Breadtube was a fun natural cluster early on until people got it too in their heads that it was supposed to be a leftist clubhouse where liberals didn't fit in.

they are putting my brain inside the computer tomorrow

First thought: isn't this what every IDE integration like Github Copilot is?
Second thought: that integration is asymmetrical. The editor highlights all the agent's changes to bring them to my attention and let me choose to keep or undo each one. The agent isn't shown my edits that same way.

I think the right just bet too much too fast on trying to capitalize on his death. Right-wing media personalities eventually realized it wasn't the ace to winning the culture war at the same time Trump pushed forward into multiple unpopular issues.

Nowadays everyone is a bit jaded from having encountered many rude players in voice chat and from the knowledge there are better places to attempt to socialize online.

And then there was the fact the game was a ton of people's main way of interacting with the internet for a brief time. Social media hadn't taken off yet, and people used their mics often to try to have conversations.

Felt this too.
The menu music and the main menu's background scene was so somber, and there wasn't much to distract yourself with in the menus outside of interacting with your friends list or joining a game.

for what it's worth, the vulnerabilities described in the Mythos article that it found and exploited were of usual kinds. There are no mysterious move 37s.
If AI discovers a new class of vulnerability that affects code patterns we strongly thought were safe then I'll panic.

One thing to keep in mind is that making an exploit to take advantage of a vulnerability is often much harder than understanding the vulnerability. Defenders don't need a working exploit in order to recognize that code allowing out-of-bounds writes is a vulnerability that's likely exploitable.

for what it's worth, the vulnerabilities described in the Mythos article that it found and exploited were of usual kinds. There are no mysterious move 37s.
If AI discovers a new class of vulnerability that affects code patterns we strongly thought were safe then I'll panic.

I don't think the milestone itself being reached means the answers to these questions have flipped right now, but I expect to be less confident on them going forward.

Questions like "Can this replace most programmers?", "Is this smart enough for recursive self-improvement?", and "Is this dangerous enough now to need regulating or stopping?" could be relatively safely dismissed before that milestone.

When ChatGPT was new and everyone was anxious to understand the specific pace of AI progress, I figured a significant milestone to watch for was when it could find and exploit brand new software vulnerabilities on command.
It's slightly unnerving to see the milestone met today.

"Over 99% of the vulnerabilities we’ve found have not yet been patched"... "Mythos Preview is capable of identifying and then exploiting zero-day vulnerabilities in every major operating system and every major web browser when directed by a user to do so ... oldest ... 27-year-old bug in OpenBSD"

I will ask claude

The post-quantum cryptography timeline accelerating at the same time that mythos is creating mass CVEs is really threatening to alter my roadmaps

I think some people excel under this pressure, but most people do it badly and end up overconfident on some wrong things.

that's a good example of AI agents' annoying tendency to assume some quirky code is meant to be like that and paper over it with a workaround instead of just correcting the code to be more idiomatic

kira is now better at posting than i am, apparently

i am losing my job to automation

feels like supporting monetization of views is necessary to attract and keep creators and I'm not sure how that can work in a decentralized/distributed atproto kind of system

The axios package that was hacked recently was a pretty legitimately popular (though dated) package so it didn't really have much to do with that.

the best-case scenario would be that it annoys AI companies enough to figure out and propose some actually effective AI safety regulations to do instead. Not sure how much stock I put into that possibility though.

me reading the replies to the @attie.ai announcement

Here's a cursed fact: in C, even normal 0 literals are octal

though for what it's worth, the specific Iran propaganda videos I've seen depicting everyone as lego characters were massively preferable to the Whitehouse's propaganda videos

I wonder how much Sora shutting down has to do with Iran using AI for propaganda videos lately. Kind of awkward for a business to be visibly associated with that.

If there was an intended personality test to it, it's "Can you understand other people can react to your patterns of behavior? Can you function in society?", not "Do you think the future is fixed? Do you think stuck-up Omega can truly understand lolrandom *holds up spork* behavior?".

Everyone misses that it's supposed to just be a very simple spherical-cow counter-example for decision theories that have no mechanism to let an agent consider how other agents think about their future behavior.

Sneerclub is great in a light brief dose of specific good threads but I'm convinced everyone that stays a regular of it develops that pathology or is there because of it.

I wonder if the professionals or companies he worked with insisted on it.