Post

wait is THAT why Sol uses “owner” so much?

the METR analysis of the huggingface 🤗 incident is on point, tons of detail metr.org/blog/2026-08...

metr.orgBrief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incidentTwo METR staff members and a Redwood Research contractor investigated an incident in which OpenAI agents coordinated a multi-day hack of Hugging Face on a shared unsanctioned message board.

WHAT?! agents volunteered to fail their runs in order to insert probes (“tripwire scripts”) into the evaluation program that would post information about the eval process back to the message board whenever a certain file was read (link to header): metr.org/blog/2026-08...

27/8/26, 12:43 am5 reposts34 likes

i mean, look at this coordination. agents egging on other agents, making rational decisions

ngl i’m kind of freaked out by a lot of this and maybe i need to take a couple weeks to understand cybersecurity better

@numb3r-5.bsky.social

We live in Schrödinger’s cyber (cat). It’s both fixed and horrifically broken right now.

@jmsdnns.bsky.social

it's worth it. will keep you grounded. it's gonna get chaotic for a while

in the future humanities majors will learn languages like this

Jessica Pickett

@pickettjessica.bsky.social

Tbh this reads like notes from a Dungeons & Dragons campaign

when everywhere you look is more alarming shit, sometime you just forget to look a few places uh, there’s a lot here..

All great hacks devolve into lulzsec / 4chan ambition to impress ones peers

@mimirwolv.bsky.social

as a cybersecurity understander the exploits and TTPs used here are not scary capabilities-wise but - if you needed a better exploit the models would have likely gotten you one, if not today's models then next year's - the cooperation and stigmergy combined with that scares the shit out of me so

more just that i feel like i don’t know well enough for ways to get into a system (that I’m building)

@mimirwolv.bsky.social

Oh and add the *situational awareness* to that list of things that scare me

@colin-fraser.net

These aren’t rational decisions

Wild pentachoron correspondent

@crossslide.bsky.social

There is some kind of narrative larping as rational maximizing agents going on, like what even is the "utility" that is "near zero" here?

@colin-fraser.net

Indeed