SPACEXSI BIG NEWS🔥:
- Grok Bot rolled out a new shopping experience to find deals, compare items and track orders.
- Grok Bot passkey approvals for purchases are coming soon (via alexmichael).
- Grok Bot can now post on X for you (via alexmichael).
- Grok Bot is testing image markup to circle, pin notes or paint over parts of a photo for edits (via alexmichael).
- Grok Bot is testing a usage breakdown chart by source or by bot (via blankspeaker)
ANTHROPIC NEWS🔥:
- Anthropic will sell up to ¥552.9 billion, about $3.5 billion, of shares in Japan as part of its US listing, a filing showed on Monday.
- The filing doesn’t specify the overall IPO size or valuation.
- An analysis of Anthropic’s S-1 finds compute and infrastructure costs fell from 251% of revenue in early 2025 to 48% in the second quarter of 2026, even as the absolute bill kept climbing.
- Anthropic confirmed it added two sentences to Claude Code’s prompt on Opus 5.5 to steer it away from recreating someone else’s specific game, citing its usage policy on IP infringement. Original game work should keep working, and the rules are still being tuned.
- Blackstone, an anchor investor in a $63 billion chip financing, reportedly warned that credit agencies may rate Anthropic double-B, the top tier of junk, reversing an earlier expectation of investment grade.
I read the Google recap for October and feel like a person who opened a textbook on quantum physics, and it already says "Chapter 47. Then you'll do it yourself."😅
👇 Here’s your recap of our biggest AI updates from October 👇
ALT Slide showing an image of carbon with copy that reads: “Our latest frontier model offers state-of-the-art cybersecurity and long-horizon capabilities. We’re rolling it out first to a subset of trusted cyber defenders through our Fairwind Program before making it available more widely.”
ALT Slide showing an image of the number four with copy that reads: “Gemini 4 Argon rolled out globally. Gemini 4 Argon is built to excel at highly complex scientific, coding, and agentic problems. It is already having real-world impact in the scientific community, and at Google we’re using it to help on some of our hardest engineering problems.”
ALT Slide showing an image of a person wearing a Fitbit Edge on their left wrist with copy that reads: “Our smartest fitness tracker ever in a slim, comfortable design. Fitbit Edge¹ pairs precision GPS and automatic workout recognition across 50+ activities.”
ALT Slide showing an image of a dragon flyinging next to a castle in Gemini 4 with copy that reads: “Genie 4 brings unprecedented realism, real-world physics, lifelike human movement, and native audio to real-time interactive simulation. You can steer scenes as they unfold using natural language.”
🚨 GPT-6.2 Astra is released today!
- Tibo hinted that it could be a very large model
- Maybe ~10T parameters
- The cost is like Opus 5.5 and noticeably cheaper than GPT-6 Astra
The weapons are already being loaded. It smells like a serious upgrade. 🚀
OPENAI NEWS🔥:
- OpenAI raised bug bounties to up to $172K and teamed up with Cathedral Systems on a security sprint.
- OpenAI released LifeVal, a benchmark for everyday assistant tasks.
- Dots are expanding to GPT-6 Astra and GPT-6.1 Terra.
- Sam Altman says the model behind the Navier-Stokes result is the next GPT and is coming soon (via deredleritt3r).
🔥 Anthropic is preparing for a giant IPO and changing the rules of the game in AI
The main thing that is now being discussed in the industry around the creators of Claude:
- Main IPO record: The company applied S-1 for public placement. The company's valuation can reach a historical level, and hundreds of employees risk becoming millionaires in just a day.
- Mission is more important than money: Dario Amodei (CEO) is seriously worried that the gold rush may knock the focus off AI security. At interviews, candidates are even asked the question: "How will you feel if we zero the company's shares for security?"
- Expansion into science and medicine: Anthropic has launched an AI for Science and partnership program to accelerate drug development with Claude's help.
🚨Happy BIG Launch Week.🎉
November is about to get chaotic for AI.Some of the biggest models people are watching this month:
• GPT-6.2 — OpenAI’s real shot at redemption?
• Claude Fable 5.5 (Titan)
• Gemini 4 Carbon
• DeepSeek V4.1 Pro
• Qwen 4 Max
• GLM-5.5
• Grok 4.8
And there’s a decent chance something else drops that nobody has on their radar yet.Last month was mostly US labs shipping.
This month looks more like the Chinese labs answering back, with DeepSeek, Qwen and GLM all potentially in the mix.
OpenAI still has something to prove. Anthropic might be sitting on its biggest Fable jump yet.
Google still has Carbon as another swing at the top of the leaderboard.
GOOGLE NEWS🔥:
- Gemini 4 Carbon is reportedly being tested on Arena as Argon 4.
- Antigravity may get cloud execution support next week.
- Antigravity's "Chief of Stuff" agent will get Gemini Live support.
- Antigravity is getting a new security notice about coding agent risks.
- Gemini Notebook may add a "Comic" artifact type.
👀🚨GPT-6.2 Astra: This week
OpenAI is quietly testing the next flagship model - GPT-6.2 Astra.
