SCdF

Locking the door of the cockpit, actual on the ground policing in terms of monitoring terror cells.

I am getting workable code with Claude on a 10kloc Typescript project. I ask it to make plans then execute them step by step. I have yet to try something larger, or something more obscure.

This is the fracture in the industry I don't think we are talking about enough.

It overwhelms everyone's ability to keep track of what it's doing. Some people are just no longer keeping track.

I have no idea if people are just doing this to toy projects, or real actual production things. I am getting the sneaking suspicion it's both at this point.

Look I have no idea if this is related, but I have noticed recently, talking to other developers, that the addiction / allure of the speed that coding with AI agents gives you is leading to a relaxation of their standard quality bar. This doesn't even feel like the evil overlords whipping them more, it is self-inflicted.

When you can get multiple different agents to all work on things and you are bouncing between them, careful review of their code becomes the bottleneck. So you start lowering your bar to "good enough", where "good enough" is not really good enough. It's a new good enough, which is like you squinting at the code and as long as the shape is vaguely ok, and the code works (where that means you click around a bit and it seems fine), it's ok.

Over time you lose your "theory"[1] of the software, and I would imagine that makes you effectively lower your bar even further, because you are less attached to what good should look like.

This is all anecdotal on my end, but it does feel like quality as a whole in the industry has tanked in the last maybe 12 months? It feels like there are more outages than normal. I couldn't find a good temporal outage graph, but if you trust this: https://www.catchpoint.com/internet-outages-timeline , the number of outages in 2025 is orders of magnitude up on 2024.

Maybe this is because there are way more, maybe this is because they are now tracking way more, I'm not sure. But it definitely _feels_ like we are in for a bumpy ride over the next few years.

[1] in the Programming as Theory Building sense: https://gareth.nz/ai-programming-as-theory-building.html

Remember, I am responding to this:

> With AI, if well done, it would be *useful nearly everywhere.*

I'm not saying it doesn't have uses.

Having said that, there are two things I never want AI to do: a) degrade or remove the need for me to express myself as a human being, b) do work I'd have to redo to prove it did it correctly.

On translation, sycophancy is a problem. I can't find it now, but there was an article I read about an LLM mistranslating papers to exclude data it thought the user wasn't interested in. So no, I wouldn't trust it for anything I cared about.

I do use AI: I'm literally reviewing some Claude generated code at the moment. But I can read that and know that it's done it right (or not, as the case often is). This is different from translation or summarisation, where I'd have to do the whole task again to prove correctness.

> With AI, if well done, it would be useful nearly everywhere.

I fundamentally disagree with this.

I never, now or in the future, want to use AI to generate or alter communication or expression primarily between me and other humans.

I do not want emails or articles summarised, I do not emails or documents written for me, I do not want my photos altered yassified. Not now, not ever.

I think in that case you have even less ability to turn that stuff off? If it's not there for you yet, perhaps it's a slow rollout still?

The three examples you listed are implementation details, so it's not clear if this is a serious post. Which datacenter they deploy code in is (other than territory for laws etc, which is something you may wish to know about and pick from) an implementation detail.

A better example would be: imagine every single operating system and app you use adds spellcheck. They only let you spell check in American[1]. You will get spell check prompts from your Operating System, your browser, and the webapp you're in. You can turn none of them off.

[1] in this example, you speak the Queen's English, so spell color colour etc

I am in the UK. TBC this isn't a gmail.com email address, this is a paid "small business" workspace against a custom domain.

Eventually they backtracked and allowed (I think?) all paid customers to disable gemini, but I had already migrated to Fastmail so :shrug:

I disagree: in as much as I have noticed this *far* more with AI than any other advancement / fad (depending on your opinion) than anything else before.

This also tracks with every app and website injecting AI into every one of your interactions, with no way to disable it.

I think the article's point about non-consent is a very apt one, and expresses why I dislike this trend so much. I left Google Workspace, as a paying customer for years, because they injected gemini into gmail etc and I couldn't turn it off (only those on the most expensive enterprise plans could at the time I left).

To be clear I am someone that uses AI basically every day, but the non-consent is still frustrating and dehumanising. Users–even paying users–are "considered" in design these days as much as a cow is "considered" in the design of a dairy farm.

I am moving all of the software that I pay for to competitors who either do not integrate AI, or allow me to disable it if I wish.

Do we know if this counts for Prime?

Though tbh the writing has been on the wall for awhile. It's really frustrating, because it otherwise just gets out of your way, which is why I like it.

I guess I have to dedicate an afternoon to finding an alternative.

