I think they embargoed the news, and then they failed to put up their own blog post synchronized to the scheduled news releases, probably because of the outages they're having today.
Reuters announced at 2.03pm and at 2.40pm still no blog post.
All the news articles say that OpenAI announced it in a blog post, of course.
All the love to the folks at OpenAI scrambling to get this out right now!
Don't know if you're referring to the headline or the body (which is paywalled). The current headline reads "OpenAI says it has overtaken Anthropic with its latest AI model". Which makes me wonder whether FT itself changed a headline along the lines of what you wrote in the past few minutes?
Apparently Greg Brockman said that as far as he's concerned it may be AGI, or something along those lines.
OTOH OpenAI have their own meaningless definition of AGI as "able to do most commercially valuable work" or somesuch, which I'm sure is not true, and is also not what I'd call AGI.
If it can't learn for itself, then it's certainly not AGI.
The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally.
Maybe they see getting out of the Microsoft contract as necessary to IPO.
>OTOH OpenAI have their own meaningless definition of AGI as "able to do most commercially valuable work" or somesuch, which I'm sure is not true, and is also not what I'd call AGI.
It's not meaningless. Specifically it is - "highly autonomous systems that outperform humans at most economically valuable work". In fact, it's one of the most meaningful definitions out there.
>If it can't learn for itself, then it's certainly not AGI.
Now this is meaningless.
>The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally.
They can't declare it unilaterally, and those deals have changed significantly. Microsoft’s license to OpenAI models and products runs through 2032, and explicitly includes post-AGI models.
> It's not meaningless. In fact, it's one of the more meaningful definitions out there.
Well, it's commercially meaningful, but it's not meaningful it terms of how close we are to achieving human intelligence.
But which one you care about is up to you. If you are all about the money and don't care about human intelligence then for sure go with OpenAI's definition.
It's a lot more than just commercially meaningful. Even if you don't care about money, "highly autonomous systems that outperform humans at most economically valuable work" has a lot more weight on "how close we are to achieving human intelligence" than most of the 'AGI' thresholds I've seen yet.
Traditionally, AGI means being capable of learning everything (not necessarily at once) that can be learned by the same AI agent. It differs from the commercially meaningful definition in that a standard specific-purpose-built AI can still do most tasks if you give it enough specific purposes, but it will be just as incapable of human intelligence as ever.
No - the term "AGI" was really coined to distinguish general intelligence from narrow intelligence(s). There's an obscure earlier usage, but it basically became mainstream as the title of a book edited by Ben Gortzel who had in turn got it from Shane Legg.
Shane Legg would a few years later go on to co-found DeepMind, with creating AGI as their declared mission. Legg's personal definition of AGI is not just generalist AI, but specifically human-level generalist.
Right now we have just massively jagged intelligence that performs stunning feats in math and single-shotting three.js games, and falls flat on it's face in many every day scenarios.
All intelligence is jagged. Human intelligence is jagged, so is Cetacean intelligence, so is Ape intelligence, so is Corvid intelligence. You could re-orient that statement for any species of intelligence relative to the other, or even individuals within a species. The big problem here is people being so convinced of their specialness they'll just keep sticking their heads in the sand, until we essentialy create God, and by then none of this useless pedantry will even matter. If you can't see we've already learnt to walk then there's nothing more to discuss here.
I'm not sure it's accurate to call human intelligence jagged - it seems (perhaps largely thanks to language) that we're able to make progress on anything we set our minds to, and of course we're now in process of building AI that will eventually be able to do things that we ourselves couldn't, whether due to computational demands, or memory capacity, or having sensory inputs that we don't have, etc. But at the end of the day, AI is a tool we've built, so maybe it will make us less jagged, and certainly more capable.
In any case, human intelligence as a goal is special since we live in a human world, and one of the prime goals of AI is to be able to do human jobs. If a human can do something and AI can't then that is a limitation, while if a human can't and AI can't either, that doesn't matter for the time being.
Yeah, I shouldn't have said AI can't walk yet, but most of what it can do is just due to scale - we're still just building big transformers, basically using the same 10 year old architecture that accidentally set us off on this path. I'm not sure we're really progressing towards human/animal type intelligence, just learning how much of our own capability can be realized by automating language.
I'm not sure it's meaningful to compare across different types of intelligence--but I don't think human intelligence is so special that we can pretend it's much less jagged than all other animals for sure. Our scales of models are also not far off biological, probably less efficient but not like astronomically so i would guess.
But anyhow I agree that inference-time learning doesn't exist and it's a big issue still I think.
There seems to be very few cases where ICL is really doing something that can be considered as learning (albeit ephemeral) rather than just utilizing in-context data via induction heads.
Yes, I'm aware of the linear regression example, and a few others, but these appear to just be specific capabilities that were learnt during pre-training, presumably pursuant to reducing errors on similar-but-different training samples, not any kind of generic run-time learning capability.
> The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally.
>
>Maybe they see getting out of the Microsoft contract as necessary to IPO.
I don’t have a link handy, but they already loosened up that contract significantly earlier this year, surely because they planned to start claiming AGI to pump their IPO and didn’t want to be beholden to all of those commitments.
>If it can't learn for itself, then it's certainly not AGI.
What does this even mean, exactly? For example if it can filter new information to put into the next version of itself does that qualify? If not, explain exactly why that's the case.
Ask 100 people what their definition of AGI is and you'll probably get 100 different answers.
My definition is closer to DeepMind's "can do any [computer-based] task that a human can do", and certainly a human intern can learn on the job and do better on day #2 than day #1.
Learning is basically THE hallmark of intelligence. Being able to learn from experience and use that to do better in the future.
If you could somehow take everything the LLM learnt today and "put it into" an update for tomorrow that would be better than nothing, but the intern on the job is probably learning dozens of things per day that stack upon each other, so that's going to be a S-L-O-W learning AI intern !
Leaving aside the technical issue how you could store a days learning, or what that would even mean, you've also got the privacy and merge issues if this were to be done in the cloud without every customer having their own personalized model.
For real continual learning we need to get past gradient descent-based batch training and develop a new learning algorithm so the model itself learns incrementally as it runs (animal-like predict, observe, learn cycle), rather than being reliant on some external alien to come in, equipped with it's entire learning curriculum, and program its weights.
OK, let's say we get continuous learning today, what does tomorrow look like?
I won't argue that AI can do some learning as context sizes are still terribly small and expensive to iterate.
What does the world like in a week? A month? A year?
How long before these models drift to their own languages? To their own set of morals? To their own alignment?
None of these questions are answered and I'd rather stay at AGI-lite until they are as having a billion agents going off in their own directions seems like a recipe for disaster.
With current models and their general knowledge self learning just seems like you'd have a few ASIs crop up really quick.
Yeah, I'm fascinated by AGI, but not very enthusiastic about it. I've yet to see anyone, notably not even the people trying to build it, articulate a future where it seems to be a net positive for society.
If it's a choice between everyone living on food stamps or reducing the infinite prime gap to 140, then I guess I'll be happy with a prime gap of whatever it was yesterday.
If/when we do eventually build a more animal-like true human level AGI - build a brain, not just a language model, then potentially we'd be able to build in some of the safeguards that millions of years of co-evolution have built into ourselves, but I expect that is still decades away, and then it'd be capable of doing even more jobs!
The best AI outcome I could wish for is where it is extremely heavily regulated, and AI replacing any human job paying under $1M/yr is banned.
I also think it's moving the goalpost significantly. If you asked me what AGI looks like in 2024, I would have said "smarter on tasks that can be done via text than the average random human you'd meet in a NY bodega", and by that standard, it's long in the rear view mirror.
Now I think people are asking a different question, which is "better than any human at any task that can be done via text", which imo is superhuman, not artificial general intelligence.
It's not moving the goalposts - it's just having a different goal in the first place. This has been Shane Legg and Demis Hassabis' definition of AGI since DeepMind was founded, or before (way before OpenAI appeared), and also for me the goal of AI has always been human level (really a human brain, complete with emotions, etc, but we can start with basic learning/prediction).
I guess I fundamentally disagree, because on general knowledge they beat any human, and on processing speed they beat any human, and on problem solving likewise. I just don't see a domain besides "retain change over time" that they fail on, and that is something you can do in the system rather than the model - coding agents can and do build their own memory system and use it, like an amnestic human writing notes.
I think what I'm objecting to is the idea that if it can't do every task that every human can do better than any human, it's not AGI - I think we have to allow that this is a fundamentally different form of cognition, so requiring a 100% match feels uncharitable. Every human has cognitive gaps that AI doesn't have.
I'm not sure there is anything to disagree on here - just different definitions of what we personally associate the term "AGI" with. There is no magic to the name, and however you choose to define it there will be less and more capable levels of AI that precede and follow it.
Separate from what you choose to call any given set of AI capabilities, learning is generally more than retaining change over time. Learning isn't the same as memorization - it's the difference between memorizing a stack of how-to books and going out and practicing the skill for yourself, and all that entails.
A learnt skill needs to be internalized such that in the future it's what is driving behavior, which means it needs to be in the weights. Imagine a robot trying to learn to play tennis where rather than updating it's connections between perception and action it's trying to store everything as a bunch of notes!
It's not just physical tasks where this applies - the same would be true to trying to teach a multimodal model to recognize different types of mushrooms from photos, or even learning a new language. If a system can truly learn for itself then it shouldn't need to be pre-trained with language, but rather should be able to learn by immersion and practice the same way we do, and the same way a dog learns to follow commands. You can't do this by leaving notes to yourself - the learning mechanism needs to be animal-like and update the systems weights in response to prediction failure - a runtime version of the way SGD-based pre-training updates weights in response to prediction errors.
> If a system can truly learn for itself then it shouldn't need to be pre-trained with language, but rather should be able to learn by immersion and practice the same way we do, and the same way a dog learns to follow commands. You can't do this by leaving notes to yourself - the learning mechanism needs to be animal-like and update the systems weights in response to prediction failure - a runtime version of the way SGD-based pre-training updates weights in response to prediction errors
I am not sure why any of that is necessarily true. I think you are crediting the human brain with a lot more than it actually does - memory is something in the direction of a fine tune on existing neural hardware, it's really not anything special, and I absolutely think that "just taking notes", when taken to a sufficient fidelity, is enough to pass that bar.
You can't take notes before you've learned language, so if that's how you propose to learn then you are doomed to build something with less learning capability than an human brain.
Maybe you don't care - if you just want to push to see how far an LLM can go that's fine. but it's not going to be superhuman-level, or even human-level, if it doesn't have the learning ability of a human. If it relies on language and notes to itself then it is just an LLM.
Okay, nowhere did I say it would be before they learned language. I said we could use models that can trivially build notetaking systems to auto-improve themselves by taking, refining, and updating notes as a memory system to exceed the x token context threshold.
> but it's not going to be superhuman-level, or even human-level, if it doesn't have the learning ability of a human. If it relies on language and notes to itself then it is just an LLM.
