At this point the most likely outcome appears to be that we live in a simulation.
The logic works like this:
1) each new model is exponentially advanced
2) we can simulate entire populations of agents and watch them interact [<—WE ARE HERE]
3) the models will become so advanced that simulations become indistinguishable from reality
4) agents in the simulations become conscious
5) simulations becomes cheaper and easier to run, so we run billions of them - each populated with billions of fully conscious agents.
6) it is now mathematically likely that you live in one of these simulations right now (rather than in the base reality you think you live in)
This is Bostrom’s simulation hypothesis playing out in realtime. Unbelievable to watch.
This isn't just a stupid thing to mock, it's truly disgusting, because AI is becoming increasingly capable of formulating novel medical solutions, which means banning it from being advanced enough to do so is a death sentence to those with diseases it could help find cures to.
When we released ARC 3, I got asked, "when do you think a frontier model will saturate it?", and I answered "in about a year, though it depends on how much it gets explicitly targeted"
That was 6 months ago, so the progress that Astra represents happened about 2x faster than I anticipated. I think the speed of progress will surprise a lot of people, and what the new models can do will challenge the views of AI that people developed by using prior generations of models.
Benchmarking AI systems is a continual process that co-evolves with the models. New benchmarks challenge AI capabilities with emerging questions to shape the directions and feedback signal of the research process. Then they adapt as models progress, targeting the residual between AI and human intelligence.
We are still working on ARC-AGI-4, which we started developing after releasing ARC-AGI-3 earlier this year. It is coming Q1 2027. We think it's going to be really special.
We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today. Team is moving mountains to give access as fast as we can.
First one will land in ~ 3 hours. There is still time to create your account if you don't have one.
The future of transport is safer, more enjoyable & gives you more time back
With theater, music & gaming apps on a 22" touchscreen, Cybercab offers you a space to be productive or unwind
Soon to be equipped with V5 Starlink hardware, too, for complete connectivity
Digital is solved. Brockman said they have multiple models in reserve. Brown said scaling laws are holding.
Focus on robotics, sensors, wet lab automation and rapid admin automation plays.
Everything that needs to be funded will be funded.
Long compute. Long compute. Long more compute.
Most of you were born into a four or five decade stretch where society lacked imagination.
Something went wrong in the 70s. We lost our mojo for a bit.
We are about to get it back. Be optimistic. The future is here. We’ve done this before you just weren’t alive for it.
Los alamos. Freedoms forge. Apollo.
Accelerate, anon 🚀
Astra had made Codex more essential than ever due to its unique strengths. The computer use, vision and self-verification capabilities work incredibly with it. Using it in a terminal almost feels silly