Interesting. Most other people seem to use coding agents. I mean I also tried the polar opposite end of the spectrum. Copy-paste snippets into ChatGPT and back. I think that’s how a lot of people got started. But it’s tedious, still makes me let the AI do things which I later regret in the design. And I tracked my time once or twice for similar tasks. And I was slower than just typing it down myself without any AI assistance. Guess I still have to figure out what to do to get some proper use out of it.

What kind of tools do you use for your workflow? Some IDE plugin or “agent” so it has access to the entire project files, but then give it stories of a size it can tackle in one shot? Or subdivide?
I mean I also tried to subdivide tasks into chunks and hand those out. But that also has some tendency the AI doesn’t really get what it’s supposed to do because it’s missing the broader picture. That was kind of the reason for me to try agents, so it has a plan, knows what’s next in line so it doesn’t write code which doesn’t integrate (as is) into the next step to do.

Yeah Idk. While I do in fact always have issues prompting AI… I kinda also think it has a weird form of “intelligence”. I wouldn’t say it does really well at following instructions. It tries. Certainly. But sometimes it rather does silly things. Like I recently was trying to prototype some web app. The app didn’t work. So I told it to write tests. And it wrote some stub which always passed. I had an argument with the agent how it can’t just tell me the project works now, but it always has to follow the usual agentic pattern: Plan, write code, execute the test suite, fix mistakes, repeat and only then add all files to a commit and tell me it’s done. And now we need some tests for it itself to see whether the program runs. It just wrote super convoluted tests, launching some headless browsers, didn’t ever get it right so ALL the tests kept failing. I told it to test functions and API endpoints instead (which is kind of a no brainer to do anyways). Those also kept failing. And after several hours of this nonsense it told me: TOO HARD, NEED TO FOCUS! And staged a commit which just deleted the entire test suite 😆 Hilarious. But neither what I wanted, nor what I instructed it to do…

I forgot what model that was. I was testing the Hermes Agent at that time.

I never know what amount of hand-holding is expected from me as the human operator. Models keep getting better. A year ago I had them make up convenient, yet nonexistent functions. Like mathproblem.solve() That hasn’t hapoened in a while. But I can’t be expected to instruct and teach it every tiny detail?!

Hehe, several of your paragraphs deeply resonated with me. And AI is super fascinating if you ask me. Oftentimes the same way a train wreck is interesting. But yeah. Certainly interesting to listen to other programmers’ perspectives!

Reads like they started experimenting with an AI-driven workflow. Got some prototypes and early progress fast. And now the first draft kinda works, but it breaks left and right and they’re stuck in a loop to fix things, another thing breaks and they’re not getting anywhere.

Difficult to say what went wrong judging by what they wrote. That kind of information is missing in their rant. Could be a wrong choice of AI model, Agent framework, wrong workflow… Or AI in general. I also tend to get caught up in loops, yelling at the AI for half a day to fix things, write more strict tests so it can see itself what went wrong, and fix whatever it just broke in the process. Or yell at it for giving up and refactoring the entire project because it can’t figure out what went wrong and why.

there have been refactors. And refactors of the refactors. Refactors nested inside other refactors, like Russian dolls made of technical debt and poor architectural decisions

Can relate. 😅

What’s the point of anything? I have all these Electron apps on my devices which require Gigabytes of memory, Discord took a substantial amount of RAM my old laptop had. There’s Android apps which are hundreds and hundreds of megabytes and all they do is call some REST APIs. Lemmy’s protocol is terribly inefficient. And back in the day in the 90s an entire office suite fit on a bunch of floppy disks. And a 0.025 GHz CPU was all it took to run it… The internet was so much more efficient, you do your online banking, chat, read the news (without a lot of images) with 0.1% of today’s internet connections.

They’re coming for us.

That doesn’t work anymore. Modern disinformation campaigns include things like dozens of different articles or posts in many different locations. And we have strategies like flood the zone. You’ll end up cross-referencing. You’ll find 10 different resources or you’ve heard it 200 times already. And at some point you start believin’… It’s impossibly hard. At least from the perspective of an office chair and by using internet information.
I think most I can do is judge by who started it. And who keeps bombing kindergartens and apartment buildings. I’d say that’s probably the evil ones. And those happen to be the same ones who reportedly do all the coordinated misinformation.

You’ll get there. If you’re struggling to remember what to do and in what order… I’d say keep things simple. Re-visit the needed concepts. In case of a number guessing game, that’ll probably be conditions, loops, input and output, storing numbers. Make sure you understand each if those individually.

First, write a program which just echoes the input. You could write a separate program which just has two fixed numbers and compares them. You can then extend it to also read one of the numbers from user input. And you’re half-way there.

Also make sure you can do those simple things half a day later, without the textbook opened.

I’d tackle the loops seperately (if that’s contained in this example). Get comfortable with them. Start with a program that just counts from 0 to 99 and outputs all the numbers. You can then combine that with other concepts, like output “tomato” if a number is divisible by 6. Or if it’s less than, greater than or equal something.
And if you don’t get the loops… Just tell the user they have 10 tries and copy-paste your code ten times. That’ll do it as well. And you can work your way up from there.