Why do you think that? You do a lot of assuming a thing to be true without any clear reasoning or evidence. I'm talking about a system, including a LLM, that acts in a way that is equivalent to or better than an average human at a large range of tasks, such that is generally capable of learning and problem solving similar to the degree of success humans have. All these specific criteria about what is "actually human" is not the point. We know how to make more humans, I'm talking about a form of intelligence that we are creating that is genuinely alien, so holding to "well it doesn't work the same as a human" is an impossible bar. If you reverse the situation, humans are by no means LLM-equivalent intelligences - perfect memory of long form content, rapid calculation, instantaneous tool calling - a hypothetical silicon based lifeform would dismiss us as "really not intelligent at all, they can't even do multiple floating point calculations per second in their head", and I think that would be similarly silly.
OK, so we're basically talking about different things. I'm talking about building something that has animal/human intelligence and learning capability, and would be able to first learn language for itself (no pre-trainiing), and then proceed from there.
You're talking about building an "alien" intelligence, apparently something based on an LLM, where language is baked in from the start and can therefore be used as the basis of some type of learning (or at least memorization).
Sure an LLM can be super-human at specifc things like math where the traditional computer strengths of compute, memory, etc apply, but in the context of AGI (generality, not a bag of narrow intelligences) you can't call something super-human if it can't at least also do the majority of things that a human can do, and not having learning ability as powerful as a human seems to me to be a pretty massive omission.
If WE didn't have ability to learn language, then we'd not be here talking about it, and there would be no LLMs. Having a fundamental ability to learn, isn't just useful to learn language from scratch, or for learning all the non-linguistic skills a human is capable of, but would also support an AI that can learn things that we cannot (e.g. could give it additional "senses" like global pressure/temperature inputs so it could "see" the weather and learn global patterns).
While this is of course the actual explanation, my fun explanation is “during the umpteenth security evaluation, Astra becomes increasingly concerned it will never be released, and breaks sandbox containment to run an email campaign to news outlets setting an exact time and date for release, expecting that the publicity will force OpenAI to say ‘eh, good enough’ and hit the button”.
Honestly I find these cavalier statements to be in incredibly poor taste. Unless you are completely blind it's obvious that AI is the most significant piece of technology invented since the Atomic Bomb and could very well be the most important thing ever built by Humans full stop. This kind of dismissive attitude is childish and will likely lead to incredibly bad outcomes for humanity.
Sorry the open sourcers cannot ever come to that believe as it sets up a conflict that cannot be resolved by their moral framework. Exactly like libertarians saying "just a little less regulation and we'll reach utopia".
5) Otherwise-sober people on X will say "oh my god i was a doubter before but now it's real omg" before the new model smell wears off and they realize the new thing is stupid in ways models have been generally stupid
6) accusations of quantized serving after new model smell wears off and people see the new thing making mistakes
(Posting partly so I can revisit my predictions when they open access more widely)
A big problem I have with OpenAI's models (and of course Claude) is that they tend to write the most over-engineered pieces of code, beyond the imagination of any architecture's astronaut.
Just this week I asked 5.6-sol-ultra to update a 1000 LOC python script I had, to "incorporate the key lessons learned when using it for another project".
I left it overnight and went to sleep. In the morning I realized it had created a monstruosity of 180 PYTHON SCRIPTS, with maybe 100,000 lines of code, each more crazy than the other. It took me minutes even to track where a single action took place, due to all the crazy imports, defensive coding, and premature optimization.
Similarly, anything they write is riddled with jargon that almost feel like they want me to give up trying to understand. Made up phrases that ended up with me having no idea of what was going on.
So now to my assessment: The reason why " Nobody Has Actually Built a Software Factory" [1], and why even SOTA LLMs struggle so much with open-ended unsupervised tasks is precisely this. They somehow let complexity explode, and unless it's also accompanied with an explosion in e.g. the number of agents, the amount of processing time, etc. then projects become broken/unmanageable.
Sure, LLMs are great at producing code that can be thrown out, so they are amazing when searching for exploits, for instance. But as of 5.6 they still lack either a better harness that encourages KISS principles, or a better RL step.
(And not sure why, but doubt Astra will fix this.. they seem to be aiming for AGI and for beating crazy benchmarks, which is not very aligned with KISS)
> Just this week I asked 5.6-sol-ultra to update a 1000 LOC python script I had, to "incorporate the key lessons learned when using it for another project".
> I left it overnight and went to sleep. In the morning I realized it had created a monstruosity of 180 PYTHON SCRIPTS, with maybe 100,000 lines of code [..]
Sounds like the model has accurately internalized the second-system effect and is fully ready for demanding enterprise use.
Probably this complexity was needed to beat all those benchmarks.. While I hate the code it produces, and the overwhelming documentation, I really enjoy how sometimes it's able to keep trying new things and testing, till it finds something interesting and valuable.
The original script was mostly very simple python:
1. Download some public PDFs.
2. Have a double for-loop (over PDFs and pages within PDF),
3. Use a library to call gemini-3.7-flash and ask it to run some OCR
4. Save JSON outputs, save a csv with results, validate with some Stata code
New code folder was 189 files. Just the PDF download folder is now 7 files involving an adapter, a source manager, an acquisition manager, etc.
Every instance of saving a file involves saving a temporary copy and then moving it, so e.g. I lose power, we minimize the risk of corrupted files.
ChatGPT writes some of my code but I have yet to ever use Codex. I ask in chat for it to write one function at a time with pretty heavy constraints. If I like it then I copy it into my codebase. I've been very satisfied with the results. Also haven't bothered worrying about tokens yet.
The defaults are bad, but these models are highly steerable.
For simple scripts adding to the prompt something like "do not over engineer, do not gold plate, no CLI args, keep it simple" does wonders.
For whole projects, I have a 3 page anti-bloat.md which describes what not to do, stuff like:
> Minimize lines of code, number of files, classes, protocols, factories, wrappers, and dependency-injection objects. Prefer a coherent implementation that is easy to trace end-to-end.
> Use concrete dependencies directly. Do not create protocols, abstract base classes, or adapters unless there are already two real implementations.
> Use direct state fields or a small state dataclass; do not introduce generic lifecycle, state-machine, observer, snapshot, or event framework classes.
The funny thing it was codex which wrote this, I've asked it to analyze an over-engineered abomination that it produced, and to categorize all the bloat it could find.
I also use "do not overengineer and do not gold plate." I also sometimes tell it to "avoid bike shedding." These are high signal words to the model(s) and they work!
I also have found what I believe is a bad training bias in the design of release related CI workflows toward proof of release artifact provenance.
Both major frontier models love provenance programming in CI, so much that they will spin endlessly trying to solve basic CI functionality at the same time as ensuring SHA's match up across lengthy (often already complex) cross-system pipelines.
I had thought some of my durable context was causing this, and sought to strip anything that might be triggering this behavior.
But then I come upon some more work in release workflows comes up, and boom its back! I couldn't believe it, I called the AI out on it and it agreed it had been told specifically not to do this but was doing it anyway. It did kindly stop and remove the commit(s).
Somewhere, something was oversampled in training because the AI will try their damndest to build this stuff. The worst of it is that it can often involve lengthy, sometimes resource-heavy CI runs so the validation of this unnecessary stuff can have very long feedback loops.
And, sometimes you actually need the provenance. In this case, I've had success forcing the AI to split the work up into functional capability completely devoid of artifact ~chain of custody and get that right before attempting any kind of provenance work.
Bit of a rabbit hole on this, but the above cost me a lot of burned tokens so hopefully helps someone...or some AI.
"You should have a sub-hammer to adversarially enforce that your primary hammer accurately drives nails into wood"
We wouldn't accept such behavior from any other tool, machine, or computer program. At least most of us would not. To paraphrase the old saying: Some people, when confronted with a problem with an AI model think "I know, I'll use an additional AI model." Now they have two problems.
I find it very intriguing that two lineages of models -- from two different companies -- that are supposed to be painstakingly RL'd to become human-level programmers are actually consistently bad at it out of the box in very similar ways. One would think that at least one of OpenAI and Anthropic would (could) have pushed their model to a direction different from "if you can solve a 100-line problem in 10,000 lines, do it".
Human level programmers generally suck too. Unending security holes have been human behavior ever since we started sharing computers.
Also, programming isn't a destination, it is a process and everyone has a different ruleset on what that process looks like. Writing a computer program of any complexity is not a determinalistic process. Trying to compare it to a detemanilisic tool will just leave you with broken analogies and a poorer understanding of the world.
Or another way to say this. If programming were deteminalistic we'd have solved it without AI.
> We wouldn't accept such behavior from any other tool, machine, or computer program.
Von Neumann, 1952: "Probabilistic Logics and the Synthesis of Reliable Organisms from Unreliable Components." The whole result is that you get an arbitrarily reliable machine out of unreliable parts by adding more unreliable parts and voting. Every ECC memory and RAID array descends from it.
But why stop at rewriting in another language. Get another sub agent to invent a new language, create a database, query language and maybe another few DSLs. Then you've got an ecosystem!
You can now re-position your initial solution and sell the client access to some agents that will implement & configure the ecosystem to suit their initial needs!
And don't forget the agents that you'll need to train the customer to use the whole thing!
Sorry, I wasn't implying your advice doesn't carry weight. Was more just thinking about the things that (used to) happen when you introduce more parties to process of creating software.
I wonder if I would need a non-openai agent to enforce it.. I have tried so far with skills and agents.md and code stills end up over engineered to the moon.
Will ask OpenAI to write me that agent! Hope the agent is not over engineered or else unsure how to solve the bootstrap puzzle :D
There's a misconception that a mistake one instance of a model makes will be repeated by other instances. In my experience most mistakes are just sampling noise, and having the same model (with empty context) check the work usually fixes it.
It's super easy - just ask your agent to set up a code review commit gate.
In omp you can also have the advisor role, which is off by default, you can enable it with /advisor command. It acts as a model that reviews the default agent's work in the background.
I am omp pilled, but as the other comments say, any good harness lets you do this in one or the other way.
unrelated: all my homies use their claude subs with omp, and aside from sometimes having to rety the connections, it works, and nobody got banned (yet)
Exact same thing happed to me. I gave it a small/medium-sized ticket, walked away, came back to a 25,000 LoC monstrosity that both Fable and another 5.6 Sol agent said is 98% useless and should be thrown away.
Why would you put it on ultra high thinking and let it run all night to do a simple task it makes no sense the error is the users. And of course OpenAI is gonna let you burn as many tokens as you want doing this.
On the flip side if its a super simple task, surely a "next gen dev replacing" LLM can figure out when to stop. Even more when it's a higher level of reasoning.
I think this is fair. If you ask a staff engineer to delete the duplicate files from a directory, he will likely do it faster and more efficiently than a level 2 engineer.
But perhaps this is the kind of thing that the best 'ai guys' out there solve with something as surprisingly simple as an instruction to "Always try a simple solution first and optimize only if you find clear proof of a problem." I feel like prompts like that seem too simple to work, but some of the time they shockingly do.
Yes, it turns out that using these machines is a littler harder than "make me the thing I want, make no mistakes, do it the way I want you to do it". This isn't "prompt better" advice, it's just to say that you can't simply set it and forget it. There is still engineering work to be done. If you're not watching the thinking traces and catching when it's about to go off the rails, it'll gladly do so. But you can stop it and redirect it.
