How did I end up on a timeline where Microsoft is talking about rolling back AI in its OS and practically acknowledging vibe coding caused problems… and Linux developers are talking about ramping up its usage?
Obviously Microsoft is still worse here, but what are these trajectories?
What I think you are also seeing is AI sucking at some things and doing better than humans in others.
AI is pretty great at adding unit tests to code, for example, where humans do a just-OK job. Or in writing code for a very direct well scoped small problem.
AI is just OK at understanding product nuance and choices during larger implementations, or getting end to end coding right for any complex use cases.
The variable you’re missing is time. There was a big shift in quality by Christmas, and the latest models are much better programmers than models from one year ago. The quality is improving so fast, that most people still think of AI as a “slop generator”, when it can actually write good code and find real bugs and secutity issues now.
As someone who has to sift through other people’s LLM code every day at my job I can confirm it has definitely not gotten better in the past three months
We require you to submit markdown plan before working on a feature, which must have full context, scope, implementation details. Also verification tests mardown file of happy path and critical failure modes that would affect customer, and how tests were performed. Must be checked in with the commit. More complex, large features require UML diagrams of architecture, sequences, etc. to be checked in too.
If your plan or verification docs have wrong context, missing obvious implementation flaws, bad coupling, architecture, interfaces, boundary conditions, missing test cases, etc then PR rejected.
Every developer’s performance is judged as a systems engineer. Thoughtless features without systems docs and continued lack of improvement in your systems thinking gets you PIPed.
How do you manage?
The other missing variable is actually knowing how to use the tools. Vibe coding still produces slop. Good AI-generated code requires understanding what you’re trying to achieve and giving the AI clear context on what design paradigms to follow, what libraries to use and so on. Basically, if you know how to write good code without AI, it can help you to do so faster. If you don’t, it’ll help you to write slop faster. Garbage in, garbage out.
This is a good answer. AI tools won’t make someone who has not yet developed programming skills into a good programmer. For someone who has a good grasp of implementation patterns and the toolkit for a given tech stack, they can speed things up by putting you into the role of a senior programmer reviewing code from multiple newbies.
I’m finding that for it to work well, you have to split things up into very small pieces. You also have to really own your AI automation prompts and scripts. You can’t just copy what some YouTuber did and expect it to work well in your environment.
I used to feel the same way, but I’ve come to realize it’s slop that just looks better on the surface not slop that is actually better.
At least it compiles most the time now. But it’s never quite right… Everytime I have Claude write some section of code 6 more things spring up that need to be fixed in the new code. Never ending cycle. On the surface the code appears more readable but it’s not
Linux kernel czar?
I’m curious about this but I refuse to click the link because that just sounds so fucking stupid.
We Brits use Czar as a colloquialism for “person in charge of…”.
So the head of the water regulator might be referred to as the water Czar (and they deserve a similar fate).
The headline is stupid but the article is interesting. Greg is saying that since last month for some unknown reason, AI bug reports have gotten good and useful, and something current Linux maintainers can handle.
Yeah, but then article says that “good” ones still need reams of human work to make them acceptable.
Article is propaganda.
Greg says they’re mostly small bug fixes and that the current maintainers can handle it, not sure where you’re getting the “reams” bit from
Says in the article that they arent good to go, needing code review, code cleanup, comment and documentation cleanup, etc
Yeah I mean, the goal is not to replace code maintainers, only to assist them in their work. Greg in general seems optimistic about it:
“I did a really stupid prompt,” he recounted. “I said, ‘Give me this,’ and it spit out 60: ‘Here’s 60 problems I found, and here’s the fixes for them.’ About one-third were wrong, but they still pointed out a relatively real problem, and two-thirds of the patches were right.” Mind you, those working patches still needed human cleanup, better changelogs, and integration work, but they were far from useless. “The tools are good,” he said. “We can’t ignore this stuff. It’s coming up, and it’s getting better.”
Can I read more about it somewhere else?
It’s not just bug reports; in the last month, AI driven development has actually gone from slop to reliably better than the average human.
That’s not saying it’s writing better code, just that managing the development process and catching regular bugs is now better than when run by a junior analyst.
Makes sense that a properly balanced model with randomization turned down should be able to recognize when something is being done outside the acceptable parameters.
Makes sense that a properly balanced model with randomization turned down should be able to recognize when something is being done outside the acceptable parameters.
I don’t know how you gathered such a sense when that not being true has been the main laughing point for AI since its inception. Meta AI security and safety researcher Summer Yue’s “Nothing humbles you like telling your OpenClaw ‘confirm before acting’ and watching it speedrun deleting your inbox” was just last month btw.
It’s not just bug reports; in the last month, AI driven development has actually gone from slop to reliably better than the average human.
Funny, I heard that same claim about 6 months ago.
And I’m sure I’ll hear it again in another 6 months or so.
