific software that was first designed for Windows 95. Or stuff like making horribly designed Wordpress plugins. Or Shopify integrations. etc. etc.

I've also seen these small businesses totally held back by incompetent programmers, and despite their best efforts and huge amounts (for them!) of investment they can never seem to fix it. These types of enterprises are having AI run circles around their current engineering practices, even if it would make most FAANG engineers gasp in horror.

Either way it will certainly be interesting to watch! I just wish I was closer to retirement.

[sfn42]

This has been a debate for ever, long before LLMs. On the one hand you have people who don't care, on the other you have people who produce good code.

Doesn't matter how fast you can make the wrong thing.

[eithed]

Don't forget that you can adjust your requirements (either via plan or skill) to ensure the mistakes do not happen. The problem is that neither LLMs, nor humans (that don't work with the domain) will know they made these mistakes. Even coders don't think about everything all the time

[brazzy]

> Don't forget that you can adjust your requirements (either via plan or skill) to ensure the mistakes do not happen.

No, you can't. Adjusting prompts ensures absolutely nothing.

[eithed]

I disagree. What I should have added is that with agents (as well as humans) you do need to have tests that verify what was done.

[vrganj]

In my experience, the refactors are just as bad, just in different ways. All you end up doing is treading water with different iterations of shitty code. By the time you get somewhere acceptable, you could've just fixed it up yourself.

My preferred workflow these days is to pair program with an LLM until it gets close-ish and then manually touch it up. Without that, it just produces junk in different forms.

[dv_dt]

It's quick to build a hut in a green field, but slow to remodel the expanded building after. I think that will remain true regardless of if a team of sw developers are doing it, or an AI with a product manager or somewhere in between.

[mitxela]

Technical debt remains the same. LLMs are found not to work as well when editing messy codebases - exactly the kind you get after using an LLM for a while. After a few weeks or months you have to either throw it away and start over, or involve a human at exorbitant prices.

[squidbeak]

> I think these authors are making a much stronger claim that AI is proficient or even an expert at software engineering.

The author specifically says:

> I am sure it is not perfect (I only spent an hour working with the results), but a software engineer would iron out the remaining potential bugs that I could not find quickly (which is one reason we may need more, not less, coders in the future, to help with the explosion of new uses for software)

which acknowledges pretty clearly that engineers bring a level of insight and experience still missing from Mythos. Saying that, I totally disagree with his contention that this will always be true. It's pretty weird that the author of an article stressing the steep improvements in a model's capability can't seem to imagine further improvements in that capability. As if Mythos is where development ends or whatever gap remains between models and experts won't steadily narrow or eventually widen in reverse.