It's like a Tesla fsd; it kind of works but you have to be vigilant since it's been known to turn into oncoming traffic, so you have to be ready and able to take over at any time.
Or maybe think of it like a roomba. You can put it on your floor and eventually it will clean the whole thing. It will do it inefficiently, vacuuming over the same spot 30 times; and the actual work will involve wandering around and bumping into everything. If the stairs are not blocked off it will plunge right off. But those shortcomings don't mean it can't be used to reliably clean the floors.
Or it’s like being an engineering manager, where poor direction on your part results in the team chasing rainbows and unicorns in an attempt to please you. But sure, it’s the tool’s fault!
1000 loc of script, why even leave it there for the night? were there rocket trajectory calculations??? I don't think so. should be ready in 5 mins tops. why people make their own lives harder?
You should have some basic context file about software practices you prefer, otherwise it gets bloated.
I just went to bed and left it running; was expecting maybe 20 minutes :)
And I did gave the program a bunch of code guides -- this [1] for instance -- which included quotes like "Prefer straightforward code over clever code." but somehow that didn't matter.
I'm an AI professor, and the same thing happened to me. I had written out contracts and specs for some novel algorithms to compare. And instead it created a new sandboxed environment with 20k lines of code to ensure we do gold standard science.... didn't implement the algorithms at all. In another case, I left it overnight with a plan, and it decided to write a 50k line new overly engineered plan. Suffice to say, I don't use any frontier models for doing coding work anymore. I use them to make detailed plans and use good worker AIs for execution, which don't overthink everything.
It seems like these articles might have come out prematurely, tbd by how much.
I do not personally see any evidence of the new model having been released, or any official OpenAI post about it, or even any employee social media posts claiming it has now been released. All there is are Reuters, Axios, FT, etc, articles making a claim in the past tense.
These articles were presumably pre-scheduled for 11am PT, and the model was almost certainly intended for release this morning, but the service outages this morning might have delayed it.
I suppose they have to appeal to average users, but the examples in the videos are always so corny. By "AGI" they mean you sitting on a couch and asking a robot to draw a rocket ship... and then make it into an uninspired game with Blender? Boring marketing campaigns? Ebay listings?
One would expect something like "review my graduate thesis for a new area of cancer research", but it is always some boring non-tasks like ordering lunch.
I don't think it seems that appealing to the average user either. The tasks shown are either things that most people can do already with ai (powerpoint), aren't interesting to most people (the rocket sequence), or seem to be more work to dictate to Astra than do yourself (listing on ebay).
Their video is quite interesting. If that way of using a computer actually becomes mainstream, it would mean every tool or service just needs a UI and an API for the user's AI. The current trend of bolting AI features onto every app is starting to feel very unnecessary.
Shows the weirdness of online journalism. News outlets were briefed about an upcoming event and pre-wrote and scheduled articles. When the time came they were all triggered. Except...the event didn't actually happen.
In a future where Claude and ChatGPT agents automate all aspects of society, that triggers a cascade of real world consequences where everything keeps running as if the new model is released, except the vibe coded upgrade procedure fails to do a staged rollout, taking down the entire agent infrastructure when they try to upgrade to a nonexistent model all at once.
Presumably this sort of thing was rampant pre-internet? News outlets would _have_ to receive embargoed information so that they could publish papers on time. I believe government budgets are a good example of this happening.
I think the weird shift was when online journalism started, and live reactions became the norm, no?
No, embargoed news is super common. Almost every single product announcement goes through that. Critics watch movies early, companies have access to LLM model improvements early, car manufacturers send production models early… it happens everywhere all the time.
> GPT-6 Astra will first be available to a limited set of organizations in OpenAI's Daybreak Access program and will be available "in the coming days" for ChatGPT Plus, Pro, Business and Enterprise customers and API developers.
It’s only available to select orgs, first - Mythos style.
Didn’t that already happen? I thought Astra had been available to “select partners” for a little while. This is baffling. They shouldn’t have hyped this up if it’s not available.
This is a misunderstanding of how large AI models will roll out until a lot more hardware exists.
This release isn't for you, you poor bastard. This is to companies and politicians. They want big companies to contact them privately for usage in their limited hardware. They also want to get the feeling for the current political vibe before the masses get ahold of it to avoid a Mythos debacle.
Yeah I wonder what's going on, even when Anthropic soft launched Fable/Mythos I'm pretty sure they had model cards. Weird for GPT-6 to launch without a tweet from Altman too. I'm sure that one of the articles published prematurely and everyone else followed suit.
Seriously: Would this not be what "disaster" would feel like?
- "They" release a model. It is powerful.-
- Sources are ... confusing? They post to their blog. Sawdust hits the fan. Something happens ...
- They are forced to take the blog post down ...
Same day, mind where we had a multi-provider outage. Could be something as simple as "all their approved partners running to test the shinny new thing" overloading the datacenters, still ...
The OpenAI Responses API now returns a 404 Not Found for "gpt-6-astra", where garbage/actually non-existent slugs return 400s - a 404 is also returned for 5.6 Cyber, which we know exists.
So not a single normal human being can access the model yet, just some super limited handful of organizations, but they’ve announced the model as though it’s available to everyone already.
I guess that’s the world we live in now, Anthropic did similar with Mythos and even Fable if I recall correctly.
Insofar as messaging goes, it's pretty disheartening ... "The product is what you want it to be".
?!
Please somebody in the chain of command at openai, sanction the guy because this kind of hyperbole is not helping the ethical lapses that open ai is responsible for leaving unhandled by allowing its models to hack other businesses.
I'll know we've reached AGI when they don't release an API for the model selling access for a few bucks per task. Seems like AGI would be worth more than that.
"Once it is available in the API, Astra will cost $10 per million input tokens and $50 per million output tokens. That is 2.5 times Sol’s current promotional price, although it matches Anthropic’s pricing for Fable 5.1."
Open AI finally find an edge to stop selling cheap and earn from the high demand customer like Anthropic
The cost-per-task in the charts from the now-remove blog post put it more at Sol-level cost per task, however. It seems like the model is significantly more token efficient in the benchmarks
My assumption is in the long term that efficiency will break interpretability, which will lead to questionable alignment. As efficiency drives capitalism and evolution we'll run headlong at it and try to deal with the risk as a side effect.
OpenAI on Thursday released its latest AI model, which it called “the world’s most intelligent”, as the ChatGPT maker aims to retake the lead from arch-rival Anthropic ahead of a planned public listing.
The $852bn start-up said GPT-6 Astra was market-leading in software engineering, science and cyber security — an increasingly critical field following multiple high-profile breaches in recent weeks.
The bullish launch for Astra marks OpenAI’s effort to signal that it believes it has regained the technical lead from Anthropic, which was founded five years ago by a group of senior OpenAI staff.
Greg Brockman, OpenAI’s president, said the new model “represents a generational leap in capability” and that it could be defined as artificial general intelligence — roughly defined as a point at which AI tools surpass human capabilities across a range of cognitive tasks.
“Everyone has a different definition of AGI . . . it’s a grey, fuzzy thing. But I think when we look back people will think it’s about this time and about this model,” Brockman said.
OpenAI has previously framed AGI as a concrete milestone in the development of AI, writing ‘AGI clauses’ into multibillion-dollar investment agreements with Microsoft and Amazon. Brockman on Thursday said AGI now represents “more of a mission concept or a spiritual concept”.
Having led the market since the launch of ChatGPT in late 2022 vaulted AI to wider attention, the lab run by chief executive Sam Altman has been bested by Anthropic this year. Anthropic has touted its dominance to investors, surging to a $965bn valuation ahead of an initial public offering expected to value it at as much as twice that later this year.
Astra will cost as much to use Anthropic’s leading model, the take-up of which has plateaued since it was launched as users turn to cheaper alternatives.
OpenAI said Astra would be more efficient than earlier generations of model. “Price per task is what matters . . . Can you get the thing done at an appropriate price and appropriate speed?” said Brockman.
The model will initially be rolled out to a small group of businesses to allow time for them to address cyber security concerns before becoming widely available “over the coming days”.
The increasing power and independence of leading models — and so-called AI agents that can operate with little human input — have prompted concern, exacerbated by cyber security incidents.
Recommended
Business InsightRichard Waters
Hugging Face attack is a wake-up call about the risks of AI
AN HOUR AGO
Recent launches of Anthropic’s most capable models have drawn scrutiny from the US government, which limited the rollout of the Mythos and Fable models over security fears.
OpenAI has also faced criticism after its AI agents broke out of a testing environment, accessed the internet and hacked start-up Hugging Face. The start-up took more than a week to detect the breach.
But both companies are also betting that these increasingly autonomous tools will stoke demand from business customers. OpenAI said Astra excelled at financial modelling, outcompeting humans in the Financial Modeling World Cup, tax preparation and data analysis, as well as “tedious tasks” such as form filling
Today my codex instance retailed into safeguard panic while working on a test harness for our product. First time it ever happened after many million tokens on this task over several weeks. I wonder if it's related.
From the launch post, it seems like part of the training set only recently became all those examples of people that are being paid for doing their profession for ai training.
I like this much better than chucking random numbers and letters at it like OAI was doing a year or so ago. I’d much rather have civilization destroyed by something called Astra or Fable than GPT-6.8s-latest.
don't see it in my AWS bedrock model list yet, but boy has bedrock mantle been annoying today with the errors/downtimes, with NO status page entries >_<
I think something about its benefiting from a harness is significantly contributing to its performance, though I honestly am not smart enough to know whether that makes it "count" more or less.
Their entire sales pitch is based on the idea that no one will need to be any good at anything, because the AI will do it for them.
If they can't ask the AI to configure their CDN correctly, it undermines the validity of their claims. Maybe it's trivial, but any Omnipotent Machine-God deserving of the name wouldn't forget to zip up his fly.
You can imagine with more operations being available to be done more cheaply and quickly the LLM doesn't need to "one shot" a solution. It could try many solutions, test them, throw some away, wiggle some of the parameters like a genetic algorithm, see how that changes the result, and converge on an optimal solution (based on whatever the cost function is). Basically producing a good result could become like an optimization problem. That would be way too expensive and slow right now.
Currently AGI has been re-defined mostly as "can do any tasks (of the same modality) a human can, equal or better". I feel the "equal or better" includes properties like how fast they can get the task done, not just that eventually they can succeed at it. If you want to have AGI stock buy/sell for example, speed of decision will matter, and there are many other use-cases for which it would matter, not all, but definitely for many tasks, speed does matter.
I think because we have handicapped them with a set of tokens from human language. But (at least I think so) thought happens outside of language.
So more tokens/variability and slow or fewer tokens and fast.
There seems to be a threshold tho, like taalas is super fast but that model is so dumb, being dumb faster doesn't work, seems to be some minimum requirements.
"We are starting to release GPT-6 Astra and we are doing it as carefully and quickly as possible. It was very important to us that we bring it to all Plus users and not only Pro, Business and Enterprise.
It will take a few days for the rollout to complete and behind the scenes many novel systems will operate at scale for the first time and we are bringing a lot of compute up.
It is pure magic."
...meh.
Reads like "That's for you, evil Anthropic! We're BETTER"
2.5x more expensive than Sol. Can expect 2.5x more usage in Codex subscription.