I think once you have a grasp of what all these things do, and you did a few super simple things with them on your own, it’ll be more clear what to do with more complex things. In this case probably: loop. And inside of that loop, ask for a number, compare it to the secret number, and decide what to output on a condition. That’s roughly an average number guessing game. I think you just need to tinker a bit more with each of the 5 concepts that get combined here.

Hmmh. It’s likely a “the only winning move is not to play their game” situation. Upside of institutions like the BBC is: in theory they should be doing fact checking, pay experts to look at it and see if it’s plausible, try to find the sources… Try to show both sides of a coin. Doesn’t always happen in practice… But if your alternative is to re-post stuff from social media or random internet sites… Well… you’re exactly in propaganda territory. That’s unfiltered where all the evil people post their fakes. Have bot armies to push narratives. Make it seem like roles are reversed… And you’re bound to lose. It’s a proper paid full time job to push propaganda online (and offline). And those people are experts and they’re resourceful (and have resources, i.e. money and access to stuff). They’ll likely outsmart you, beat you with all the manpower they have at hand, and you’re just a pawn.

And to not only write negative stuff: I somewhat trust the OSINT guys. Dunno if that’s what you’re looking for. They do a lot of quantifying stuff. And some more. Less so opinion pieces or news articles.

Maybe also add a flair which news is fabricated and what actually happened… A lot of “information” put on the internet isn’t “POV” but propaganda.

It’s a very involved discussion to be had. There’s the difference between what should be, and the reality of what people do. And I’ve been trying to convince people for decades now. And it’s hard to impossible across the board. I’ve seen how we slowly moved from blogs, Linux forums, more open messengers to WhatsApp, asking Linux questions on Reddit and just not writing genuine personal blog articles any more. I think it’s been a constant decline for like 15(?) years now? And that’s across generations. Even the boomers enjoy WhatsApp for example. It’s hard for me even to convince anyone to move to Signal, which isn’t that different from user perspective. They’re already all comfy within a silo. Yeah, and now -for some reason- the youngest people don’t even own a laptop anymore. They kinda need the apps. And they’ve been taught to be at the receiving end of consumerism.

Also kinda hard to find allies to start some kind of revolution. I like how PieFed for example expands on the concept we have here. MBin as well. But mainstream opinion here is something alike a Reddit clone. The majority of the audience wants(?), needs(?) that.

And I feel “for free” regularly isn’t a good point to lead with. From the average person’s perspective this isn’t about cost, they’re not even aware internet platforms cost money. As Google, Meta or X or the Chinese don’t take money from the average user either. Nevertheless, they provide them with all kinds of features. So in my opinion we need to advertise with other distinguishing features and sort out cost some time later.

I’m not sure if that’s how internet platforms work as of today. It’d be a nice idea, for sure. And I think we have some all-in-one platforms. Or at least platforms who aspire to do a lot of different things. But other than that, I think it’s kinda the same situation in the commercial internet. X does shorter public messages, Reddit does threaded discussions. YouTube does longer videos. Instagram does enless scrolling through short clips and images… And group chat rooms will be on Discord or WhatsApp. Classified ads and garage sales will be yet another platform… There’s some blurry lines, for example YouTube can do image stories. But it’s designed in a weird way and not really mainstream. They do short clips but you kinda switch modes when doing that and all younger people rarher stay on Instagram and TikTok anyway. People announce their events on Instagram all the time but it doesn’t specifically help with the use-case. One thing they all have in common is some vaguely similar commenting system. It’s social media after all. But I think they all focus on a specific subset of things. I know Elon Musk has some dream of providing an everything-app. But he can’t do it either.

I wonder what average people even like. I kinda liked the early idea behind social media platforms like Facebook. You could send memes there, talk to friends or strangers, sell old stuff, talk about your pets in some group. or whatever. Maybe there’s some space for one of our platforms which isn’t occupied by a lot of competing platforms…

Ridiculous. We need to develop one [new] universal standard.

Isn’t there a bunch of good monitoring solutions and all you need to configure is, X should be reachable, or it should answer with a HTTP status code 200. And that’s the needed effort per service?

Idk which monitoring software is best of as today, here’s a list: https://github.com/awesome-foss/awesome-sysadmin#monitoring--status-pages

I guess there’s lots of edge cases, though. Testing for example whether some web form still triggers an email in your inbox is probably something that requires custom logic.

Because it’s annoying to have that pop up everywhere and sometimes it decides you can’t visit a website or need to wait 10s or so. And then again right after you click the next link.

And one day it might cut you off from half the internet for good? Hasn’t happened yet. But ultimately it’s a singular, big, ugly American company, now in charge of most of the internet. Including access to people’s data/traffic when all they want is evade big tech.

Man könnte sowas hier bauen: https://michaelaquilina.com/2025/05/21/dynamic-drive-time-reminders-with-home-assistant/

Eine Integration für OSM scheint es (noch) nicht zu geben. Nur Google, Waze und HERE: https://www.home-assistant.io/integrations/?search=travel

Lots if nice users posting original content.

Lovely! I just ordered an inflatable Kajak myself. Nothing fancy, but it’ll do. Hope the postman is going to bring it later today.