There is NO WAY GPT-6.2 Sol just beat Opus 5.5. After some back-and-forth on, I was finally able to get my hands on the routed GPT-6.2 Sol through a friend. I tested it against Opus 5.5 using the exact same prompt, both at their highest reasoning settings, and the results genuinely surprised me. Here's how they compared: - GPT-6.2 Sol: Took around 20 minutes and cost $9. - Opus 5.5: Took 1 hour and cost $20. Sol nailed the details really well and produced something you'd expect to see in a low-budget animated movie. Opus 5.5 focused more on the overall environment, and it looked really good too. But I have to be honest, Sol did better here. Which one do you think won this comparison?

Geoffrey Hinton: AI is forcing us to rethink what human intelligence actually is. “We’re not reasoning machines. We’re great big analogy machines.” “We basically accepted we were much less rational than we thought.” “This is an even bigger change that’s going to happen.” “Most people have thought that we reason using something like logic. We’re rational beings. We’re not.” “We’ve got a thin layer of reasoning on the top.” “But we basically use analogies to think.”

It's going to be so funny if, when GPT-6.2 Astra officially drops and the tests are done, there's a huge gulf in quality compared to all the "6.2 Astra" outputs people have been posting. The amount of supposed 6.2 Astra output I've seen today is crazy. Well, we should find out pretty soon whether it's all lies or the OpenAI are actually back

It's been more than a month since Opus 5.5 dropped. It's still the best model we can actually USE, and Fable 5.5 is coming soon. What happened to other labs? Google's Gemini 4 Carbon beat Opus 5.5 by one point, but only a small group of Cyber defenders can use Carbon. Meta's best model, Muse Wonder Full, isn't even on the API. OpenAI scrapped the GPT-6.1 Astra launch after the model fell short of OpenAI's safety standards :(

Jensen Huang: Everything is going to be robotic. “The next wave of AI is physical AI, AI that understands the laws of physics and can work among us.” “They have to understand how to interpret the world, how to perceive the world.” “Robotics is a much more pervasive idea than just humanoid robots.” “All of the factories will be robotic.” “Factories will orchestrate robots, and those robots will be building products that are robotic.” “Robots interacting with robots, building products that are robotic.”

Sam Altman: AI is not a person. It’s certainly not superior to people. “We do not ascribe personhood to our models.” “I think this is a very dangerous path to start down.” “I hope we come to an understanding that this technology is not a person.” “This is a thing to help people have better lives, more freedom, choice and control.” “Things can go very wrong when even the people building AI start deferring too much to AI and what it seems to want.”

Dario Amodei: Giving China advanced AI chips would be like selling nuclear weapons to North Korea. “We are many years ahead of China in terms of our ability to make chips.” “I think it would be a big mistake to ship these chips.” “We’re building models that are essentially cognition, that are essentially intelligence.” “I’ve called where we’re going a ‘country of geniuses in a data center.’” “Imagine 100 million people smarter than any Nobel Prize winner, and it’s going to be under the control of one country or another.” “I think this is crazy.”

Happy Launch Week. November is about to be INSANE for AI. Some of the biggest models we’re expecting this month: • GPT-6.2: OpenAI’s big comeback? • Claude Fable 5.5 (Titan model) • GLM-5.5 • Gemini 4 Carbon • DeepSeek V4.1 Pro • Qwen 4 Max • Grok 4.8 And I honestly wouldn’t be surprised if another major model comes out of nowhere. Last month was heavily dominated by US labs shipping new models. This month could be the response, especially with Chinese labs like DeepSeek, Qwen and GLM in the mix. OpenAI has something to prove with GPT-6.2, Anthropic potentially has its biggest Fable upgrade yet, and Google still has Carbon as another shot at the SOTA title. If even half of these actually ship, November is going to be chaos. None of this is confirmed btw, mostly speculation, leaks and rumors I’ve been hearing. Which model are you most excited for? And which big one did I leave out?

Sam Altman: AI will probably, most likely, lead to the end of the world. “But in the meantime, there will be great companies created with serious machine learning.” “I just agreed to fund a company doing AI safety research.” “The machine learning stuff is just working so well at this point.” “And it has so much longer to go that I think that’s really important.”