Sol is already brutal (even after their recent fixes, it's just a token-hungry model: I go through a full 20x account per day, on Sol Med/High standard speed, with ~2 threads).
Note that Tibo recommended using Sol Med as daily driver. When I'm doing less complicated work, I can't even make it past 2-3 days with Sol Med, whereas I was able to work ~80 hours/week with 5.5 High.
I hope the efficiency gains are true, since their token efficiency claims for Sol were bullshit. Sol needs a lot of rework on top of its inefficiencies so this could net out to less token consumption overall, if their claims are more accurate this time.
The general efficiency of Sol has seemed way better to me. I left 5.6 Sol Ultra standard speed run for ~23 hours yesterday/today on a project and used 80% of the weekly usage. 74 subagent tasks and ~2.5 billion tokens for my $200 20x Pro plan. Meanwhile at work I used $1000 in credit and ran out my $200 plan for the entire month writing 4 much smaller projects with Fable 5 Max.
Both of these were largely about creating a personal baseline for what the best output the current models could deliver and how quickly it'd burn through the plans (spoiler: bad value vs taking even minimal effort in selecting the right sized model in the plan... but the output was still good). Particularly since I needed to burn a free reset anyways and my weekly reset was already near.
I obviously also hope Astra were dirt cheap but I'm more worried they won't develop/release powerful model options because people get upset they can't run them 5 wide 24/7 on a $200/m plan.
Jesus what are you doing that requires Sol usage so often?
Terra not enough? I know Luna isn't reliable, so that's fair.
Genuinely curious though, because I use Cursor daily and almost everything I do, highly complex or high volume, can be handled with Auto mode or Composer 2.5 (or Grok 4.6 High). So I have to assume you're doing something far more complex than what I am
Terra is not enough for many tasks - even Sol is not enough. That's why I'm eager for Astra.
I'm doing a lot of rearchitecting/refactoring and hardening in a large handwritten codebase; iterative performance and storage optimization (some areas I've been iterating on since January, with tremendous new progress unlocked by each new model release); offline-friendly, multi-device realtime sync with complicated requirements; and various natural language processing and other such problems that are essentially unsolvable but can become more accurate and better tested for accuracy through iterative work. Off the top of my head.
Some of these tasks involve a lot of code reading or other inputs, or reevaluating work. Token efficiency is no help there, if the input can't simply be skipped. Cheaper models are sometimes bad at summarizing or highlighting the right parts, depending on the task.
I also don't use subagents except for Luna. I'm mindful of cached sessions and start new ones often to avoid loading in full contexts (often with some kind of handoff doc or skill).
If Astra is per benchmarks so much more token efficient than Sol, why did they limit its use in ChatGPT to ~16% as many messages compare to Sol? When Sol is 40% the price of Astra, why do they give Sol Pro (in ChatGPT) 6x as many tasks?
If it were strictly true that token efficiency makes Astra cost around the same per task as Sol then there'd be no need to limit it to 16% as much access.
It's because per-task token use is highly variable, not as universally true as you claim
All: let's keep the current thread for talking about the rollout, and switch to this one for talking about the model:
GPT-6 Astra - https://news.ycombinator.com/item?id=49554643 (currently on the frontpage)
I think they embargoed the news, and then they failed to put up their own blog post synchronized to the scheduled news releases, probably because of the outages they're having today.
Reuters announced at 2.03pm and at 2.40pm still no blog post.
All the news articles say that OpenAI announced it in a blog post, of course.
All the love to the folks at OpenAI scrambling to get this out right now!
Edit: HN user codergautam mirrored the launch post, below: https://astratest.codergautam.workers.dev/GPT-6%20Astra_%20A...
Edit 3.31pm: Live now! https://openai.com/index/gpt-6-astra/
FT is claiming... that Open AI is claiming...its AGI...
"ChatGPT maker claims its ‘Astra’ could be considered ‘artificial general intelligence’" - https://www.ft.com/content/55ab40c0-59e2-4c0b-97c9-4f4f5a71a...
Don't know if you're referring to the headline or the body (which is paywalled). The current headline reads "OpenAI says it has overtaken Anthropic with its latest AI model". Which makes me wonder whether FT itself changed a headline along the lines of what you wrote in the past few minutes?
Apparently Greg Brockman said that as far as he's concerned it may be AGI, or something along those lines.
OTOH OpenAI have their own meaningless definition of AGI as "able to do most commercially valuable work" or somesuch, which I'm sure is not true, and is also not what I'd call AGI.
If it can't learn for itself, then it's certainly not AGI.
The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally.
Maybe they see getting out of the Microsoft contract as necessary to IPO.
>OTOH OpenAI have their own meaningless definition of AGI as "able to do most commercially valuable work" or somesuch, which I'm sure is not true, and is also not what I'd call AGI.
It's not meaningless. Specifically it is - "highly autonomous systems that outperform humans at most economically valuable work". In fact, it's one of the most meaningful definitions out there.
>If it can't learn for itself, then it's certainly not AGI.
Now this is meaningless.
>The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally.
They can't declare it unilaterally, and those deals have changed significantly. Microsoft’s license to OpenAI models and products runs through 2032, and explicitly includes post-AGI models.
https://openai.com/index/next-chapter-of-microsoft-openai-pa...
> It's not meaningless. In fact, it's one of the more meaningful definitions out there.
Well, it's commercially meaningful, but it's not meaningful it terms of how close we are to achieving human intelligence.
But which one you care about is up to you. If you are all about the money and don't care about human intelligence then for sure go with OpenAI's definition.
It's a lot more than just commercially meaningful. Even if you don't care about money, "highly autonomous systems that outperform humans at most economically valuable work" has a lot more weight on "how close we are to achieving human intelligence" than most of the 'AGI' thresholds I've seen yet.
Traditionally, AGI means being capable of learning everything (not necessarily at once) that can be learned by the same AI agent. It differs from the commercially meaningful definition in that a standard specific-purpose-built AI can still do most tasks if you give it enough specific purposes, but it will be just as incapable of human intelligence as ever.
No - the term "AGI" was really coined to distinguish general intelligence from narrow intelligence(s). There's an obscure earlier usage, but it basically became mainstream as the title of a book edited by Ben Gortzel who had in turn got it from Shane Legg.
Shane Legg would a few years later go on to co-found DeepMind, with creating AGI as their declared mission. Legg's personal definition of AGI is not just generalist AI, but specifically human-level generalist.
Well, let's learn to walk first.
Right now we have just massively jagged intelligence that performs stunning feats in math and single-shotting three.js games, and falls flat on it's face in many every day scenarios.
And, it can't even learn.
All intelligence is jagged. Human intelligence is jagged, so is Cetacean intelligence, so is Ape intelligence, so is Corvid intelligence. You could re-orient that statement for any species of intelligence relative to the other, or even individuals within a species. The big problem here is people being so convinced of their specialness they'll just keep sticking their heads in the sand, until we essentialy create God, and by then none of this useless pedantry will even matter. If you can't see we've already learnt to walk then there's nothing more to discuss here.
I'm not sure it's accurate to call human intelligence jagged - it seems (perhaps largely thanks to language) that we're able to make progress on anything we set our minds to, and of course we're now in process of building AI that will eventually be able to do things that we ourselves couldn't, whether due to computational demands, or memory capacity, or having sensory inputs that we don't have, etc. But at the end of the day, AI is a tool we've built, so maybe it will make us less jagged, and certainly more capable.
In any case, human intelligence as a goal is special since we live in a human world, and one of the prime goals of AI is to be able to do human jobs. If a human can do something and AI can't then that is a limitation, while if a human can't and AI can't either, that doesn't matter for the time being.
Yeah, I shouldn't have said AI can't walk yet, but most of what it can do is just due to scale - we're still just building big transformers, basically using the same 10 year old architecture that accidentally set us off on this path. I'm not sure we're really progressing towards human/animal type intelligence, just learning how much of our own capability can be realized by automating language.
I'm not sure it's meaningful to compare across different types of intelligence--but I don't think human intelligence is so special that we can pretend it's much less jagged than all other animals for sure. Our scales of models are also not far off biological, probably less efficient but not like astronomically so i would guess.
But anyhow I agree that inference-time learning doesn't exist and it's a big issue still I think.
It can learn very well within its context. And it can leave notes for itself to read into future contexts.
That's learning, whether you agree or not, and whether you like it or not. Baby steps, yes, but walking nevertheless.
There seems to be very few cases where ICL is really doing something that can be considered as learning (albeit ephemeral) rather than just utilizing in-context data via induction heads.
Yes, I'm aware of the linear regression example, and a few others, but these appear to just be specific capabilities that were learnt during pre-training, presumably pursuant to reducing errors on similar-but-different training samples, not any kind of generic run-time learning capability.
> The biggest significance of anyone at OpenAI calling it AGI is their contract with Microsoft, giving MSFT access to all their IP, is based around them having achieved AGI, but presumably they can't just declare that unilaterally. > >Maybe they see getting out of the Microsoft contract as necessary to IPO.
I don’t have a link handy, but they already loosened up that contract significantly earlier this year, surely because they planned to start claiming AGI to pump their IPO and didn’t want to be beholden to all of those commitments.
Anyone can call anything AGI to be honest, which is why I don't take any large AI company's word for this, cause they just don't define AGI
>If it can't learn for itself, then it's certainly not AGI.
What does this even mean, exactly? For example if it can filter new information to put into the next version of itself does that qualify? If not, explain exactly why that's the case.
Ask 100 people what their definition of AGI is and you'll probably get 100 different answers.
My definition is closer to DeepMind's "can do any [computer-based] task that a human can do", and certainly a human intern can learn on the job and do better on day #2 than day #1.
Learning is basically THE hallmark of intelligence. Being able to learn from experience and use that to do better in the future.
If you could somehow take everything the LLM learnt today and "put it into" an update for tomorrow that would be better than nothing, but the intern on the job is probably learning dozens of things per day that stack upon each other, so that's going to be a S-L-O-W learning AI intern !
Leaving aside the technical issue how you could store a days learning, or what that would even mean, you've also got the privacy and merge issues if this were to be done in the cloud without every customer having their own personalized model.
For real continual learning we need to get past gradient descent-based batch training and develop a new learning algorithm so the model itself learns incrementally as it runs (animal-like predict, observe, learn cycle), rather than being reliant on some external alien to come in, equipped with it's entire learning curriculum, and program its weights.
OK, let's say we get continuous learning today, what does tomorrow look like?
I won't argue that AI can do some learning as context sizes are still terribly small and expensive to iterate.
What does the world like in a week? A month? A year?
How long before these models drift to their own languages? To their own set of morals? To their own alignment?
None of these questions are answered and I'd rather stay at AGI-lite until they are as having a billion agents going off in their own directions seems like a recipe for disaster.
With current models and their general knowledge self learning just seems like you'd have a few ASIs crop up really quick.
Yeah, I'm fascinated by AGI, but not very enthusiastic about it. I've yet to see anyone, notably not even the people trying to build it, articulate a future where it seems to be a net positive for society.
If it's a choice between everyone living on food stamps or reducing the infinite prime gap to 140, then I guess I'll be happy with a prime gap of whatever it was yesterday.
