Posts by breakfast 1964 publicly visible posts • joined 24 May 2007
When I was setting up the DNS for a site last year I decided to go full IPV6 because why not? It turned out the why not is because a bunch of browsers seemingly can't work with IPV6 addressing and just wouldn't find the site. I shifted to V4 and everything worked as normal. That was a genuine surprise to me, but there's not much point in a website that a proportion of users simply cannot access, so I'm stuck with V4 for the time being.
Of all the things I loved about being self-employed, accounting wasn't one of them. Turning a horrible once-per-year job into four horrible quarterly jobs is not an improvement. It will ultimately cause a certain proportion of businesses to close down and another proportion who were previously law-abiding to go off-books. Will it increase revenues? Unlikely.
Good news for accountants, though, I guess. I wonder what discipline most people at HMRC trained in.
Over the last couple of days since this story broke it seems increasingly that the "bugs" found are tiny edge-cases of the kind that one might also spot with a regular fuzzing tool and, importantly, the kind that often don't get fixed because they don't create any meaningful attack surface.
Another massively hyped nothingburger of a story from the AI guys.
I've been a steady user since it was in beta and I recently had a perfectly answerable question closed for not containing a complete reproduction code, a bar to entry that would effectively prevent any non-trivial problems in more complex systems from being viable topics for questions. If I could reproduce it in a ten line sample I would have fixed the problem already, obviously.
To me that's the biggest sign that the site is really dying - I was part of another StackExchange site that got a single super-zealous moderator who destroyed it as a viable community in a couple of months by closing every question that didn't match his incredibly specific and limited set of ideas about what it should be about. I still wonder whether he grasped what he had done or whether he felt he had accomplished something useful in that role.
Weird that at this conference sponsored by companies who only make money if you buy more tokens from them they have a lot of presentations suggesting the solution to whatever your problem is has to be even more agents.
Just keep spending the money and losing touch with how to do your actual job and somehow their businesses will become profitable in spite of unbelievable costs and ridiculous inefficiencies.
Same way they do everything else - they copy somebody else's code.
It's other people's code all the way down, the gamble vibe coders make (and it is a gamble - the psychology of vibe coding is very much the psychology of gambling addiction) is that it will copy the code from somebody competent.
It's worth checking out this mastodon thread from @Jonny on Nuromatch.social in which he tries to unravel the spaghetti. It is hilarious and horrifying by turns.
Yn union gywir.
Even if it was able to track Cymraeg, would it be able to follow Gog, vs Hwntw? Could it distinguish a Cardi or someone from the south-west? If it was trained on standard written Welsh it would be miles away from how anybody talks, and even if it was very well trained it would be as baffled as me trying to follow anyone who talks like our local farmers do.
I have a pretty simple theory as to why people are impressed with generated code and may believe it more secure - aside from the gambling addiction that actually drives its use, of course - which is that it's the average of the internet and necessarily 50% of programmers are below average.
If you're a worse-than-average coder, AI code is going to look really great to you. If you're better than average it's going to look mediocre. If you're a bad enough programmer then maybe AI could make your code more secure - you certainly don't know! Now you can let a computer create problems for your colleagues and users instead of doing it by hand. For people who are content to be bad at their jobs, who don't have any interest in improving their own abilities, that's a real draw.
This makes companies deliberately recruiting engineers who love AI particularly ironic - they're actively seeking worse developers.
The trouble with projects designed to support vibe coding is that they're largely vibe-coded themselves because the people working on them are all-in on AI. Given that generated code doesn't care about security (or know that it exists) (or know anything) this kind of high-level compromise is the tip of the iceberg in terms of the kind of liabilities you are incorporating into your projects.
Kids are at least coming in with a degree of scepticism and awareness that this is a dystopian technology designed to cause them harm.
It's absolutely old people who seem to be most profoundly impacted by this technology - like most adults, I use facebook as little as possible, only keeping up with a few wider family members who don't use other channels much and weird village drama in the local community group, but the generation who told us that we shouldn't trust anybody on the internet certainly seem to have gone all-in on trusting what people on the internet say. It has made them horribly vulnerable to the worst scammers, grifters, and liars, whose ravings they share with the same wide-eyed wonder they use for obvious AI images (that they think are real for some reason) and agonisingly trite doggerel about things from The Old Days.
It's painful to witness and I hate to think what it is doing to the aging brains subjected to it relentlessly.
It's not common, but a few companies who are less drooling and craven than their contemporaries have at least added a hidden checkbox down on page 26 of their "Settings" menu that brave AI haters can use to flag that we don't want our work used to train AI.
I propose this be called a slop tout opt-out.
This is pretty straight up-and-down Microsoft get to decide what goes in a package and if the high end models are too expensive Students will have to live without them. That's just business. It sucks to lose something you had access to, but sometimes it be like that. The only way to stop people complaining when you withdraw a service is not to give it to them in the first place.
