I’m not surprised it’s moving this direction, but I am surprised it’s moving this direction now.

Hardware is extremely difficult to find right now and token prices are falling off a cliff in terms of price per intelligence. About 6 months ago I was talking about how it’s only going to get cheaper from here and still seems to hold up today, we really don’t need anything better than Opus 5.5 and compute is only going to keep increasing (price decreasing) who knows what the field will look like next year.

Either way I’m happy to live in a world pushing towards that, just surprised.

I was also impacted, you need to remove the payload that got dumped:

Payload: /tmp/.pxb/{curl,wget}, /tmp/.spid

Two processes named khungtaskd / cron

They also stole your admin token so go to the Plex UI and log out of all devices for whichever account has ownership of the instance.

Then you can reclaim ownership and it resets the admin token.

You should likely be fine, I didn’t have any crypto miners on mine, it was just in a docker instance and didn’t have access outside of it. I would be much more worried if it was a direct install on a main PC.

Also install watch tower

These research papers talking about AI leading to cognitive decline seem so asinine when certain people’s productivity has just skyrocketed under AI.

Sure the hand coding capabilities decline, but that’s replaced by thinking around the actually product or architecture.

A huge sect of new Linux users and Selfhosters are talking about how much AI has enabled them yet we’re still bringing up papers on cognitive decline that were written on models from 3-4 years ago when Claude Code didn’t even exist.

Love this, opus 5.5 is absolutely nuts at optimization. I was able to use it to get some 10x performance improvements recently.

It seems like the tide is finally starting to turn here in terms of LLM usage. Even 4 months ago this thread would be a nightmare lol

It’s a good thing Sam Altman isn’t Anthropic !

Probably the right decision given there’s no way to truly tell what level of AI generation or effort within the project itself with AI has taken place.

I understand the willingness to ban low effort low quality one shot projects, but given most software nowadays incorporating some level of LLMs, I can’t ever see a blanket ban working out anywhere.

It’s interesting I have the exact opposite perspective with AI in regards to selfhosting.

The customizability is now endless for me, I can do anything, customize anything, problems that are typically 1-2 hours are fixed in <5 minutes.

Don’t want to fiddle w/ a GUI? Shoot off a message and it’s done.

I had tons of stability issues within the last year unfortunately, I would still recommend against it.

OP if you don’t mind running this through a Hermes / OpenClaw harness this is probably the easiest way with pre built email connectors. Also if you don’t mind spending a few bucks a month DeepSeek v4 will blow out anything you run locally for pennies (or gpt luna as of last wed.)

It can also set up its own custom connector if needed and run as needed via CRON.

This is a pretty solid list, besides Code Quality. Most of the links there are user error. It’s akin to blaming a Junior dev for breaking production imo and undermines the entire page.

The most recent release of SOTA models being both more intelligent and 20-30% cheaper is really fighting against this argument (Opus 5.5 & Sol 6).

Anyone have any recommendations?

This may have been true last year but nowadays AI is far better and more thorough, and faster than 90% of devs when driven by someone knowledgeable.