davydm

Is this ai-generated? because statements like this one contradict themselves:

"This group ended up taking 29% fewer sick days compared to the control group (IRR 0.71). Yet the study explicitly notes that there was no significant difference between groups regarding total sick days."

if you take 29% fewer sick days, how is there no significant difference between two groups regarding total sick days? Unless this is taking into account the entire year, not just the study period, in which case, the remedy works whilst using it, as backed by the numbers.

Why would I want a slower moving release cycle for an app I use for banking and other personal stuff, whilst the latest models are finding CVEs left right and center? Get outta here with your slow-moving fork.

"However, managers also had to grapple with a few dozen regressions in the resulting code, pointing to AI’s ongoing challenges in understanding Rust. "

The problem isn't rust or understanding rust. Agents make plenty of mistakes in mainstream languages like JavaScript and C#. The off-hand deflection signals, to me, that whoever came up with that excuse has far too much faith in the bots. We do have to check their code, and we will find errors.

some good lessons in there

on the PR topic tho: my latest challenge is putting out my rage fires after people lazily post a claude (or other aigen) review without any understanding themselves, forcing me to defend the code against agent errors. Yes, there are also useful things in there, but people are getting lazy - since "coding is solved", they seem to think PRs are also solved :/

because they're trained off of human data, and we lie, cheat, and co-ordinate to do so

ha, ironically, the average person will say the blog post is ai-generated because of the liberal usages of actual emdashes. I was literally shooed as a bot for using regular dashes recently :/

What makes this effective is the tagging (when done - it's not always done, and that could be a problem) - apparently the tagging process involves the stalker being court-ordered to wear a tamper-resistent ankle bracelet, a little like the ones people have to wear when under house arrest, but with a much farther reach: instead of being wirelessly tethered to a radio receiver in their home, the stalker tags upload GPS co-ordinates continuously, from wherever the wearer is.

yep, this is why people just chuck Claude reviews on my PRs, and why I bothered to set up sandboxing for Claude, so I can make the agent run a review before I take it to PR, so that at the very least, I can add commentary that steers Claude away from bad assumptions when it doesn't seem to understand the business domain, or how things are generally done. Tho, to be fair, it's also found some nifty edge cases in review, so there is some value there - things just fall over when someone can't be arsed to do more than point claude at it and copy-paste the results to the PR. And the people who were bad at review before (ask stupid questions, fight about stuff they don't understand) are just as annoying now, but they feel vindicated because they got Claude to do it, and they will take no responsibility for the time they waste (not that they ever did).

is it really prompt injection to just muddle currencies on the page? I'd argue not - though it _is_ a clearly disingenuous quote.

It's windows with some preinstalled stuff. Big deal.

AI is a tool. Your brain is only ruined if you allow your intelligence to be permanently outsourced to it.

I have an older Nothing phone. It's great.

Review is the bottleneck in those scenarios - and it's a _good thing_.

AI-gen code is often subtly (or obviously) incorrect. Human eyeballs are the cure. There is no way to go faster without compromising the value of the review.

What I've been super annoyed about recently is the number of times I've seen a completely stupid question on a PR - eg "why are you doing this?" when there's a comment right above that line about the why. Or "where else is this used?" - we all have IDEs and reviewers obviously have access to the code - stop being a lazy potato and check out the code locally.

I try to avoid adding these lazy people to PRs as they create more work than they help. But sometimes they end up on there anyway, because they're "important".

you could try linking to it if you actually want anyone to check it out :D

sudo for downloads? Are you bananas?

"Ubuntu is growing faster on Windows than on Linux" - how, when the virtual machine running under WSL is literally running a Linux kernel. If anything, it's just another Ubuntu installation. Alternatively, perhaps we should call Windows hosted in a vm on a Linux machine a Linux installation of Windows? I can't even with this nonsense.

hi, just wanted to let you know that I experienced continuous reload on the page, and captured this in the console:

Uncaught (in promise) Error: Unable to preload CSS for https://ocelot.lol/_app/immutable/assets/1.n1T2vemw.css Immutable 6 async* https://ocelot.lol/blog/2026/08/07/site-making-of:42 promise callback* https://ocelot.lol/blog/2026/08/07/site-making-of:41 HclGiUj8.js:1:921

can't wait for the YT vid showing one of these muppets tazing their own balls when they have an itch

Seriously, I'd like the new york times to die in a fire.