Whether students benefit from having access to machines that allow them to bypass learning is a more tricky question. It really depends on whether the desired output of university is an endless stream of student projects or educated people who know how to think about the world. It used to be the latter, but given how willing universities seem to be to integrate the slop machines, I suppose we must assume that what matters to them is now the former.
Several times I thought my old netbook had died because I had somehow accidentally knocked the precise combination of keys that disabled the screen. It was a secondary use for a function key IIRC, labelled with an empty rectangle which only ever made sense in retrospect. I still can't think of many practical uses for that.
Geothermal has potential as well - there are a lot of interesting innovations coming in around geothermal systems that can increase pressure in times of high supply and effectively act as grid-scale batteries for energy from renewables.
They provide power and they offer jobs for people whose expertise in geology and deep drilling would otherwise limit them to working in the petrochemical industry.
I feel like if I'm running a business and I have to make a quick decision, I've already lost a certain amount of control.
It's more like an cargo ship than a motorbike - your whole job is to be enough steps ahead to direct things so you have time to get into position. If there's a situation where you need to make a snap decision you're probably in trouble already.
Delegating the choice to somebody else's computer probably is as good as deciding based on a flipped coin, dice roll or tarot reading, but it's no substitute for actual human judgement, which is literally the only reason these people are being paid a thousand times more than anybody else in the company.
I have noticed this too, but what really blows my mind is that I'm using Reaper as a DAW and the download for that is fifteen megabytes. There's nothing missing from it- it's a full-featured audio workstation and it fits into fifteen meg.
Makes everybody else look like they're standing still.
The one thing I will give to Labour voters is that they had some reason to expect a Labour government after voting for the Labour party.
That wasn't really on the manifesto - Labour basically promised nothing ahead of the election, which makes it a little surprising how disappointing they have been in power. You promise nothing and still manage to be worse than expected? Extraordinary.
However, historically Labour have stood for a set of centre-left values - expressed by many of their candidates and MPs - and it is reasonable for people to vote on those values. Most of the reason for their amazing plummet in the polls is that after election they inscrutably chose to become continuity-Tories/Reform-lite, something nobody wanted and nobody voted for.
Voters in the USA have no such excuse. The Republicans promised authoritarianism, white nationalism, and a concerted effort to end American democracy. If people voted for that, we can assume it's what they wanted.
"adopt [AI] too slowly, and businesses could lose out on increased productivity and growth opportunities,"
Could miss out on opportunities, is laying on the hypotheticals really deep. In the old days they used to tell us that it's important for businesses to make decisions based on evidence, but I suspect it was always hype and faddish buzzword nonsense.
Adopt AI too slowly and businesses could find their offices filling up with confused unicorns.
Adopt AI too slowly and businesses could find their offices have turned into custard.
Adopt AI too slowly and businesses could find that their employees have formed a choir and insist in singing songs in a language nobody understands with the intention of "awakening the sky angels."
There is as much hard evidence for all of these as there is for productivity gains due to AI.
AI can generate documentation, but the best technical writers I have worked with spent most of their time asking awkward questions of the business analysts, architects and design team to establish how exactly the system will work. There's no way AI will have that kind of initiative, it can potentially flesh out a skeleton but it can't ask the product team what they expect to happen when a product is deleted from the system.
I have a reasonably tidy application I wrote using a JS framework a while ago and although it works well and was pleasant to put together, having had to focus on other work for a year or so if I want to upgrade to the current version of the framework and the UI component library I chose (both of which are relatively simple by the standards of these things) it will take weeks of work. In fact, by the time I get to the end of that work they'll probably have updated again and that update will probably also not be backward-compatible.
Perhaps I'm just old and grumpy but the constant move-fast-and-break-things updates that ignore backwards-compatibility strike me as a tad frustrating, to say the least.
It's not only wages - sure I may have experience that makes me several orders of magnitude more effective in my work than a new graduate, but I also have decades of learning that I have rights, and that there are things my employer cannot and should not demand of me, and things I don't have to do if they don't pay me extra for them. That is one weird trick employers hate.
I haven't read in detail but the other very important thing is whether they're recording time spent in detail with and without AI or just surveying users on what they reckon. Enough other surveys have shown people believe AI is saving them time, but when you evaluate how their time is spent it turns out to have no benefit or work out slower, that we clearly can't trust vibes-based metrics.
Unlike everywhere else, of course, where vibes-based metrics are entirely scientific (as long as they get results I agree with.)
Big corporations always hated Open Source and through the introduction of AI they have found a way to perform a DDOS on the community that cares for and about it.
I doubt that is a deliberate plan, but I bet they're quite happy to see it happen. The harder it becomes to maintain FOSS the fewer viable competitors there are for corporate products and the less pressure there is to improve those products.
