My recent videos

PSA: Claude Fable 5.2 is not cheaper than Opus 5.5.

It costs $8 per million input tokens and $40 per million output, while Opus 5.5 is $4 and $20, so Fable is exactly double. It's only cheaper than the last Fable, 5.1, which was $10 and $50. Fast mode is $16 and $80. Anthropic does claim that on one coding test, Fable 5.2 matches Opus 5.5 at max effort for about half the cost per task, so test it on your own work before switching. Are you sticking with Opus or switching? Save this video and follow for more AI news.

This could be AI's biggest week of the year.

Theo from t3.gg says a big week is coming, and four launches are rumored. Claude Fable 5.2 already dropped today. GPT-6.2 Sol hasn't been announced, but testers say they're being routed to it. A new Gemini model rumored as "Krypton" is still leaks only. And Alibaba has confirmed Qwen 4 is in training, with prediction markets giving a release before December a 93.5% chance. If they all ship, the leaderboard could flip by Friday. Which launch are you most excited for? Save this video and follow for more AI news.

This could be AI's biggest week of the year.

Theo from t3.gg says a big week is coming, and four launches are rumored. Claude Fable 5.2 already dropped today. GPT-6.2 Sol hasn't been announced, but testers say they're being routed to it. A new Gemini model rumored as "Krypton" is still leaks only. And Alibaba has confirmed Qwen 4 is in training, with prediction markets giving a release before December a 93.5% chance. If they all ship, the leaderboard could flip by Friday. Which launch are you most excited for? Save this video and follow for more AI news.

Anthropic just dropped Claude Fable 5.2, and it beats Opus 5.5, GPT-6 Astra, and Gemini 4 Carbon on almost every benchmark Anthropic tested.

It scored 49.1% on FrontierCode 2.0 for agentic coding, ahead of Opus 5.5 at 45.5% and Astra at 42.8%. A team of 1,000 Fable agents turned the proof of Fermat's Last Theorem into computer-checkable code in 30 hours. It's also 20% cheaper at $8/$40 per million tokens and up to 30% faster. It's live on all paid Claude plans, and Pro users get a week to try it plus a bonus reset. Are you switching? Save this video and follow for more AI news.

GPT-6.2 Sol isn't officially out, but people say they're already testing it.

One tester turned Hokusai's The Great Wave into a 3D scene you can fly through. He calls it a massive leap over GPT-6.1 Sol, says it beats GPT-6 Astra in Max mode, and prefers its design taste to Claude Opus 5.5. He also ran dozens of 3D and HTML tests on about 10% of his $200 Pro allowance. Another tester says it knows events from June 2026, a sign it's a newer model. OpenAI hasn't confirmed anything yet. Would you switch from Claude for this? Save this video and follow for more AI news.

People keep saying we'll never get a model better than Opus 5.5, but better models already exist.

You just can't use them. Gemini 4 Carbon is on Artificial Analysis, and only vetted cyber defenders have access. Claude Mythos is limited to about 150 vetted organizations, while the public gets Fable, the same model with extra safeguards. GPT-6 Cyber is limited to select security customers. Labs say these models are too good at hacking for open release, and the government can now review frontier models before launch. Should the best AI be locked to a few companies? Save this post and follow for more AI news.

It’s really not that hard.

Do what works and then if it works for you, great. If it doesn’t, ditch it.

Here's what you actually get with ChatGPT's $500-a-month Pro plan.

It launched at 25 times the usage of Plus, and OpenAI just doubled all Pro limits on October 29. You also get Ultrafast mode, up to 8x faster responses on GPT-6 Astra and GPT-6.1 Sol in Codex and ChatGPT Work, which Pro 100 and Pro 200 don't get. Plus everything in Pro, like GPT-6 Pro with parallel agents and dots. It's built for developers and power users who hit limits every day. Would you pay $500 a month for AI? Save this post and follow for more AI news.

OpenAI just launched GPT-6 Cyber, a new model built for security research.

OpenAI says it pairs stronger capabilities with better alignment, and it powered the Linux kernel security research OpenAI announced yesterday. Security teams at Mozilla, Palo Alto Networks, and Cisco are already using it, but it's only available to select vetted teams in OpenAI's Daybreak Red program for now. That's the cyber model from The Information's scoop, with new Sol and Astra models still rumored. Should AI this good at hacking be locked down? Save this post and follow for more AI news.

ChatGPT can now make GIFs.

OpenAI just launched ChatGPT Images 2.6, which turns any prompt or picture into an animated GIF. You can also save reference images of yourself, your pets, or your characters and reuse them anytime without uploading them again. It's live today for everyone on every tier, on web, desktop, and mobile. That's one of the three leaked OpenAI models confirmed, with two to go. What's the first GIF you're making? Save this video and follow for more AI news.

ChatGPT family plans could be the fix for usage limits.

Leaked code shows family plans from $8 to $500 a month, and right now power users juggle multiple accounts just to avoid hitting their limits. OpenAI even just added multi-account switching on mobile. A family plan could mean one bill with several accounts, or a bigger shared usage pool, so if you hit your limit, you borrow from the household instead of paying for a second account. OpenAI hasn't confirmed how it'll work yet. Would this solve your usage problems? Save this post and follow for more AI news. #

Is the Anthropic IPO canceled?

A viral post claims the biggest IPO in history just got pulled, and tech YouTuber Theo says he's heard similar news from a few sources. But X slapped on a community note: Anthropic filed its S-1 on October 26 and an amendment on November 2, with plans to list on Nasdaq under the ticker ANT. The Wall Street Journal reported a target of the week of November 16, and Anthropic hasn't said anything about canceling. Rumor or real? Save this post and follow for more AI news.

