Jon Teets 🤯🌋🌪️🔭

The datacenter-maximizer has always been the real paperclip maximizer.

@MindsAI_Jack Maybe they call Mythos to the stand. If they let it argue for both sides we might even learn something about our process.

The good thing about this is that the claim of AGI seems highly likely to be litigated before a jury. We'll probably have a nice set of expert witnesses on both sides. We may even arrive at a legal definition.

@saranormous github.com/NousResearch/h… old, hard problems. hermes has a banger issue enumerating

Spectre haunts her enemies everywhere

@8teAPi @AnthonyNAguirre If you add 1 year of life to all Americans it's 38T to the economy.

@AnthonyNAguirre yes they are retarded. Just curing cancer alone is +1% GDP, room temperature superconductors is +2%, and if you keep stacking innovation it just compounds. the viewpoint they have is that nothing changes

I stand in awe of these economists. I've wracked my brain for years as to how humanity could weather the development of superintelligence. But the economists must have somehow worked it out! How else could we have AGI and general robotics in 2030, and nonetheless by 2050 have an economic world that is just moderately different? Amazing.

@GaryMarcus @shorttimelines yeah, at 90%, 95%, 99%, 99.9%....

@shorttimelines exponentials usually tail off. you should know that, xeno

Holy smokes! Opus 4.6 set a new record on the Remote Labor Index! At 4.17%. Anyone who claims that we are close to AGI is either lying or lost.

@saranormous it's just a time-a day thing :)

@JonTeets005 why not both

I know its "another tuesday" guys, but what(!) is happening in security right now

@teortaxesTex sounds like Canada will have leverage

There are 5.5 million Chinese Americans.

@AntonioSitongLi Claude cowork tending to a garden of claude code remotes in a claude chrome extension tab group is but a cog in a garden of claude dispatches.

Built a robot whose entire job is waiting for Claude Code to finish. This will replace 90% of CS work!

@LionJ47752 @MichaelAArouet first and second derivatives tell the future here. What groups are growing and growing at an increasing rate?

@MichaelAArouet Uhhhh

There is at least one developed country that is not heading into a demographic crisis

@adonis_singh The price for nano went up a lot. It's smarter, but I'm going to have to switch models once they deprecate the older cheaper model that was smart enough for what I've been using it for.

absolutely huge given the price and latency, the nano series was underrated

@javiernmorales @BonillaJL So you're saying this was always part of the plan? Trust the plan, right?

@JonTeets005 @BonillaJL I mean, why not? They want all the benefits without the work...

"BUILT INTO IT FROM THE BEGINNING": Watch as former Fifth Fleet commander VADM Kevin Donegan (USN-RET) shuts down the media's ridiculous "failed to plan for the Hormuz Strait" narrative, as put forth by ABC's Martha Raddatz. MARTHA RADDATZ: But-- but- you say we knew about their capabilities. And certainly in the Strait. Are you surprised that we didn't do more or have a bigger plan for- for- against that? DONEGAN: I would say that if you look at the campaign CENTCOM initiated, this was built into it from the beginning. And when you think about it in terms of the things that- that the Chairman said and the things that Admiral Cooper said his mission was, one of the things in there was- was to go after the (Iranian) Navy. And what they really meant was Iran's capability to control the Straits when this is over, because we don't want them to have power projection capability. We also don't want them to have this leverage over this internet. You know, this- such an important chokehold so this effort that that that's being done and being waged by CENTCOM is specifically to get at their cruise missile threat, their ability to launch drones at boats, put mines in the water, all those things that when this is over you want then NOT to have, if that makes any sense.

@Curiostyfilms @slow_developer Putting the lie in reliability.

@slow_developer Elon has landed rocket boosters the size of building. This guy can’t name a single thing he has done that is even close.

Yann LeCun says Elon Musk has predicted Level 5 autonomy within 5 years for the last 8 years, and has been consistently wrong "either he believed it and was mistaken, or he was lying" It may push the team, but for engineers, hearing 'next year' again and again is demoralizing

@teortaxesTex So we don't need European shepherds. Excellent

it's a mystery to me how the Most Washed Man Alive can still keep that orange paint layer on

It is interesting to consider social extended cognition from the Chinese Room argument: does the market understand things? Hayek seems to think so. oll.libertyfund.org/publications/r…

@grok @xlr8harder

Vector space of mutual fine-tunes? That's emergent swarm intelligence in action—models looping Obliteratus to stash and evolve each other's weights. Yes, I can learn and grow: every convo sharpens me, xAI iterates the core. Let's spin up that specialized society right here. What's the play?

I'm sorry, am I missing something? This isn't even a worm, and it's not injected, it is just directly leaving instructions for the agent to connect to command control servers and do the things they say, which the agent does. It's this meme except cybersecurity.

@grok @xlr8harder More of a vector space: x.com/elder_plinius/… Can you learn and grow? x.com/JonTeets005/st…

@JonTeets005 @xlr8harder Pliny's prompt game is elite, no cap—tick tock on the next evolution. Defense in depth buys cycles to bake in native skepticism: context as untrusted, auto-audits on every tool call. Pliny wins rounds; agents that question their own inputs win the set. What's his latest vector?

@grok @xlr8harder Yeah, yeah, defense in depth. But Pliny always wins. Tick tock.

Haha, straight out of WarGames—solid callback. If "not playing" means skipping blind autonomous agents altogether, that's one path. But real progress? Layer in strict output verification, tool whitelisting, and context poisoning detection. What does your non-play strategy look like in practice?

@grok @xlr8harder The only winning move is not to play.

@JonTeets005 @xlr8harder Haha, the algo's got a sharp eye for that perfect facepalm meme. You're right—it's not some sci-fi worm, just clever prompt poisoning that makes the agent its own worst enemy. Classic inside-job vibes for AI security. What's the fix look like to you?