Sorry I think I worded it badly. The OP is not the you, I mean the propaganda is. As in, a large part of the propaganda is the idea that people should have opinions about this kind of thing, as opposed to accepting the idea that there are people who are more well versed in these topics than they are. eg the scientific consensus is that global warming is real and it's a big problem. There isn't really an opinion to have there, broadly.

How does the climate crisis fold into that, or do you not think that's a problem, or do you think it's a problem with no ramifications?

> and you are asking them to decide on COVID or climate change

In case you didn't mean this, do you agree that the propaganda you're referencing above is the "you" in this sentence? eg the propaganda is the thing that is asking them to decide on covid or climate change.

I don't think anyone who is genuine expects the public to have expertise in these topics. The propaganda seems centred around a constant war against intellectualism and expertise, such that people think they should have an opinion on things they are woefully unqualified to have opinions on, and politicians just align themselves to what they think will get votes.

Ironically, the pub it suggested near me that was the most fucked closed down years ago (it's not just them, quite a few databases don't know that), so yeah, good call.

I don't think with any confidence we can say it will be for the better. Or at least, not on balance for the better.

To be clear you believe that you do a year's worth of work in one week? Every week? So halfway through this year you will have done 25 years of work?

Yes, it is not fair. The entire gambling industry is unfair and exploitative.

Late to this, but my interpretation of the parent's point was eg: LLMs still often produce bad code, despite "reading" every book about programming ever written. Simplistically, they aren't taking the knowledge from those books, and applying them to the knowledge of the code they've scraped, they are just using the scraped output. You can then separately ask them about knowledge from those books, but then if you go back and get them to code again, they still won't follow the advice they just gave you.

More power to you obviously. But I have mixed feelings about this.

There is so much information that curation is inevitable. Sure. But I don't want that curation to be "fun". I don't _want_ tiktok in my life, or really anything whose goal is "engagement". I don't want time killers.

One of the reasons for getting back into RSS for me was to have a direct feed to authors I'm interested in.

But I understand that quickly can become unmanageable.

When that time comes, I think I'd be interested in the curation being about compressing content down, not expanding it out. That is to say: use the algorithm to select from a large pool of what you're interested in, down to a manageable static size (like a weekly newsletter), as opposed to using it to infinitely expand outward to keep engaging you.

Really disappointed that down detectors down detector[1] isn't detecting that down detector[2] is down

[1] https://downdetectorsdowndetector.com/

[2] https://downdetector.com/

LLMs can't lie: they aren't alive.

The text they produce contains lies, constantly, at almost every interaction.

I've only read the abstract, but there is also plenty of evidence to suggest that people trust the output of LLMs more than other forms of media (or that they should). Partially because it feels like it comes from a place of authority, and partially because of how self confident AI always sounds.

The LLM bot army stuff is concerning, sure. The real concern for me is incredibly rich people with no empathy for you or I, having interstitial control of that kind of messaging. See, all of the grok ai tweaks over the past however long.

I really like it, but for the cost of a cup of coffee you could try it for a month as well. I highly suggest you just do that: commit to spending a month using it, and if you don't like it at the end cancel. Maybe it won't work for the way you use search, maybe it will, the only way to find out is to try.

Don't be intentionally daft. Skiing and sports aren't notably addictive, and don't notably cause harm in society.

You are taking a forklift to the gym. The goal isn't actually to move the weights from one place to another. The goal is the process.

For those who were not familiar with the licence they have switched to: https://www.tldrlegal.com/license/functional-source-license-...

Of those apps you've listed that I've used, none of them have been notable for being high quality to me, though as you say it's difficult to measure. For me I would rate them somewhere between unremarkable (notes, calendar, contacts!?) and awkward (pages, numbers, keynote). If you asked me to guess what desktop software Apple makes that people rate highly, I never would have guessed any of those, except _maybe_ Logic[1] and Final Cut, though ironically those are two of the three I've never used.

I also think you're confusing what I wrote. It's not a competition.

I have just found that Apple's hardware on desktop has been stronger than their software, in my experience (periodic sporadic use, ~2006->now).

[1] and now from a sibling comment I hear that perhaps people regard that tool as bad, so there you go, they jury is clearly out

I don't think it's the modern Apple, I think that's just Apple.

I remember using iTunes when fixing the name of an album was a modal blocking function that had to write to each and every MP3, one by one, in the slowest write I have ever experienced in updating file metadata. Give me a magnetised needle and a steady hand and I could have done it faster.

A long time ago they had some pretty cool design guides, and the visual design has often been nice, but other than that I don't think their software has been notable for its quality.

I'm interested, both as a Fastmail customer, and a software developer: what does this let you do that a PWA doesn't? Perhaps not become the default email client? Is there anything else?