Meanwhile my workflow of creating anything:

• Develop locally in an IDE (JetBrains) or remotely in neovim
• Init repo, push repo
• Clone PKGBUILD template
• Edit PKGBUILD, change remote, push repo

And then I have a native thingy working on every Arch machine I have (a server, a workstation, two DNS/DHCP servers, a PC and a laptop). And as the servers are chilling at my dad’s place … it’s not even costing me much (excluding 1000/500 fiber for like 100€/month, which is technically mine)

Ollama is in extra, open-webui in the AUR.

I don’t think so

Let’s acknowledge the good: AI-assisted coding can be a game-changer. It lowers barriers for new programmers and non-programmers, allowing them to produce working software by simply describing what they need.

CW made a simple software in C# as MVC with ChatGPT. Let it change it once. HE DID ESSENTIALLY NOTHING FOR 7 WEEKS, EXCEPT ASKING ME TO FIX ANOTHER BUG EVERY WEEK. WE’RE SUPPOSED TO LEARN SOMETHING. GIVING JUNIORS AI TO “PROGRAM” IS LIKE GIVING A TODDLER THE TOP TEN DEADLIEST SUBSTANCES TO LEARN WHAT IS TOXIC AND WHAT ISN’T.

That’s like saying “Enjoying to cut peoples throats and see the life vanish from their panicked eyes is not an excuse for murder”

I only do hardcore horny chats with my LLMs lol

CW be like:
Spend 7 weeks learning C# (we learned Java in Uni, it’s not that hard ffs) and implementing a proper base for a project: Nah
Let ChatGPT generate fucky code and let $me fix it over the span of 7 weeks: Hell yeah

Had I realized that it all was generated, and he didn’t have a single little clue how it works, I would’ve just rewritten it with django or something. Hell, technically the whole server part wasn’t needed, it could’ve been index.html, style.css and scripts.js and that’s it.

As I understand it, they want to train AI with AI?

Except in the 10% of times, in 30% of those you’ll have a hell of a lot of fun finding which exact line has one little variable name mismatch. But if you’re actually very careful, it’s a nice feature.

Co"worker" spent 7 weeks building a simple C# MVC app with ChatGPT

I think I don’t have to tell you how it went. Lets just say I spent more time debugging “his” code than mine.

Yes. (n)vi(m) > Emacs, and Bing >= Google. So it should be vi SearxNG

But remember, on a third device. Not the one where your KeePass DB is one fingerprint away, and your private SSH key too.

Tbh, I myself still have SSH on port 22. Firstly, because I’m lazy, and secondly … yeah that’s it. I’m honestly just lazy. But spam bots trying office/cookie123 are not a real threat, and anyone trying to actually target me will either have somehow acquired my key + password, use one of the probably many security issues that exist in the dozen services I selfhost, social engineer me into doing something (not saying I’ve given out my (old) KeePass password once, but it could be, as love makes blind (I still love her)), or just smash my kneecaps until I give out everything.

Move SSH to non-standard port, make endlessh use the default port. Only use SSH keys. Only allow correct users (so eg. your user and git/forgejo). Use fail2ban to aggressively ban (redirect to default port, so 22) and report to abuseipdb everything that fails to authenticate first try (wrong user, password instead of key), has non-compatible ciphers (generally, only allow TLS1.3 etc.), or fails in any other way. Just be sure that if you accidentally get banned yourself (eg. Ctrl+C-ing during authentication), you can use another IP (eg. force v4) for connecting.

Reminder that old games, ported correctly, run on anything. We just need the source.

Stop calling “Vibe Coding” “Coding”. It’s as much coding as shitting on a plate is cooking.

And as soon as you enter corporate stuff, LLMs are useless again, because most things are integrated into existing ecosystems which LLMs don’t know and/or libraries are only used for closed source code.

Are three lettuce heads enough for you?

Accurate

PowerfulTurtle and (Not) BurningTurtle meetup