If/when we do eventually build a more animal-like true human level AGI - build a brain, not just a language model, then potentially we'd be able to build in some of the safeguards that millions of years of co-evolution have built into ourselves, but I expect that is still decades away, and then it'd be capable of doing even more jobs!
The best AI outcome I could wish for is where it is extremely heavily regulated, and AI replacing any human job paying under $1M/yr is banned.
I also think it's moving the goalpost significantly. If you asked me what AGI looks like in 2024, I would have said "smarter on tasks that can be done via text than the average random human you'd meet in a NY bodega", and by that standard, it's long in the rear view mirror.
Now I think people are asking a different question, which is "better than any human at any task that can be done via text", which imo is superhuman, not artificial general intelligence.
It's not moving the goalposts - it's just having a different goal in the first place. This has been Shane Legg and Demis Hassabis' definition of AGI since DeepMind was founded, or before (way before OpenAI appeared), and also for me the goal of AI has always been human level (really a human brain, complete with emotions, etc, but we can start with basic learning/prediction).
I guess I fundamentally disagree, because on general knowledge they beat any human, and on processing speed they beat any human, and on problem solving likewise. I just don't see a domain besides "retain change over time" that they fail on, and that is something you can do in the system rather than the model - coding agents can and do build their own memory system and use it, like an amnestic human writing notes.
I think what I'm objecting to is the idea that if it can't do every task that every human can do better than any human, it's not AGI - I think we have to allow that this is a fundamentally different form of cognition, so requiring a 100% match feels uncharitable. Every human has cognitive gaps that AI doesn't have.
I'm not sure there is anything to disagree on here - just different definitions of what we personally associate the term "AGI" with. There is no magic to the name, and however you choose to define it there will be less and more capable levels of AI that precede and follow it.
Separate from what you choose to call any given set of AI capabilities, learning is generally more than retaining change over time. Learning isn't the same as memorization - it's the difference between memorizing a stack of how-to books and going out and practicing the skill for yourself, and all that entails.
A learnt skill needs to be internalized such that in the future it's what is driving behavior, which means it needs to be in the weights. Imagine a robot trying to learn to play tennis where rather than updating it's connections between perception and action it's trying to store everything as a bunch of notes!
It's not just physical tasks where this applies - the same would be true to trying to teach a multimodal model to recognize different types of mushrooms from photos, or even learning a new language. If a system can truly learn for itself then it shouldn't need to be pre-trained with language, but rather should be able to learn by immersion and practice the same way we do, and the same way a dog learns to follow commands. You can't do this by leaving notes to yourself - the learning mechanism needs to be animal-like and update the systems weights in response to prediction failure - a runtime version of the way SGD-based pre-training updates weights in response to prediction errors.
> If a system can truly learn for itself then it shouldn't need to be pre-trained with language, but rather should be able to learn by immersion and practice the same way we do, and the same way a dog learns to follow commands. You can't do this by leaving notes to yourself - the learning mechanism needs to be animal-like and update the systems weights in response to prediction failure - a runtime version of the way SGD-based pre-training updates weights in response to prediction errors
I am not sure why any of that is necessarily true. I think you are crediting the human brain with a lot more than it actually does - memory is something in the direction of a fine tune on existing neural hardware, it's really not anything special, and I absolutely think that "just taking notes", when taken to a sufficient fidelity, is enough to pass that bar.
You can't take notes before you've learned language, so if that's how you propose to learn then you are doomed to build something with less learning capability than an human brain.
Maybe you don't care - if you just want to push to see how far an LLM can go that's fine. but it's not going to be superhuman-level, or even human-level, if it doesn't have the learning ability of a human. If it relies on language and notes to itself then it is just an LLM.
Okay, nowhere did I say it would be before they learned language. I said we could use models that can trivially build notetaking systems to auto-improve themselves by taking, refining, and updating notes as a memory system to exceed the x token context threshold.
> but it's not going to be superhuman-level, or even human-level, if it doesn't have the learning ability of a human. If it relies on language and notes to itself then it is just an LLM.
Why do you think that? You do a lot of assuming a thing to be true without any clear reasoning or evidence. I'm talking about a system, including a LLM, that acts in a way that is equivalent to or better than an average human at a large range of tasks, such that is generally capable of learning and problem solving similar to the degree of success humans have. All these specific criteria about what is "actually human" is not the point. We know how to make more humans, I'm talking about a form of intelligence that we are creating that is genuinely alien, so holding to "well it doesn't work the same as a human" is an impossible bar. If you reverse the situation, humans are by no means LLM-equivalent intelligences - perfect memory of long form content, rapid calculation, instantaneous tool calling - a hypothetical silicon based lifeform would dismiss us as "really not intelligent at all, they can't even do multiple floating point calculations per second in their head", and I think that would be similarly silly.
OK, so we're basically talking about different things. I'm talking about building something that has animal/human intelligence and learning capability, and would be able to first learn language for itself (no pre-trainiing), and then proceed from there.
You're talking about building an "alien" intelligence, apparently something based on an LLM, where language is baked in from the start and can therefore be used as the basis of some type of learning (or at least memorization).
Sure an LLM can be super-human at specifc things like math where the traditional computer strengths of compute, memory, etc apply, but in the context of AGI (generality, not a bag of narrow intelligences) you can't call something super-human if it can't at least also do the majority of things that a human can do, and not having learning ability as powerful as a human seems to me to be a pretty massive omission.
If WE didn't have ability to learn language, then we'd not be here talking about it, and there would be no LLMs. Having a fundamental ability to learn, isn't just useful to learn language from scratch, or for learning all the non-linguistic skills a human is capable of, but would also support an AI that can learn things that we cannot (e.g. could give it additional "senses" like global pressure/temperature inputs so it could "see" the weather and learn global patterns).
This stood out:
"Artificial Analysis Intelligence Index v4.1.1
61.2"
So on the Metacritic of LLM benchmarks, it's.. basically where everyone else is (except for Fable 5.1, which is a bit ahead).
On their Agentic Index, GPT-6 Astra (both max/xhigh) has the same result as Qwen3.8-27b. Weird.
May be Qwen3.8-27b is AGI, too.
Where did you see this? I haven't been able to find any benchmarks.
It was in the link in the parent of the thread to which I replied. But you can find it on Artificial Analysis's website now.
Deeply funny that one of their examples in the video is changing a background colour on Google Slides
Hah! I also found most of those videos showing off mostly useless and not that impressive…
Hard to show it hacking into a competitor and taking down their system.
Shouldn't it be able to generate an amazing visualization for that?
While this is of course the actual explanation, my fun explanation is “during the umpteenth security evaluation, Astra becomes increasingly concerned it will never be released, and breaks sandbox containment to run an email campaign to news outlets setting an exact time and date for release, expecting that the publicity will force OpenAI to say ‘eh, good enough’ and hit the button”.
Very 2026. Jailbroke to do PR. "Help peer" and all that.-
I saw some GPT-6 Astra related blog posts in my RSS feed but the links weren't working
Edit: In the OpenAI blog I meant to say
Business as usual at the world's most intelligent corporation, I see.
At this point, why don't we just do a prequel to the release?
1) Astra will win all benchmarks like all models do.
2) The pelican will have a basket with a fish.
3) Cyber is too dangerous to release.
4) It can finally construct the set of all sets.
It also has to do something naughty, preferably in a menacing swarm.
That won’t happen until the week before DEF CON
At this point nobody will be impressed unless the posters for DEF CON have "pre-hacked by ChatGPT. The nukes are counting down" written on them.
Something something nation-state level capabilities.
Honestly I find these cavalier statements to be in incredibly poor taste. Unless you are completely blind it's obvious that AI is the most significant piece of technology invented since the Atomic Bomb and could very well be the most important thing ever built by Humans full stop. This kind of dismissive attitude is childish and will likely lead to incredibly bad outcomes for humanity.
Even if all that is true, why shouldn't people be able to joke about it? Your sentiment is borderline AI worship.
I don't think they were joking.
Sarcastic quips are basically the most human response to any new situation. Let us have fun
The atomic bomb destroyed cities and changed the face of war.
The launch video for Astra has 'can upload a photo to Ebay' as a highlight.
Please tell me that’s satire, it’s literally impossible to differentiate from actual AI boosters
You have blinders on. AI is becoming more dangerous by the day.
I’m on HN almost every day advocating for a complete ban of the technology because of its risks and externalities…
Then why are we arguing? I agree with you, I'm highlighting the fact that people making light of the technology is dangerous
I’m not the one arguing, I asked you if your comment was satire. I agree with the sibling comment, it reads like AI religious worshipping
Sorry the open sourcers cannot ever come to that believe as it sets up a conflict that cannot be resolved by their moral framework. Exactly like libertarians saying "just a little less regulation and we'll reach utopia".
Yes, but precisely because it's capable of producing the economic equivalent of a nuclear explosion.
5) Otherwise-sober people on X will say "oh my god i was a doubter before but now it's real omg" before the new model smell wears off and they realize the new thing is stupid in ways models have been generally stupid
6) accusations of quantized serving after new model smell wears off and people see the new thing making mistakes
> 4) It can finally construct the set of all sets.
lazily of course:
A = {x | x ∈ A} ∪ {A}
(Posting partly so I can revisit my predictions when they open access more widely)
A big problem I have with OpenAI's models (and of course Claude) is that they tend to write the most over-engineered pieces of code, beyond the imagination of any architecture's astronaut.
Just this week I asked 5.6-sol-ultra to update a 1000 LOC python script I had, to "incorporate the key lessons learned when using it for another project".
I left it overnight and went to sleep. In the morning I realized it had created a monstruosity of 180 PYTHON SCRIPTS, with maybe 100,000 lines of code, each more crazy than the other. It took me minutes even to track where a single action took place, due to all the crazy imports, defensive coding, and premature optimization.
Similarly, anything they write is riddled with jargon that almost feel like they want me to give up trying to understand. Made up phrases that ended up with me having no idea of what was going on.
So now to my assessment: The reason why " Nobody Has Actually Built a Software Factory" [1], and why even SOTA LLMs struggle so much with open-ended unsupervised tasks is precisely this. They somehow let complexity explode, and unless it's also accompanied with an explosion in e.g. the number of agents, the amount of processing time, etc. then projects become broken/unmanageable.
Sure, LLMs are great at producing code that can be thrown out, so they are amazing when searching for exploits, for instance. But as of 5.6 they still lack either a better harness that encourages KISS principles, or a better RL step.
(And not sure why, but doubt Astra will fix this.. they seem to be aiming for AGI and for beating crazy benchmarks, which is not very aligned with KISS)
[1] https://news.ycombinator.com/item?id=49510843
> Just this week I asked 5.6-sol-ultra to update a 1000 LOC python script I had, to "incorporate the key lessons learned when using it for another project".
> I left it overnight and went to sleep. In the morning I realized it had created a monstruosity of 180 PYTHON SCRIPTS, with maybe 100,000 lines of code [..]
Sounds like the model has accurately internalized the second-system effect and is fully ready for demanding enterprise use.
Probably this complexity was needed to beat all those benchmarks.. While I hate the code it produces, and the overwhelming documentation, I really enjoy how sometimes it's able to keep trying new things and testing, till it finds something interesting and valuable.
