It would seem so: https://www.theregister.com/security/2026/07/29/word-worm-crawls-into-copilot-spreads-chaos/5280588

It makes sense: with LLMs and agents based on them, even plain text becomes effectively executable, so any document can carry malicious instructions. It’s worse with complex document formats like DOCX, ODT, or PDF, since text can be hidden from the human eye and escape detection by review. Very risky tech…

Our company develops software for medical devices. Risk assessment is everything, and we currently do “You can use AI, but you are responsible that the result works and is correct” and “You can use it, but do not input any company secrets, confidential data, medical records, or things like passwords and security-relevant data”.

I guess the difference is we’re too busy making things and we don’t have the time to brag about AI on social media? Also we’re small and don’t have the reach anyway.

I think this will depend on the industry. Slow vs fast moving, regulated vs not, whether someone depends on your API or not, etc.

But regardless, having had to deal with some legacy code written 10 years ago by someone who isn’t working at the company anymore, I would take an outdated spec over none at all. At least then I know what people intended back then, what they cared about, what they had and hadn’t considered. As long as the spec is written by a human, that information is surely valuable.

You end up with a nice spec that you wouldn’t have had otherwise? That has value even if you discard all the LLM-generated code.

I think experienced engineers are able to go straight from a vague requirement to an implementation without that intermediate step, and that power is easy to abuse. We build an implicit spec on our head, which gets translated to code on the fly, and then gets forgotten.

Not defending the LLM technology, but i do think this is one of the upside.

Have not used it for programming.

At work, this is to some extent because we haven’t decided yet whether to trust sending away our codebase to be potentially trained on.

In general, this is because I find the agressive FOMO-inducing marketting revolting, and because it would be increasing our dependence on big tech companies where I believe we should do just the opposite.

I did use it once or twice to get pointers on topics I wasn’t familiar with (e.g., new standards or protocols, …), mostly just to get a helicopter view and some sources to follow it up when i was in a rush at work. Would not do that on my personal time.

I wouldn’t object to using an open-source/open-training, ethically trained, self hosted model.

As much as I dislike AI, programming is a field where there’s always something for someone to hate. E.g., should we ban C++ articles because there’s a lot of Rust fans that hate them?

Your two options are not exclusive. Just create a community to discuss AI programming, cross post to the generic programming community if you think it’s relevant, and let people upvote/downvote as they see fit.

Why would their experience be relevant? They’re asking a question, so obviously they have things to learn. You could be nicer about it.

Then it’s a problem of the platform, if there’s no way to either tag content on a particular topic, which people can filter if they wish, or a place for meta discussions, which people can choose not to visit. I still agree with the OP that simply deleting/forbidding this content isn’t a good option.

That’s a bit like saying “I’m not interested in compiler warnings, my program works for me.” The issues this article discusses are like compiler warnings, but for the community. You should be free to ignore them, just by scrolling past. But forbidding compiler warnings would not fly in any respectable project.

I hadn’t bought a bundle in a long time, maybe I just don’t remember it being that bad, but really? Even with the “extra to charity” preset, the charity gets less than Humble themselves? That’s kind of gross.

GitHub Desktop works well for me and my workflow; even though the Linux version is only supported by the community (possible thanks to it being open source). The UI is very neat and simple. Yet you can do squash, reorder commits, ammend, commit hunks etc. Dark theme available of course! It integrates with GitHub (for PRs mostly) but afaik isn’t tied to GitHub repos.

Very relevant read: https://staffeng.com/

There’s no specific AI detection at the moment, as far as I can tell. But it has “listicle” detection. If you ask “best lawn mower”, all these “the 5 best lawn mowers of 2023” websites with affiliated Amazon links get pooled into a compact Listicle section, that you can just scroll past and ignore.

That’s crazy. Google/DDG bloat from SEO websites had already driven me out a while ago, so I hadn’t noticed. I’ve been using Kagi for a few months now, and I find I can trust my search results again. Being able to permanently downgrade or even block a given website is an awesome feature, I would recommend it just for that.