GPT-6.2 Astra could drop this Friday.

The Information's Stephanie Palazzolo just reported new Sol and Astra models are coming, and leakers are calling them GPT-6.2 Astra and GPT-6.2 Sol, possibly as early as this Friday. This could finally be OpenAI's answer to Claude Opus 5.5 and Gemini 4 Carbon, and it's not one model, it's a whole combo. Still holding your two banked resets? This is what you saved them for. What's the first thing you'll test? Save this post and follow for more AI news.

Don’t act like we didn’t notice.

OpenAI may be about to drop three new models.

Three unreleased models have reportedly shown up in A/B tests and on Arena: a new image model, a small and highly efficient model, and a larger, much more powerful, more expensive model. They're reportedly labeled release candidates, so OpenAI hasn't decided which ones ship or when. And Tibo just posted, "Things will change again and again and again." Which one are you most excited for? Save this post and follow for more AI news.

Replying to @Randomize relax

Are we cooked or Nah?

Don't use your banked resets yet.

OpenAI's 28 days ended with two banked resets saved to your account, and you choose when to use them. GPT-6.2 Astra is rumored to be coming soon, and a new flagship model will burn through your usage fast. So hold both resets for launch day and go all out. Are you saving yours or already used them? Save this post and follow for more AI news.

Your ChatGPT dot is getting a brain upgrade.

OpenAI's Tibo says they're working on expanding the models that power dots, including GPT-6 Astra and GPT-6.1 Terra, with more coming. Your dot will automatically pick the best model for each part of a task. Terra brings more speed and a personality upgrade, and Tibo says it'll help bring dots to more people. No release date yet. Would you trust your AI to pick its own model? Save this post and follow for more AI news.

Hackers just cooked Europe's supermarket freezers.

Overnight, they broke into the refrigeration systems of supermarkets in seven countries, including Belgium, Sweden, Ireland, and the Netherlands, and cranked freezers up to 50 degrees Celsius, that's 122 Fahrenheit. "Everything is destroyed," one Swedish store manager told Reuters. About 40 Delhaize stores were hit, plus Carrefour and Spar, and it's the second attack in a month after France. Nobody knows who's behind it. Should your fridge be online? Save this post and follow for more tech news.

Is GPT-6.2 Astra coming?

Rumors say OpenAI's next model is one of the most powerful it's built internally, able to solve math problems the world's greatest minds struggle with. Meanwhile, OpenAI just launched LifeVal, a new benchmark built from thousands of ChatGPT users' feedback on the everyday tasks they want help with, because OpenAI says AGI has to help with real life to benefit all of humanity. AI that solves math, or AI that runs your day? Save this post and follow for more AI news.

Is Claude secretly testing Fable 5.5?

This viral animated forest wasn't made by a video AI. It's pure code, and the creator says Claude made it in one shot, claiming Opus 5.5 is quietly routing to a smarter unreleased model. Theories say Fable 5.5 or Opus 5.6, and a Mythos 5.5 reference reportedly just showed up on Anthropic's website. But Anthropic hasn't confirmed any of it, and secret-routing claims have been made before without panning out. Leak or hype? Save this post and follow for more AI news.

OpenAI's 28 days of shipping are officially over, and Day 28 ended with one last banked reset, no surprise model.

The final tally: 14 usage resets and 17 feature drops. The best: GPT-6.1 Terra, doubled limits on every Pro plan, and texting your dot like a friend. The flops: a disk-space fix and a cooking show. And the low point was a weekend bug that drained usage on 40,000 accounts, which OpenAI apologized for. What was your favorite day? Save this post and follow for more AI news.

Day 27 of OpenAI's 28 days, and Tibo's entire post is "Banked 1/2." It's the first of two banked resets for the final two days, saved to your account so you can use it whenever you hit your limit.

Some users are already cranking GPT-6 Astra to max. That's reset number 10 in 27 days, with one more coming tomorrow for the finale. Are you saving yours or using it right away? Save this post and follow for more AI news.

Day 26 of OpenAI's 28 days: you can now switch between multiple ChatGPT accounts on mobile, with desktop coming soon.

But the real news is the finale. Tibo says the next two days will be banked resets, meaning they're saved to your account and you use them whenever you want. That would make 11 resets in 28 days. His tease: "do with this what you may." No tricks, just treats this Halloween. When would you use yours? Save this post and follow for more AI news.

Claude just killed the 5-hour limit, at least on its top plan.

Anthropic is renaming Max 5x to just "Max" at $100 a month and Max 20x to "Ultra" at $200, and verified Ultra accounts no longer have a 5-hour limit, starting today. Newer accounts keep a temporary limit until they're verified, and weekly limits haven't changed. This lands one day after OpenAI doubled limits on its Pro plans. Claude Ultra or ChatGPT Pro: which is worth $200? Save this post and follow for more AI news.

Claude Code reportedly won't clone video games anymore.

Users spotted a new rule around October 27 saying Claude can build original games, genre games, or games from your own designs, but not a recreation of a specific existing game, even if you rename it. There's no setting to turn it off, and Anthropic hasn't commented. Meanwhile, OpenAI says its blocks on reverse engineering, like cleaning spyware off your TV, were accidental and are getting fixed. Copyright protection or AI censorship? Save this post and follow for more AI news.

I tested Claude Duet, and honestly, this is what an AI agent should be.

It's Anthropic's new app where Claude works as a proactive agent. It watches your inbox, calendar, and follow-ups, and messages you when something needs your attention. It's intuitive, it's organized, and it checks with you before it does anything on your behalf. Right now it's a research preview, waitlist only, for US Pro and Max subscribers. Muse, dots, or Duet: which agent are you trusting with your day? Save this video and follow for more AI news.