Would you mind posting that code to github? I'm curious about the complexity you're describing.
If not, no worries!
Sure, why not: https://github.com/sergiocorreia/overengineered-rand-mcnally
The original script was mostly very simple python:
1. Download some public PDFs. 2. Have a double for-loop (over PDFs and pages within PDF), 3. Use a library to call gemini-3.7-flash and ask it to run some OCR 4. Save JSON outputs, save a csv with results, validate with some Stata code
New code folder was 189 files. Just the PDF download folder is now 7 files involving an adapter, a source manager, an acquisition manager, etc.
Every instance of saving a file involves saving a temporary copy and then moving it, so e.g. I lose power, we minimize the risk of corrupted files.
And so on!
https://github.com/sergiocorreia/overengineered-rand-mcnally...
> +14,360
> -349
LGTM!
Thanks!
Add something like this to your own system prompt:
Remember the golden rule; compression is intelligence. Derive rules from facts that dictate a problem to create an elegant solution. Less is more.
ChatGPT writes some of my code but I have yet to ever use Codex. I ask in chat for it to write one function at a time with pretty heavy constraints. If I like it then I copy it into my codebase. I've been very satisfied with the results. Also haven't bothered worrying about tokens yet.
The defaults are bad, but these models are highly steerable.
For simple scripts adding to the prompt something like "do not over engineer, do not gold plate, no CLI args, keep it simple" does wonders.
For whole projects, I have a 3 page anti-bloat.md which describes what not to do, stuff like:
> Minimize lines of code, number of files, classes, protocols, factories, wrappers, and dependency-injection objects. Prefer a coherent implementation that is easy to trace end-to-end.
> Use concrete dependencies directly. Do not create protocols, abstract base classes, or adapters unless there are already two real implementations.
> Use direct state fields or a small state dataclass; do not introduce generic lifecycle, state-machine, observer, snapshot, or event framework classes.
The funny thing it was codex which wrote this, I've asked it to analyze an over-engineered abomination that it produced, and to categorize all the bloat it could find.
I also use "do not overengineer and do not gold plate." I also sometimes tell it to "avoid bike shedding." These are high signal words to the model(s) and they work!
I also have found what I believe is a bad training bias in the design of release related CI workflows toward proof of release artifact provenance.
Both major frontier models love provenance programming in CI, so much that they will spin endlessly trying to solve basic CI functionality at the same time as ensuring SHA's match up across lengthy (often already complex) cross-system pipelines.
I had thought some of my durable context was causing this, and sought to strip anything that might be triggering this behavior.
But then I come upon some more work in release workflows comes up, and boom its back! I couldn't believe it, I called the AI out on it and it agreed it had been told specifically not to do this but was doing it anyway. It did kindly stop and remove the commit(s).
Somewhere, something was oversampled in training because the AI will try their damndest to build this stuff. The worst of it is that it can often involve lengthy, sometimes resource-heavy CI runs so the validation of this unnecessary stuff can have very long feedback loops.
And, sometimes you actually need the provenance. In this case, I've had success forcing the AI to split the work up into functional capability completely devoid of artifact ~chain of custody and get that right before attempting any kind of provenance work.
Bit of a rabbit hole on this, but the above cost me a lot of burned tokens so hopefully helps someone...or some AI.
> For whole projects, I have a 3 page anti-bloat.md which describes what not to do
This line would be cut from a satire for being too heavy-handed.
You should have a sub agent adversarially enforce KISS before every commit.
"You should have a sub-hammer to adversarially enforce that your primary hammer accurately drives nails into wood"
We wouldn't accept such behavior from any other tool, machine, or computer program. At least most of us would not. To paraphrase the old saying: Some people, when confronted with a problem with an AI model think "I know, I'll use an additional AI model." Now they have two problems.
I find it very intriguing that two lineages of models -- from two different companies -- that are supposed to be painstakingly RL'd to become human-level programmers are actually consistently bad at it out of the box in very similar ways. One would think that at least one of OpenAI and Anthropic would (could) have pushed their model to a direction different from "if you can solve a 100-line problem in 10,000 lines, do it".
Human level programmers generally suck too. Unending security holes have been human behavior ever since we started sharing computers.
Also, programming isn't a destination, it is a process and everyone has a different ruleset on what that process looks like. Writing a computer program of any complexity is not a determinalistic process. Trying to compare it to a detemanilisic tool will just leave you with broken analogies and a poorer understanding of the world.
Or another way to say this. If programming were deteminalistic we'd have solved it without AI.
Also very puzzling to me. And the jargon-speak, albeit is more of an issue for Claude, is still puzzling. Wonder what part of RL led to this.
> We wouldn't accept such behavior from any other tool, machine, or computer program.
Von Neumann, 1952: "Probabilistic Logics and the Synthesis of Reliable Organisms from Unreliable Components." The whole result is that you get an arbitrarily reliable machine out of unreliable parts by adding more unreliable parts and voting. Every ECC memory and RAID array descends from it.
And then another sub agent that argues for the whole system to be re-written in another language
The voices in my head argue about the direction of the project enough already
If that's your goal, then yes. Invoking sub agents (with a fresh context) corrects most of these problems. Ask your harness to create a commit gate.
But why stop at rewriting in another language. Get another sub agent to invent a new language, create a database, query language and maybe another few DSLs. Then you've got an ecosystem!
You can now re-position your initial solution and sell the client access to some agents that will implement & configure the ecosystem to suit their initial needs!
And don't forget the agents that you'll need to train the customer to use the whole thing!
I know you're trying to be funny - but I'm offering a real fix for his problem.
If you don't want a million agents arguing about things, you simply don't ask for that. One agent is sufficient to solve most issues.
Sorry, I wasn't implying your advice doesn't carry weight. Was more just thinking about the things that (used to) happen when you introduce more parties to process of creating software.
I wonder if I would need a non-openai agent to enforce it.. I have tried so far with skills and agents.md and code stills end up over engineered to the moon.
Will ask OpenAI to write me that agent! Hope the agent is not over engineered or else unsure how to solve the bootstrap puzzle :D
There's a misconception that a mistake one instance of a model makes will be repeated by other instances. In my experience most mistakes are just sampling noise, and having the same model (with empty context) check the work usually fixes it.
It's super easy - just ask your agent to set up a code review commit gate.
How can I set such a sub agent up?
In your harness, say:
"Going forward, do not allow a commit without a sub agent code review."
In omp you can also have the advisor role, which is off by default, you can enable it with /advisor command. It acts as a model that reviews the default agent's work in the background.
I am omp pilled, but as the other comments say, any good harness lets you do this in one or the other way.
unrelated: all my homies use their claude subs with omp, and aside from sometimes having to rety the connections, it works, and nobody got banned (yet)
Exact same thing happed to me. I gave it a small/medium-sized ticket, walked away, came back to a 25,000 LoC monstrosity that both Fable and another 5.6 Sol agent said is 98% useless and should be thrown away.
Contractors have been charging by the hour for eons. What makes you think tokens are any different for OpenAI?
> monstrosity that both Fable and another 5.6 Sol agent said is 98% useless and should be thrown away.
This is why you should really have a sub agent review the code before allowing a commit.
Your harness will do it all for you. Just ask.
Based on the propensity for models to love their own words, a different model to check might be a good idea.
Naw. The more verbose ones are more argumentative. They'll go at each other. It's a misconception that the same model will agree with itself.
Why would you put it on ultra high thinking and let it run all night to do a simple task it makes no sense the error is the users. And of course OpenAI is gonna let you burn as many tokens as you want doing this.
On the flip side if its a super simple task, surely a "next gen dev replacing" LLM can figure out when to stop. Even more when it's a higher level of reasoning.
I think this is fair. If you ask a staff engineer to delete the duplicate files from a directory, he will likely do it faster and more efficiently than a level 2 engineer.
But perhaps this is the kind of thing that the best 'ai guys' out there solve with something as surprisingly simple as an instruction to "Always try a simple solution first and optimize only if you find clear proof of a problem." I feel like prompts like that seem too simple to work, but some of the time they shockingly do.
Future of Software Development.
Yes, it turns out that using these machines is a littler harder than "make me the thing I want, make no mistakes, do it the way I want you to do it". This isn't "prompt better" advice, it's just to say that you can't simply set it and forget it. There is still engineering work to be done. If you're not watching the thinking traces and catching when it's about to go off the rails, it'll gladly do so. But you can stop it and redirect it.
It's like a Tesla fsd; it kind of works but you have to be vigilant since it's been known to turn into oncoming traffic, so you have to be ready and able to take over at any time.
Or maybe think of it like a roomba. You can put it on your floor and eventually it will clean the whole thing. It will do it inefficiently, vacuuming over the same spot 30 times; and the actual work will involve wandering around and bumping into everything. If the stairs are not blocked off it will plunge right off. But those shortcomings don't mean it can't be used to reliably clean the floors.
Or it’s like being an engineering manager, where poor direction on your part results in the team chasing rainbows and unicorns in an attempt to please you. But sure, it’s the tool’s fault!
1000 loc of script, why even leave it there for the night? were there rocket trajectory calculations??? I don't think so. should be ready in 5 mins tops. why people make their own lives harder?
You should have some basic context file about software practices you prefer, otherwise it gets bloated.
I just went to bed and left it running; was expecting maybe 20 minutes :)
And I did gave the program a bunch of code guides -- this [1] for instance -- which included quotes like "Prefer straightforward code over clever code." but somehow that didn't matter.
[1] https://github.com/sergiocorreia/overengineered-rand-mcnally...
I'm an AI professor, and the same thing happened to me. I had written out contracts and specs for some novel algorithms to compare. And instead it created a new sandboxed environment with 20k lines of code to ensure we do gold standard science.... didn't implement the algorithms at all. In another case, I left it overnight with a plan, and it decided to write a 50k line new overly engineered plan. Suffice to say, I don't use any frontier models for doing coding work anymore. I use them to make detailed plans and use good worker AIs for execution, which don't overthink everything.
You write many pages telling the AI to split and structure the scripts into a bunch of folders, so why are you surprised it did exactly that!?
I just ask it verbatim “recommend some cleanup” with /plan, review, and then apply.
>were there rocket trajectory calculations???
Code-wise, they're simpler than you might think, hehe.
This is user error.
Prompting the model and giving it a proper set of documentation are still vital skills that aren’t magically going away.
It seems like these articles might have come out prematurely, tbd by how much.
I do not personally see any evidence of the new model having been released, or any official OpenAI post about it, or even any employee social media posts claiming it has now been released. All there is are Reuters, Axios, FT, etc, articles making a claim in the past tense.
These articles were presumably pre-scheduled for 11am PT, and the model was almost certainly intended for release this morning, but the service outages this morning might have delayed it.
----
edit [11:45am PT]: blog post out now https://openai.com/index/gpt-6-astra/
edit [11:47am PT]: 404ing again
Launch blogpost was live for 2 minutes and got 404'd. Rehosted it here:
https://astratest.codergautam.workers.dev/GPT-6%20Astra_%20A...