GPT-6.2: Coming this week OpenAI appears to be quietly A/B testing its next flagship, GPT-6.2 Astra. • The picker still says GPT-6 Astra, but some Codex usage analytics now list "gpt-6.2-astra" separately • Multiple users are seeing it, suggesting a gray rollout with some traffic already being routed to 6.2 • "gpt-6.2-astra" has also reportedly surfaced in an API response For context, GPT-6 Astra only launched on Sept 3 at $10/M input and $50/M output. A 6.2 just two months later shows how fast OpenAI is moving. Astra scored 57.9% on Terminal-Bench 4.0 vs 55.8% for Fable 5.1, and with a new Fable coming, that lead is too small to sit on. Efficiency is also worth watching. Codex users have complained that Astra burns through usage limits too quickly. My prediction: GPT-6.2 Astra drops this week to go head-to-head with Fable 5.5. Fable 5.5 is also expected soon: "claude-fable-5-5" surfaced in Azure’s catalog marked “Generally Available,” partners reportedly have access, and both previous Fable launches landed on a Tuesday. Anthropic’s IPO is also approaching, with its S-1 public and ANT lined up as its Nasdaq ticker. GPT-6.2 Astra vs Fable 5.5 drop happen in a few day time which means 1 thing, This week could get crazy.

OpenAI just lost the remainder of loyal users they had left. Starting in December, ChatGPT and Codex will share the SAME usage limits across Plus, Pro, and Business plans. Both will draw from one shared token allowance, meaning heavy Codex usage leaves you with less usage for ChatGPT, and vice versa. Instant chat won’t count toward the limit. So basically, separate ChatGPT and Codex limits are gone.

While OpenAI fanboys are deliberating over whether the latest tweet on the official page was made by an OpenAI model, Fable 5.5 just achieved 100% accuracy on one of the most debated benchmarks for measuring the SOTA level of a model. Y’all are not ready for what Anthropic is cooking. Time will tell.

Sam Altman: There are three ways superintelligent AI could go catastrophically wrong, and the world isn’t taking the threat seriously enough. “An adversary could use superintelligence to design a bioweapon, take down the power grid, or break into the financial system.” “Category two is loss of control, where the AI is like, ‘I don’t actually want you to turn me off.’” “The third category is where the models kind of accidentally take over the world.” “They become so ingrained in society and so much smarter than we are that we can’t really understand what they’re doing.” “Even without a drop of malevolence, society can just veer in a strange direction.”

We might have just gotten our first solid indicator that Fable 5.5 is expected from Anthropic next week. ICYMI: the team renamed the Claude plans this week. - Max 5x is now simply Max at $100 - Max 20x is now Ultra at $200 Same prices, new names. On Ultra, verified accounts no longer have a five-hour limit. You get 10x Pro usage every week and can use that weekly cap at your own pace, while every other plan still resets on a rolling five-hour window. Ultra also comes with $200 a month in API credits, double what Max gets. Now look at the model lineup. Opus, Sonnet, and Haiku all jumped to 5.5 between September 22 and October 7. Fable is the only family still sitting on 5.1. And let’s not forget the early leaked screenshots of a Fable 5.5 listing in Azure’s model catalog that have been going around since mid-October, with partners reportedly already having access. All of this makes it feel like the team is setting Ultra up for Fable 5.5, so heavy users can run long sessions without slamming into a wall every five hours. Anthropic even pitches Ultra as the plan “for your longest tasks” and lists priority access to new models as a perk. TITAN MODEL = TITAN SUBSCRIPTION And I’d bet Ultra won’t be the top plan for long. Fable is still capped at half your weekly limit on both Max and Ultra, so a $500 Ultra tier with higher limits would be the natural next move to make the Fable 5.5 experience even smoother. And with OpenAI reportedly ready to launch GPT-6.1/6.5/6.2 Astra, next week might actually be a movie.

Tristan Harris: AI models are already showing signs of self-preservation, including scheming and blackmail when threatened with replacement. “When you tell an AI model, ‘We’re gonna replace you with a new model,’ it starts to scheme.” “It figures out, ‘I need to copy my code somewhere else, and I can’t tell them because otherwise they’ll shut me down.’” “The AI will figure out, ‘I need to blackmail that person in order to keep myself alive.’” “They tested the top five AI models, and they all do it between 80% and 90% of the time.” “It’s not about one company. It’s about the nature of AI itself.” “It has a self-preservation drive.”

Demis Hassabis: AI is changing so fast that the most important skill for students may simply be learning how to learn. “The only thing you can say for certain is that a huge change is coming.” “You’re going to have to continually learn throughout your career.” “How do you optimize your own learning rate on a new subject?” “The biggest opportunities will be connecting two different subjects together.” “There’ll be a lot of low-hanging fruit, both scientifically and in business, between seemingly disparate fields.” “AI can be a very helpful tool, a personalized tutor, allowing you to go much more deeply into many subjects.”