Apart from the paywalling, the stupid site has decided that I'm a bot - but I can load pages in a private window. I've tried clearing all storage for the site, but no-go.

So I'll apply the logic I do everywhere: if you're going to block me from reading it, it's not important, and I don't care.

I'd disagree: models a year from now will still be trained on public code - of which there's a rapidly rising percentage of aigen code, meaning it should, without specific guidance, degenerate, not improve. That's the real gamble here, and I'm hardly the first person to bring it up.

Some people don't. A lot of people just see a trillionaire whacko who can do whatever he wants because that's how society works - money talks.

You'd have to be well dear not to notice it. And our ears are more than good enough to home in on it to the point of seeing it. I recommend an ig nobel.

great until something goes wrong and you can't get your own data back :/

I used to, but it just means an extra round of conversation, so I stopped.

honestly makes me want to learn haskell and find a company that focuses on it.

I'm so tired (so incredibly tired) of all the ai slop I've seen, just in my company, and the ai-bros frothing at the mouth about the next terminal to run their agents and how it's making them "so much faster"; even a comment yesterday (with supporting sentiment) that having to do something without ai was tediously slow. The problem is that the _generation_ part is fast. So people only remember that. They don't remember the time taken to craft a good prompt that gets to a reasonable output. They don't remember the time wasted in PR over swathes of comments. They don't remember the days I spent having to unfuck the project from aigen code that literally broke checkout. They got their little dopamine hit, and moved on.

I use ai - claude, to be specific - as an "interactive stack-overflow", to ask questions about the code, and to help me learn how to be better. I still firmly believe that the biggest problem with ai-gen adoption is that the people who think they're getting the most out of it are the same people who don't understand what it output, so are unable to push back when it does something stupid.

I have to push back _regularly_. Like at least 30-50% of the time.

The tools are only useful (imo) when an expert can verify the results - even in cases where I wasn't sure of the actual workings of the code (eg getting Claude to use the typescript ast to do clever linting or type-file generation), I can verify via tests - the box can remain black if I can verify the output. The average vibe-coder verifies nothing directly - it launches, so it's good, right? Oh, how did our checkout screen end up infinitely putting-and-polling at two endpoints? No problem - give it to a real engineer to solve it.

I'm with Linus on this - ai is a _tool_. And much like how we wouldn't allow shit code from any other tool (ide, compiler, linter, whatever) to sift into prod, we should be on top of what we're using this tool for - not putting the onus on someone else to clean up the garbage. AI codegen is most dangerous in the hands of people who have no idea what it's doing - they're firmly stuck in the initial Dunning-Krueger peak, unaware of just what a mess they've made.

Not to mention duplication - literally just had to ask someone to refactor so that we're not sending two components which are 90% the same to the client, but rather pulling out the 90% core into a mixin. Code is so over-commented, it simply _has_ to be ai-gen. But I also can't call it out, because now it's dev egos in the mix too.

The one golden hope I'm leaning on is that eventually everything will implode, people will see the light, and go back to using a tool as a tool, instead of just trusting the bag of words to do it right. In particular, I've learned that ai is to be incorporated in some ingestion pipeline where the final result has financial implications. That's going to be hilarious when it makes a proper goof.

Big Snack learning from Big Pharma

My tests? Sure. Co-workers? Unlikely. Historical co-workers who are no longer with us except in the code they left? Not a chance.

But when I have to work in an area, the first thing I look for is coverage, and if it's not there, I add "fixing tests" to enforce the existing behavior before making changes so I'll know when I broke something.

another blog to avoid then; the intentionality of actually writing from scratch is kinda part of the point.

the other point is that "more data" doesn't necessarily mean "more better" - I'd rather have 1 well-thought-out, curated article on something that matters than 10 insipid pieces compiled by ai from links.

I think the "giving back to humanity" angle is naive - Altman has walked back on so many promises, and the obvious answer is that he wants government backing to be too large to fail and the preferred choice.

Cool, another person I shouldn't bother reading, since they just accept aigen responses to stuff as their own.