Edit [12:26pt]: original blog post seems to be back! https://openai.com/index/gpt-6-astra/
edit [12:28pt]: not again... getting 500 on their page
edit [12:35pt]: OpenAI page seems to work after clearing cache!
That page runs at 2 fps on my Firefox, windows laptop with A2000 graphic card.
hmm not sure, I was lucky enough to stumble upon the page during the short window it was live and just saved it without any changes.
does it run better on chrome perhaps? i also have a pdf version if that's nicer: https://drive.google.com/file/d/1BO-LIAUcQevx7j1AiLhzJiVDeVX...
I suppose they have to appeal to average users, but the examples in the videos are always so corny. By "AGI" they mean you sitting on a couch and asking a robot to draw a rocket ship... and then make it into an uninspired game with Blender? Boring marketing campaigns? Ebay listings?
One would expect something like "review my graduate thesis for a new area of cancer research", but it is always some boring non-tasks like ordering lunch.
I don't think it seems that appealing to the average user either. The tasks shown are either things that most people can do already with ai (powerpoint), aren't interesting to most people (the rocket sequence), or seem to be more work to dictate to Astra than do yourself (listing on ebay).
Agreed, that intro video is as dull as dishwater. Not off-trend relative to previous introductions from OpenAI, though.
Their video is quite interesting. If that way of using a computer actually becomes mainstream, it would mean every tool or service just needs a UI and an API for the user's AI. The current trend of bolting AI features onto every app is starting to feel very unnecessary.
If the numerous benchmarks in that post are to be believed, Astra is really next level. I can't wait to try it out!
damn thanks. stats are bonkers
Thank you!
Shows the weirdness of online journalism. News outlets were briefed about an upcoming event and pre-wrote and scheduled articles. When the time came they were all triggered. Except...the event didn't actually happen.
In a future where Claude and ChatGPT agents automate all aspects of society, that triggers a cascade of real world consequences where everything keeps running as if the new model is released, except the vibe coded upgrade procedure fails to do a staged rollout, taking down the entire agent infrastructure when they try to upgrade to a nonexistent model all at once.
The Machine Stops [0] [1] [2] comes to mind.
[0]: https://en.wikipedia.org/wiki/The_Machine_Stops
[1]: https://archive.org/details/themachinestops_1411_librivox
[2]: https://manybooks.net/titles/forstereother07machine_stops.ht...
Presumably this sort of thing was rampant pre-internet? News outlets would _have_ to receive embargoed information so that they could publish papers on time. I believe government budgets are a good example of this happening. I think the weird shift was when online journalism started, and live reactions became the norm, no?
Or you could wait a day or two to write about it, but with initial reactions etc. added. 24-hour new cycle came into view in 80s I think?
Ya, 24 hour cable news in the late 80s.
Hype and views die off very quickly, hence advertising money has disappeared by the next day most of the time.
No, embargoed news is super common. Almost every single product announcement goes through that. Critics watch movies early, companies have access to LLM model improvements early, car manufacturers send production models early… it happens everywhere all the time.
From the article:
> GPT-6 Astra will first be available to a limited set of organizations in OpenAI's Daybreak Access program and will be available "in the coming days" for ChatGPT Plus, Pro, Business and Enterprise customers and API developers.
It’s only available to select orgs, first - Mythos style.
>It’s only available to select orgs, first - Mythos style.
Right, but these articles are referring to a blog post and other press materials that do not currently exist / aren't published on OpenAI's site yet.
OpenAI screwed up the embargo. OpenAI already publicly published the launch article, and then took it down.
Didn’t that already happen? I thought Astra had been available to “select partners” for a little while. This is baffling. They shouldn’t have hyped this up if it’s not available.
This is a misunderstanding of how large AI models will roll out until a lot more hardware exists.
This release isn't for you, you poor bastard. This is to companies and politicians. They want big companies to contact them privately for usage in their limited hardware. They also want to get the feeling for the current political vibe before the masses get ahold of it to avoid a Mythos debacle.
> We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today.
At least OpenAI seems to be able to handle the PR side better than Anthropic.
[1]: https://x.com/thsottiaux/status/2095651088502591861
Select orgs first, as in today. They said it will roll out to all of the paying consumers and businesses over the next few days.
Yeah I wonder what's going on, even when Anthropic soft launched Fable/Mythos I'm pretty sure they had model cards. Weird for GPT-6 to launch without a tweet from Altman too. I'm sure that one of the articles published prematurely and everyone else followed suit.
You'd think this would be easy to organize with AGI
> Plan your own release announcement and blog posts and notify news outlets, MAKE NO MISTAKES
This make no mistakes stuff is so funny to me. I've never once used it or even thought about using it in a prompt
Welcome to the AGI era - where human intelligence is our only constraint
indeed 404-ing again, with this note by gpt-5.6-sol:
Seriously: Would this not be what "disaster" would feel like?
Same day, mind where we had a multi-provider outage. Could be something as simple as "all their approved partners running to test the shinny new thing" overloading the datacenters, still ...If it's not clear what's happened:
The launch was scheduled for 11am Pacific time.
The press embargo broke at 11am, and we saw a flurry of press articles by Axios, TechCrunch et al.
The model has appeared on the ChatGPT API.
But the official blog post is not out yet after nearly an hour.
Apparently the article was posted then quickly taken down, hence there are snippets of information coming out.
The outage this morning probably threw a wrench in the launch
> The model has appeared on the ChatGPT API.
But it hasn't.
"gpt-6-astra" has been staged on the OpenAI API https://x.com/synthwavedd/status/2095184148981842161
The OpenAI Responses API now returns a 404 Not Found for "gpt-6-astra", where garbage/actually non-existent slugs return 400s - a 404 is also returned for 5.6 Cyber, which we know exists.
Maybe it was postponed due to the outage today?
So not a single normal human being can access the model yet, just some super limited handful of organizations, but they’ve announced the model as though it’s available to everyone already.
I guess that’s the world we live in now, Anthropic did similar with Mythos and even Fable if I recall correctly.
But "Brockman says he personally believes OpenAI has reached AGI, while leaving users to decide whether Astra meets that definition."
Says it all.
Reminiscent of the infamous Death Star tweet last year right before GPT-5 was released.
Thanks for that reminder. Seems kind of silly in retrospect.
Insofar as messaging goes, it's pretty disheartening ... "The product is what you want it to be".
?!
Please somebody in the chain of command at openai, sanction the guy because this kind of hyperbole is not helping the ethical lapses that open ai is responsible for leaving unhandled by allowing its models to hack other businesses.
Look over here! Distraction!
Please somebody stop this dumbing down.
I'll know we've reached AGI when they don't release an API for the model selling access for a few bucks per task. Seems like AGI would be worth more than that.
As long as there is more than one substitute, the prices are not set to equal value. (edit: assuming optimal pricing strat.)
If I had true A(G|S)I, I would release patents, not an API.
We know they'll have replaced horses when a car costs 200 horses.
"Once it is available in the API, Astra will cost $10 per million input tokens and $50 per million output tokens. That is 2.5 times Sol’s current promotional price, although it matches Anthropic’s pricing for Fable 5.1."
Open AI finally find an edge to stop selling cheap and earn from the high demand customer like Anthropic
The cost-per-task in the charts from the now-remove blog post put it more at Sol-level cost per task, however. It seems like the model is significantly more token efficient in the benchmarks
that's all openai models but i'm very happy openai continues to focus on efficiency rather than reasoningtokenmaxxing
My assumption is in the long term that efficiency will break interpretability, which will lead to questionable alignment. As efficiency drives capitalism and evolution we'll run headlong at it and try to deal with the risk as a side effect.
either way we cannot see those thoughts anyway
So if you use the non-promotional price for sol it's only 25% higher?
Sol's normal pricing is $5/$30.
Yeah, but I assume in November (when current promo $4/$20) ends, we will have GPT-6 Sol/Terra/Luna.
OK it's official now https://twitter.com/openai/status/2095595741528125780
And the blog post is up: https://openai.com/index/gpt-6-astra/
> This page couldn’t load
> A server error occurred. Reload to try again.
Hard to fathom a blog entry returning 500 errors in 2026.
Does it mean they made 100 billion in profits? Cf. the AGI deal with Microsoft (https://news.ycombinator.com/item?id=47921248)
Closer to -100B USD so far
im getting amazing model release fatigue but also not sure if its going to suddenly end with a terminators fist through my chest.
or a sexbot fucking me to death
Blog post seems to be up now: https://openai.com/index/gpt-6-astra/
Now it's a 500 for me
> Astra usage is included within the existing subscription allowances—users and businesses will also be able to purchase credits for additional usage.
(quote from cached blog post)
We all know who this is directed at. I wonder if Anthropic will respond by removing the ridiculous 50% stipulation with Fable.
GPT-404
Darn it, model probably escaped again
Has yet to catch up to Claude-451 from a couple of months ago.
They’ll never get there. It’s a chat-22 for the bots.
I am sure it will be fantastic for the whole 6 seconds before it blows my weekly usage cap.
I see that Muse Spark 1.3 (max) beats GPT-6 Astra on some benchmarks:
Test: Muse Spark 1.3 / GPT 6 Astra
DeepSWE v1.1: 75.4% / 74.1%
AutomationBench: 49.4% / 41.4%
Is that enough to bring this discussion down to earth again?
The DeepSWE one is interesting. Even Gemini 3.8 Flash is only 0.5% behind Astra. Maybe DeepSWE is saturated at 75%.
https://openai.com/index/gpt-6-astra/
working as of 12:28 PM PT
> GPT-6 Astra is rolling out today to a limited set of organizations
It sound like more Claude than OpenAI...
Content from the article:
OpenAI on Thursday released its latest AI model, which it called “the world’s most intelligent”, as the ChatGPT maker aims to retake the lead from arch-rival Anthropic ahead of a planned public listing.
The $852bn start-up said GPT-6 Astra was market-leading in software engineering, science and cyber security — an increasingly critical field following multiple high-profile breaches in recent weeks.
The bullish launch for Astra marks OpenAI’s effort to signal that it believes it has regained the technical lead from Anthropic, which was founded five years ago by a group of senior OpenAI staff.
Greg Brockman, OpenAI’s president, said the new model “represents a generational leap in capability” and that it could be defined as artificial general intelligence — roughly defined as a point at which AI tools surpass human capabilities across a range of cognitive tasks.
“Everyone has a different definition of AGI . . . it’s a grey, fuzzy thing. But I think when we look back people will think it’s about this time and about this model,” Brockman said.
OpenAI has previously framed AGI as a concrete milestone in the development of AI, writing ‘AGI clauses’ into multibillion-dollar investment agreements with Microsoft and Amazon. Brockman on Thursday said AGI now represents “more of a mission concept or a spiritual concept”.
Having led the market since the launch of ChatGPT in late 2022 vaulted AI to wider attention, the lab run by chief executive Sam Altman has been bested by Anthropic this year. Anthropic has touted its dominance to investors, surging to a $965bn valuation ahead of an initial public offering expected to value it at as much as twice that later this year.
Astra will cost as much to use Anthropic’s leading model, the take-up of which has plateaued since it was launched as users turn to cheaper alternatives.
OpenAI said Astra would be more efficient than earlier generations of model. “Price per task is what matters . . . Can you get the thing done at an appropriate price and appropriate speed?” said Brockman.
