Currently burning a lot of tokens fixing PHP on my public facing webservers 😳 Who else is having fun tonight πŸ˜‚

Replies (40)

Aedifico's avatar
Aedifico 2 weeks ago
And then faulty SW is replaced by faulty AI πŸ˜‚
GhostπŸ‘»'s avatar
GhostπŸ‘» 2 weeks ago
Yes Mike, I don't run any servers and wouldn't have the first clue where to start. I'm drinking beer and watching football, so yes, I'm having funπŸ₯³. Enjoy your SysAdmin😜.
I'm on the OpenAI plus plan. I normally don't hit my caps, but today, I'm spreading work out to avoid that 5 hour timeout πŸ˜‚ I have a local coding agent, but he's dumb and I wouldn't trust him for this work.
heh, I run my own http11 and http2 server + templating engine + htmx (pretty much an inhouse php, written in c++). Nobody read the code nobody hack, no worries :D Most of what I get are both trying to GET .env files. Noobs
One way off the treadmill: let your distro backport the fixes for you. If you're on Debian, the LTS programme issues php8.2 package updates until 30 June 2028, so the version chase can wait.
They've cornered both sides of the market. Breaking in and then offering to fix it. I had a locksmith turn up once who was a policeman during the day, breaking into criminals houses and a locksmith by night. He observed the locksmiths he called out to fix his damage earned more than he did, so he decided to get into that racket too. That's the only other time I've met somebody playing both hands in a game. At least the locksmith was human though πŸ˜‚
So this isn't going the way I expected, because I don't think you're expecting the answers I'm going to give. The "proletariat" line was a joke, except it wasn't. I don't work, which is kind of the point I was making. Right now, I've just started building (training) my own model, Brian. It will evolve and eventually it will become a robot, or perhaps something greater to take over from me both during my life and after I'm gone. I post about it quite a lot here and I have previously written articles on my, now patched, website. I'm constantly testing public LLMs to see how they are doing while going in my own direction. I'm not quite a Yann LeCun, but I'm very much favouring his world view and journeying down a different, but adjacent path
Yes you are, but that's OK, they normally achieve more than your average normie πŸ˜‚ What you just done is ask Elon what colour paint was on that flying car of his and then get annoyed because he wants to tell you about his rocket engines all night πŸ˜‚
GM I'm gradually migrating all my agents to an Omarchy box and managing them through herdr. When I wanted to fix the PHP issues, my Omarchy Codex agent couldn't SSH out, so I switched back to Codex on Mac. My agent recently migrated a server to a new VPS. It turned out it was running an unusual config and couldn't download the updates from the repository because it was too busy. So I told it to stop trying. This morning, I told it to try again and it turned out the server did the updates itself last night, automatically. Meanwhile, I'm using Chatty to manually walk through herdr config files to find a balance between agents being able to do useful work vs. YOLO mode, which allows them to destroy everything πŸ˜‚ When you automate stuff, you just abstract problems to a higher level View quoted note β†’
I'll come through for you again: List the servers in your inventory file, install the role, run the playbook. It takes three commands to not only upgrade all your servers today, but make sure they're always patched automatically in the future. The only exception is if you install software outside of apt, but in that case it should have its own auto-update mechanism. My solution there is: don't do that.
↑