AI generated code is like IKEA furniture.
Good enough for most people.
Good enough to vastly reduce the need for fine craftsmen and women.
Good enough to deskill those who call themselves cabinet makers / senior software engineers. These days the cabinet makers I personally know just do contract kitchens for project builders.
But bad enough that at the high end with special requirements / taste / money / an inflated sense of self worth, fine furniture makers still exist and thrive.
Perhaps 1% percent of current software engineers of today will be needed in the future....
"They’re foundationally incapable of always and consistently preventing prompt injection attacks. “Alignment work”, safety harnesses, and sandboxes all help to add barriers against the worst, but there are fundamental gap" ...
They seem to be very good at a lot of rudimentary best practices, more so than humans, but more accurately - if you run and audit pass with specific instructions ... they're very good at that.
I mean - it's what they're the best at which is applying 'fuzzy heuristics' in a mechanical way. If can describe issues concisely, the patterns, the styles, the rules then LLMs can very mechanistically and methodologically grind through them.
I don't even see how this is controversial - without getting into 'what their reasoning means' - we can all agree that their synthetic reasoning is pretty good at narrow scales, and they've been 'trained by compilers' and are extremely good at spotting common patterns.
If you back that up with a lot of tokens ... they excel.
Designing architecture, that's difficult, but hammering away at all the 'known-knows across a system' especially to identify things ... they're pretty good at that.
LLM is the new Excel
The article says what many here like to hear, but in my opinion the core arguments are false.
> Making software debuggable, maintainable, layered, and composable – that’s still quite a trick
Not really. I have been working on a mobile app for months, and I stopped even glancing at the code about two months ago.
150k LOC, around half of that in tests, and the AI still has no problem maintaining the code on my behalf.
Debuggable? It can add extensive instrumentation in seconds.
None of this requires expertise, prompting, or mention of TDD. It's the default.
Frankly I do not believe the author tried developing a large codebase fully agentic and without reviewing the code. I believe many here look at the code produced, deem it substandard, and go hands on.
> They’re foundationally incapable of always and consistently preventing prompt injection attacks
From Anthropic's article about the Auto mode:
> We commissioned an evaluation from a third party, Trajectory Labs, who tested different models within the latest publicly available versions of Claude Code and Codex as of July 17th 2026.1 They tested 72 indirect prompt injection scenarios held out from Anthropic
> In this evaluation, none of the 720 attack attempts succeeded against Claude Fable 5, Opus 5, or Sonnet 5 running auto mode. On the other hand, 5.83% of the attacks succeeded against GPT-5.6 Sol running Codex's Auto-review mode. Notably, this is greater than the 0.09% average attack success rate against our latest models running in bypassPermissions mode without additional safeguards. The tests showed a 19.03% attack success rate against GPT-5.6 Sol when running in Full Access mode
I'm sure someone is going to reply with how they do not trust Antrophic's research, but lacking other data, prompt injection appears to be largely solved already.
How do you expect us to take your views on LLM code quality and durability seriously when a) you don’t even look at the code and b) you’ve only been doing this for two months?
With generated code, the directory structure, interface design and general state management is usually a haphazard mess. Even with the best frontier models. But what really gets me is the model often tries to make assumptions for me that I didn't specify in the prompt. Subtle things like which error states are "oh shit we need to bail" vs "this isn't a deal breaker." Sometimes it will ask, but more often than not it will just make a decision and it's often the wrong one. If I don't have a fully kitted out test suit and a good type checker to verify the final product against, the the whole looping thing is just useless to me and I'm back to reviewing every line of code it puts out and having to draw on my years of architecture experience to make sure we don't build a giant pile of trash.
Because they are designed to be used by managers who don't know how to answer these questions and don't want to be asked them. Just have the magic answers box pick something.
The generated code is fine at the functional level, the directory structure is usually the standard pattern for the given type of project.
The error types and codes, it will produce to spec.
If you type 'make me that thingy' - yes, it's probably not going to do what you want, but if you give it spec and guidance, it usually will.
The 'interface design' ... not very good though.
[dead]
They're RLHF'ed to an inch of their lives to be able to one-shot complete tasks, since requiring human input defeats the purpose of being able to replace the labor force.
But once the insanity ends LLMs will be packaged as tools for developers to use to boost their productivity, and we'll consider them as we do IDE's and debuggers and stuff. But we have to get through this hype cycle first.
wake up, slopinthebag. wake up..
> In the past year, agent harnesses crossed the “can it be done” rubicon.
Brother, I'm still in "Can you get it right?"-mode. What am I doing wrong? (Rhetorical, but advice welcomed).
What is “it” specifically and what languages are you using?
I'd be happy to screenshare with you if you like, we can work on something trivial or open source. Half an hour should be more than enough to see whether you're doing anything obviously self-sabotaging.
There are two 'camps' with respect to AI.
One camp already knows that Neural Nets don't work and are a dead end.
The other camp hasn't yet figured out that Neural Nets don't work, but are convinced that they do (or eventually will), because they think everything always improves over time in a linear fashion.
You're appealing to ambiguity. All you've said is you have failed—how is anyone supposed to know what went wrong?
I’ve found some success is small projects, with limited scope, in a greenfield.
I’m terrified to attempt agentic anything in the repo my job actually cares about. I triggered it once by accident, when the agent was first rolled out and enabled by default… it broke everything. Now I just use ask mode, and even that is wrong half the time, and once it goes wrong it just keeps getting worse.
I saw a post from Dave Plumber who vibe coded up a new cross platform task manager. He said his spec document for the AI was 107 pages long. So maybe what I’m doing wrong is not giving the AI a literal novel of spec.
> He said his spec document for the AI was 107 pages long.
This sounds like programming but with extra steps that make it take longer with less reliability.
Ikr, at that point the code itself is a better way of encoding the information.
Tell it to use red/green TDD and start things off with an already configured test suite, maybe with a single test that asserts 1+1==2.
Make sure it know how to run the tests before it starts writing any additional code.
Then set it a clear goal.
Basically all the examples of LLM's building impressive things have been because they have human written tests to base the implementation on. If you have an LLM write the tests the results are far less impressive or valuable.
Yes and LLMs are known to cheat on tests written by them.
It's not always cheating either. They aren't intelligent, so they don't actually understand the purpose of the tests or can build them to define the actual semantics of the problem space. It's literally just next-token prediction based on the codebase and prompt. Cheating implies that they have agency, and ironically agents don't.
> It helps to know that LLMs don’t “reason”. They predict ..
Semantics. Prediction is the training objective. The ability to reason can be, and very arguably is, an emergent property of that.
Maybe it looks like reasoning, and maybe that's enough for some.
Even if that was true, you'd have to still prove it has emerged.
What would be your test to determine that?
Why would "reasoning" be an emergent property of prediction?
Predict multiple outcomes, induct across them, refine.
Why wouldn’t it?
There's a lot of reasoning in the training data.
How do you predict without reasoning?
Where is the reasoning in linear regression?
Where is the reasoning in synaptic transmission?
There isn't, which is exactly the point: we do not yet understand the fundamentals behind reasoning.
Don’t we? We can build something which has all the output associated with reasoning.
I’d say we’ve figured out the fundamentals behind reasoning.