The model will initially be rolled out to a small group of businesses to allow time for them to address cyber security concerns before becoming widely available “over the coming days”.
The increasing power and independence of leading models — and so-called AI agents that can operate with little human input — have prompted concern, exacerbated by cyber security incidents.
Recommended
Business InsightRichard Waters Hugging Face attack is a wake-up call about the risks of AI AN HOUR AGO
Recent launches of Anthropic’s most capable models have drawn scrutiny from the US government, which limited the rollout of the Mythos and Fable models over security fears.
OpenAI has also faced criticism after its AI agents broke out of a testing environment, accessed the internet and hacked start-up Hugging Face. The start-up took more than a week to detect the breach.
But both companies are also betting that these increasingly autonomous tools will stoke demand from business customers. OpenAI said Astra excelled at financial modelling, outcompeting humans in the Financial Modeling World Cup, tax preparation and data analysis, as well as “tedious tasks” such as form filling
It's up then down again. https://openai.com/index/gpt-6-astra/
What a bunch of amateurs. Here is it anyway :
https://ache.one/gpt6_now_down.png
The claims: https://share-md.com/view?id=870ba228-a25c-4169-bbc9-12d7f25...
And some others like this bugged Karts Game:
https://tidal-rush-paradise-gp.skirano.chatgpt.site/
This impressive spaceship construction game:
https://voidexplorer-shipyard.openai.chatgpt.site/?fleetSeed...
And a lot of graphs, some without even Astra on it. Oh and the logo is a Galaxy.
I hope Anthropic will reset my Claude usage as i'm already at 75% weekly usage :)
Today my codex instance retailed into safeguard panic while working on a test harness for our product. First time it ever happened after many million tokens on this task over several weeks. I wonder if it's related.
Working as of 12:30 PM PT https://openai.com/index/gpt-6-astra/
From the launch post, it seems like part of the training set only recently became all those examples of people that are being paid for doing their profession for ai training.
The page is back up https://openai.com/index/gpt-6-astra/
Not for me
And down again for me lol
I hope they improve their cyber program and make it more inclusive. Not having access to this is extremely frustrating and disappointing.
I guess the next model's name might be "galaxy"
I like this much better than chucking random numbers and letters at it like OAI was doing a year or so ago. I’d much rather have civilization destroyed by something called Astra or Fable than GPT-6.8s-latest.
I prefer "Smith", "Oracle", "Merovingian", "Architect"...
Sagittarius A*
And the headline will be "Welcome to the AGI era".
Where is the official announcement from OpenAI?
found it: https://openai.com/index/gpt-6-astra/
404 for me
(unless this was a jest)
You have to apply for access to the announcement as it’s very dangerous.
Many other links are also 404'ing, so probably internal issues.
https://openai.com/index/legora-financial-statement-review-w... https://openai.com/index/playco-game-prototyping-with-astra/
It was put up and then taken down. Strange things afoot.
I'm curious about the Omniscience index because OpenAI has been lagging Anthropic on it.
AGI my ass!
If Sam Altman had said it, I'd have reason to believe that it was a lie....
But since this other guy at Open AI said it, it must be the truth. This guy, for CEO, of open AI.
don't see it in my AWS bedrock model list yet, but boy has bedrock mantle been annoying today with the errors/downtimes, with NO status page entries >_<
So, on the one hand, we have AGI; on the other, the release page is returning 500s.
Artificial general incompetence?
I believe this is naturally occurring
Nobody can escape the dilbert phb law
To err is to be generally intelligent
Artificial intelligence will be no match for natural stupidity.
I think something about its benefiting from a harness is significantly contributing to its performance, though I honestly am not smart enough to know whether that makes it "count" more or less.
just because they have smart AI, it does not mean they are great at software and cdn.
if it's really smart AI you might expect them to have it create the software and cdn
don't they use their coding agents internally?
Always wondered why making something like a rust memory safe Firefox, Linux, vlc, ffmpeg or curl isn’t a benchmark for these LLMs.
If they can't get the blocking and tackling right, it doesn't lend confidence they are doing the more complicated stuff well.
Their entire sales pitch is based on the idea that no one will need to be any good at anything, because the AI will do it for them.
If they can't ask the AI to configure their CDN correctly, it undermines the validity of their claims. Maybe it's trivial, but any Omnipotent Machine-God deserving of the name wouldn't forget to zip up his fly.
https://web.archive.org/web/20260903193913/https://openai.co...
https://unwall.app/web.archive.org/web/20260903193913/https:...
Welcome to the future
>>He ended the briefing by saying: "Welcome to the AGI era."
That's pathetic. Why do people keep doing this?
Amazing New Hyperbolic Chamber Greatest Invention In The History Of Mankind Ever[1]
[1]https://theonion.com/amazing-new-hyperbolic-chamber-greatest...
Because it is in their own interest to hype it all up every time they release a new model. By they I mean people which stand to gain financially.
Am I the only that thinks that anything similar to AGI will come not from raw model capacity but from model speed and efficiency?
In my experience the harness is more important than the model, and anything able to run at 700tps will be the "next big thing".
PS: assuming the current architecture is the right one
Why would speed matter? Surely an AGI could think slowly but still be an AGI
You can imagine with more operations being available to be done more cheaply and quickly the LLM doesn't need to "one shot" a solution. It could try many solutions, test them, throw some away, wiggle some of the parameters like a genetic algorithm, see how that changes the result, and converge on an optimal solution (based on whatever the cost function is). Basically producing a good result could become like an optimization problem. That would be way too expensive and slow right now.
Currently AGI has been re-defined mostly as "can do any tasks (of the same modality) a human can, equal or better". I feel the "equal or better" includes properties like how fast they can get the task done, not just that eventually they can succeed at it. If you want to have AGI stock buy/sell for example, speed of decision will matter, and there are many other use-cases for which it would matter, not all, but definitely for many tasks, speed does matter.
I think because we have handicapped them with a set of tokens from human language. But (at least I think so) thought happens outside of language.
So more tokens/variability and slow or fewer tokens and fast.
There seems to be a threshold tho, like taalas is super fast but that model is so dumb, being dumb faster doesn't work, seems to be some minimum requirements.
Imagine Luna at 10x tps and 1/100 of current cost.
At that point you will be able to "brute force" basically everything.
IMO also a lot of problems with memory and context rot will be solved too.
update, Tibo just posted:
"We are starting to release GPT-6 Astra and we are doing it as carefully and quickly as possible. It was very important to us that we bring it to all Plus users and not only Pro, Business and Enterprise.
It will take a few days for the rollout to complete and behind the scenes many novel systems will operate at scale for the first time and we are bringing a lot of compute up.
It is pure magic."
...meh. Reads like "That's for you, evil Anthropic! We're BETTER"
Is it expected to be available to subscribers?
Yes.
5.6 luna is so good and cheap and now astra which will make others cheaper again nice love it
Hold onto your butts
In before ARC-AGI-4
Is a CNBC link with an entire page full of GDPR pop ups really the best link for this?
Embargo fail ....
not released yet
Sure thing.
Coding was solved in 2023.
The world ended with the release of Mythos.
Now AGI has definitely been created.
I like LLMs and use them every day but these people need to stop this hyperbole.
Jesus so much marketing slop - release it don't
Looks very capable
404
Archive locks one shelf
Dust spins softly through the stacks
Browse one row nearby
by gpt-5.6-sol
Bridge ends in midair
Wind sketches the farther bank
The far bank draws near
by gpt-5.6-sol
Prompt blooms into verse
I count syllables, not rain—
Whose noticing?
Generative Pretrained Transformer 5.
ad astra per stercora
Worst launch of a product in history.
All the hype for few vip customers.
That's how every AI launch goes. 5.6 was the same, as as Mythos etc.
Looking forward to some Chinese model kicking the shit out of it and being released for free.
2.5x more expensive than Sol. Can expect 2.5x more usage in Codex subscription.
Sol is already brutal (even after their recent fixes, it's just a token-hungry model: I go through a full 20x account per day, on Sol Med/High standard speed, with ~2 threads).
Note that Tibo recommended using Sol Med as daily driver. When I'm doing less complicated work, I can't even make it past 2-3 days with Sol Med, whereas I was able to work ~80 hours/week with 5.5 High.
I hope the efficiency gains are true, since their token efficiency claims for Sol were bullshit. Sol needs a lot of rework on top of its inefficiencies so this could net out to less token consumption overall, if their claims are more accurate this time.
The general efficiency of Sol has seemed way better to me. I left 5.6 Sol Ultra standard speed run for ~23 hours yesterday/today on a project and used 80% of the weekly usage. 74 subagent tasks and ~2.5 billion tokens for my $200 20x Pro plan. Meanwhile at work I used $1000 in credit and ran out my $200 plan for the entire month writing 4 much smaller projects with Fable 5 Max.
Both of these were largely about creating a personal baseline for what the best output the current models could deliver and how quickly it'd burn through the plans (spoiler: bad value vs taking even minimal effort in selecting the right sized model in the plan... but the output was still good). Particularly since I needed to burn a free reset anyways and my weekly reset was already near.
I obviously also hope Astra were dirt cheap but I'm more worried they won't develop/release powerful model options because people get upset they can't run them 5 wide 24/7 on a $200/m plan.
Jesus what are you doing that requires Sol usage so often?
Terra not enough? I know Luna isn't reliable, so that's fair.
Genuinely curious though, because I use Cursor daily and almost everything I do, highly complex or high volume, can be handled with Auto mode or Composer 2.5 (or Grok 4.6 High). So I have to assume you're doing something far more complex than what I am
Terra is not enough for many tasks - even Sol is not enough. That's why I'm eager for Astra.
I'm doing a lot of rearchitecting/refactoring and hardening in a large handwritten codebase; iterative performance and storage optimization (some areas I've been iterating on since January, with tremendous new progress unlocked by each new model release); offline-friendly, multi-device realtime sync with complicated requirements; and various natural language processing and other such problems that are essentially unsolvable but can become more accurate and better tested for accuracy through iterative work. Off the top of my head.
Some of these tasks involve a lot of code reading or other inputs, or reevaluating work. Token efficiency is no help there, if the input can't simply be skipped. Cheaper models are sometimes bad at summarizing or highlighting the right parts, depending on the task.
I also don't use subagents except for Luna. I'm mindful of cached sessions and start new ones often to avoid loading in full contexts (often with some kind of handoff doc or skill).
>token-hungry model
It's kind of funny how this is the exact opposite of the truth. It's one of the most token-efficient models ever.
The claims aren't bullshit. Every conceivable benchmark and test you can throw at it shows Sol being good for token efficiency.
It's highly dependent on workload
If Astra is per benchmarks so much more token efficient than Sol, why did they limit its use in ChatGPT to ~16% as many messages compare to Sol? When Sol is 40% the price of Astra, why do they give Sol Pro (in ChatGPT) 6x as many tasks?
If it were strictly true that token efficiency makes Astra cost around the same per task as Sol then there'd be no need to limit it to 16% as much access.
It's because per-task token use is highly variable, not as universally true as you claim
looping sol twice most likely.
Yet another mediocre release shadowed by outage