I’m a xennial developer. I"ve been coding for 30 years. AI now codes better (and a thousand timed faster) than most mid-level developers. The company I work for has not hired a single junior dev for months now. The new paradigm is a senior dev controlling a team of AI agents. It feels like it doesn’t even make sense to think of training juniors, because at this rate even seniors will be obsolete in a year or two.
AI in the software dev world is not hype.
Every single comment made by this person in the past three months is pro-AI. Every. Single. One.
Do you work for Anthropic? Perhaps, you are an LLM?
AI now codes better (and a thousand timed faster) than most mid-level developers.
You, if you are indeed a real person, might be overestimating your proficiency, it happens.
Huh, and here I am thinking I’m dumb because it’s such a struggle getting the ai to produce usable code.
I mean. It clearly helps in some well defined areas, but actual code? like for a feature? Of a product you expect people to pay for? And you have to maintain?
I have a few questions.
Who ultimately owns/controls this particular AI? A single company? Is this a local agent they’re running themselves or are they renting?
Who’s supposed to replace the senior running all the AI?
Besides the senior, who can discern error from function?
Are they fabricating their own chips?
And how will we continue to have senior devs to coordinate teams of AI agents if there’s no more room for junior devs? Regardless of how good a tool is, it needs to be wielded by someone who knows what they’re doing.
Terrifying
It’s an affectation of The Register they like reporting real news with a sometimes quirky voice. It’s also British so some of the language and humour doesn’t quite work as well in other parts of the world.
That’s The Register’s style. Their a little weird with their copy, but their reporting has been solid, in my experience.
Either a lot more tools got a lot better,
That’s what it was. Even the free, open source models are vastly superior to the best of the best from just a year ago.
People got into their heads that AI is shit when it was shit and decided at that moment that it was going to be stuck in that state forever. They forget that AI is just software and software usually gets better over time. Especially open source software which is what all the big AI vendors are building their tools on top of.
We’re still in the infancy of generative AI.
I tried one for the first time yesterday. It was mediocre at best. Certainly not production code. It would take just as much effort to refine it as it would to just write it in the first place.
If you read AI critics, you will see people presenting solid financial evidence of the failure of AI companies to do what they promised. Remember Sam Altman promised AGI in 2025? I certainly do, and now so do you.
Do you have any concrete evidence that this financial flop will turn around before it runs out of money?
Assume all the big AI firms die: Anthropic, OpenAI, Microsoft, Google, and Meta. Poof! They’re gone!
Here would be my reaction: “So anyway… have you tried GLM-7? It’s amazing! Also, there’s a new workflow in ComfyUI I’ve been using that works great to generate…”
Generative AI is here to stay. You don’t need a trillion dollars worth of data centers for progress to continue. That’s just billionaires living in an AGI fantasy land.
I’m sick and tired of AI fans making statements like
Generative AI is here to stay
without evidence.
Citation needed.
Um… Where would it go? I’ve got about 30 models on my machine right now and I download new ones to try out all the time.
Are you suggesting that they’d all just magically disappear one day‽
Where do you think the “new ones” are coming from?
Oh wow, comparing a thing to a completely different thing without demonstrating the comparison is valid.
Exactly the non-evidence I expected.
Whether AI can reliably detect issues and generate working code is a whole different thing from CEO’s delusions and hyperbole to game the market. Their financial success is also irrelevant, in fact it’s better if the sub/token model fails and we are left with locally ran models.
They should all be destroyed
Traditional software was developed by humans as an artifact that, and to the degree that humans improved the software for some task, got better, but it was not guaranteed. Windows 11 is proof of that, and there are a laundry list of regressions and bugs introduced into software developed by humans. I acknowledge you say usually and especially for open source — I lukewarm agree with that statement but disagree that large LLMs or other generative models will follow this trend, and merely want to point out that software usually introduces bugs as it’s developed, which are hopefully fixed by people who can reason over the code.
Which brings us to AI models, and really they should just be called transformer models; they are statistical tensor product machines. They are not software in a traditional sense. They are trained to match their training input in a statistical sense. If the input data is corrupted, the model will actually get worse over time, not better. If the data is biased, it will get worse over time, not better. With the amount of slop generated on the web, it is extraordinarily hard to denoise and decide what’s good data and what’s bad data that shouldn’t be used for training. Which means the scaling we’ve seen with increased data will not necessarily hold. And there’s not a clear indication that scaling the model size, which is largely already impractical, is having some synergistic or emergent effect as hoped and hyped.
Also, we’re really not in the infancy of AI. Maybe the infancy of widespread hype for it, but the idea of using tensor products for statistical learning algorithms goes back at least as far as Smolensky, maybe before, and that was what, 1990?
We are in the infancy of I’d say quantum style compute, so we really don’t have much to draw on beyond theoretical models.
Generative LLM models have largely plateaued in my opinion.
We’re in the infancy of AI in the sense that widespread use, testing and properly-funded development of these technologies only began a few years ago when massively parallelized processing became affordable enough, even though the concepts are older. You could say we’re in the infancy of practical AI, not theoretical.
Video killed the radio czar?
Sounds like time for a new czar