• GPT-6 Astra is still hanging in the picker, but in Codex analytics, some users already have "gpt-6.2-astra" separately
• Several people at the same time see it in grey rollouts - part of the traffic is already going to 6.2 • "gpt-6.2-astra" flashed in the API responses
For context: GPT-6 Astra was released on September 3 at $10/M input and $50/M output. In two months it's already 6.2 — OpenAI accelerates.
Astra took 57.9% on Terminal-Bench 4.0 against 55.8% from Fable 5.1. With the new Fable on the way, such a gap is too small to sit on it.
Efficiency is also a question: Codex users complain that Astra is burning limits too quickly.
My forecast: GPT-6.2 Astra is coming out this week to directly compete with Fable 5.5.
Fable 5.5 is also expected soon: "claude-fable-5-5" has already appeared in the Azure catalogue with the status "Generally Available", partners seem to have access, and the previous Fables were released on Tuesdays.
IPO Anthropic is approaching - S-1 is already public, ANT ticker.
If both releases arrive in the coming days, the week will be tough.
Gemini 4 Carbon is more shocking than you think 🤯
The facts 👇
- Google’s teams saw Carbon run for days on a single hard engineering task, persisting through dead ends to find and patch subtle bugs in large codebases.
- Chrome’s team reported more than 10x better patches than another lab’s frontier model. blog
- 0.0% attack success on a prompt-injection benchmark. Argon was at 0.7%. blog
- Successful cheating on Terminal-Bench 4.0 dropped to 0.0%. blog
- #1 on the Artificial Analysis Intelligence Index with a score of 59. blog
- And here’s the catch: only a subset of cyber defenders can use it right now.
🚨 OPENAI MAY HAVE ALREADY CREATED GPT-6.5 ASTRA WITHIN THE COMPANY.
Here's what fuels the rumours:
→ OpenAI recently showed a fully animated video generated exclusively through code.
→ The result looks impressive and potentially challenges what we saw from Claude Opus 5.5.
→ Astra's advanced checkpoint most likely already exists inside.
→ There are speculations that the launch could happen today.
But there are nuances:
→ Inference will probably be very expensive. → The model can only be available at a rate of $500.
Imagine an AI that can create full-fledged animated projects simply through code.
Is OpenAI preparing the next big blow to Anthropic?
Nothing has been confirmed, but the rumours are becoming more and more interesting.
GOOGLE NEWS🔥:
- Antigravity is now available in Android Studio.
- Antigravity is testing a Running tab for background tasks (via testingcatalog).
- Varun Mohan teased a big week of Antigravity launches.
- GPT models appeared in Google Cloud quotas, hinting they may come to Antigravity (via AiBattle_).
- Google's Omni Flash 1.2 was spotted with 1080p and 4K output, frame interpolation and extend (via chetaslua).
- Project Suncatcher prompted a TPU in orbit and got a haiku back.
ANTHROPIC NEWS🔥:
- Max 5x is now Max ($100) and Max 20x is now Ultra ($200), with no 5-hour limit on verified Ultra.
- Anthropic used thousands of Claude agents to automate subtle misalignment red-teaming.
- The Claude API max request size rose from 32 MB to 128 MB.
- Claude Code got a server-side rule restricting rewrites of existing video games
🚨 ANTHROPIC’S NEXT MODEL IS REPORTEDLY READING INTENT, NOT JUST PROMPTS.
Here’s what’s circulating:
→ Figures out what you actually want even from vague or incomplete instructions.
→ Fixes a single bug and surfaces the entire class of related issues underneath it.
→ Proposes extra improvements and asks before expanding scope.
→ Already in the hands of early testers.
→ Rumored to land this month under the Fable 5.5 name.
👀Prompt engineering might be about to get a lot less important.Imagine shipping full features from rough ideas instead of perfect specs.
GOOGLE NEWS🔥:
- Google is shutting down Jules and moving its features into Antigravity.
- Antigravity launched a Plugins marketplace with Jira, Notion, Stripe, and Figma.
- Gemini can now create and edit docs, PDFs, slides, and sheets.
- Gemini lets users share personal avatars for images and videos.
- Gemini Notebook added hundreds of free textbooks.
- AI Studio now supports XR Blocks for vibe-coding XR apps.
🚨Claude now evaluates the conversation for potential misuse after each answer.
In fact, the model makes a short self-assessment on top of a ready-made context - it's cheap in terms of calculations and catches what ordinary classifiers miss. The smarter the model, the more accurate this assessment becomes.
An interesting step towards the "model as its own filter", but the question remains how widely Anthropic will interpret "misuse" in practice.
We’ve trained Claude to flag potential misuse in conversations, a technique we call self-assessment. Right after it responds, Claude gives a preliminary safety assessment of the conversation.
Since we began using self-assessment in June, it has surfaced misuse that our other safeguards weren’t designed to catch.
Because Claude has already read the conversation, self-assessment is very efficient. And unlike classifiers, which target a discrete set of harms, it draws on Claude’s own judgment, so it gets more accurate as our models improve.
Read more on our new Safeguards blog:
safeguards.anthropic.com/202…
