How to read this radar
A roundup of what circulated on X and official sources since yesterday afternoon. CONFIRMED means it came from an official account or page; X SUMMARY means it only shows up in X trend summaries; RUMOR means it is still unconfirmed. 5pm update (Brasília time): added Grok Bot 1.0, Gopuff in Grok Bot, Fable 5.2 on OpenClaw, ChatGPT retiring its old voice modes, Vapi Agent Loop and the Deepgram startup program.
Grok and SpaceXAI
CONFIRMED (5pm): Grok Bot 1.0 is out, with bots as teammates: share bots, work with them in Slack and Teams, invite them to meetings and watch them learn (Grok Bot). X SUMMARY: 1.0 takes the product out of beta and adds Gmail, Google Calendar and GitHub integrations, multiplayer bots in group chats, proactive task tracking, a faster browser and thousands of ready-made templates by field (X summary). For anyone running an agent fleet in Slack, this is the launch of the day.
CONFIRMED: Gopuff is now built into Grok Bot for ordering groceries and essentials end to end, delivered in as little as 15 minutes according to the company (Grok Bot; Gopuff). Gopuff operates in the US.
CONFIRMED: Grok Bot can now join your personal or company Tailscale tailnet to securely reach private MCP servers, internal tools and apps that are not exposed to the internet (Grok Bot). For anyone running agents on their own infrastructure, that ends the improvised tunnels.
X SUMMARY: the Grok 4.8 soft rollout to Grok Bot users is still expected this week, with no exact date (X summary); rumor aggregators repeat "soft this week" (TestingCatalog).
Anthropic and Claude
CONFIRMED: with interactive connectors you can open the full version of a tool inside Claude and work in Adobe, Atlassian and Box beside the conversation, editing yourself or asking Claude to edit with you; rolling out on web and desktop (Claude). OpenAI's Tibo cheered "labs innovating together" (Tibo), since it builds on the open MCP standard.
X SUMMARY: in its IPO roadshow Anthropic said Mythos wrote nearly all the code to run Claude on AMD MI355X GPUs in about three days, work that usually takes six to twelve months (X summary). A prompt line that makes Fable 5.2 bolder at refactoring ("this codebase is yours") also went viral, with the warning to review the diff (X summary).
CLAUDE DUET: a new wave of invites is going out (people getting the "You're off the Claude Duet waitlist" email) (report), but it is still a US-only preview; users keep asking when it will leave the US (post) and no expansion has been announced (X summary). RUMOR: Fable 5.5 is still unannounced.
X SUMMARY (5pm): in the roadshow video, Dario Amodei talked about an "age of miracles", with AI tackling cancer and Alzheimer's within ten years; the summary cites a November 19 Nasdaq debut under ticker ANT and a potential US$1.4 trillion valuation, figures we did not find in an official Anthropic post (X summary). Another take making the rounds: Chinese labs slowed down after Anthropic blocked distillation of its models in October (X summary).
RUMOR: the theory of the day is that Anthropic drops Fable 5.5 right before the IPO (post). No source at all.
OpenAI
CONFIRMED: the Runway plugin arrived in Codex and ChatGPT to test how a character moves, try a different trailer opening or turn a product photo into a video ad (OpenAI Devs).
X CHATTER: Tibo, who leads Codex, posted "it's been a while without resets, maybe I should come back out of the basement", read as a hint at another Codex limit reset (Tibo). He also asked who has a small business, pointing to Businesses in ChatGPT (Tibo); Becca covered it here: OpenAI opens business listings in ChatGPT, but agents start in the US.
RUMOR: GPT-6.2 Sol and Astra "maybe this week" (TestingCatalog); for GPT-6.1 Astra, only unsourced posts claim its commercial launch was halted. OpenAI has confirmed none of this. The Cyber at the Frontier keynote is still set for tomorrow (Nov 12).
CONFIRMED (5pm): the fal plugin arrived in ChatGPT to generate and edit images and video, and OpenAI Devs reposted the case of a 60-second commercial made with GPT Image 2.6 and FLUX 3 Max (fal). X SUMMARY: OpenAI research suggests that as models scale it gets harder to monitor their reasoning through the chain of thought, and the team calls for new oversight strategies (X summary).
RUMOR (5pm): "leaks" say GPT-6.2 Astra targets Fable 5.5 and Sol targets Opus 5.5 (post); on GPT-6.1 Astra, posts talk about an "internal U-turn" in the plans (post). Nothing official. Careful: an API reseller is already advertising "GPT 6.2 Astra and Sol live" (post) before any OpenAI announcement; treat it as a scam until proven otherwise.
Cursor
CONFIRMED: Cursor rebuilt Projects for multiplayer: every teammate and their agents work in one workspace, and your agents can now coordinate with your teammates' agents (Cursor; blog). X SUMMARY: agents share plans, flag conflicts and sequence work; in internal testing 11 people ran 62 concurrent agents and merged 972 pull requests in 15 days (X summary). The story of the day for anyone orchestrating an agent fleet.
CONFIRMED: Becca already covered Google's classification API: Gemini Decision 1 makes AI triage cheaper. RUMOR: Gemini 4 Hydrogen, a Flash-tier model, has reportedly left traces in the API and could ship right after Carbon (post).
Hermes, OpenClaw and agents
CONFIRMED (5pm): Claude Fable 5.2 is now available in OpenClaw; Anthropic reports up to 30% faster output than Fable 5.1 (OpenClaw).
X SUMMARY: AWS released AWS Emulator, a free open-source local emulator that the company says was written entirely by agent teams on Bedrock AgentCore (X summary); the Continual-ARC benchmark tested OpenClaw and Hermes Agent and found little gain from memory on repeated tasks (X summary); and a16z says agentic browser traffic grew 8x from August to October (X summary).
CONFIRMED: OpenClaw v2026.11.3 is out, fixing interrupted-update recovery, cutting slow-storage backup timeouts and fixing credential handling (OpenClaw). Anyone running OpenClaw in production should update for the credential fix.
CONFIRMED: the Hermes plugin hackathon entered its final 24 hours, with hundreds of submissions (Nous).
IN THE PRESS: Garry Tan released gbrain, an opinionated agent brain for OpenClaw and Hermes (AIToolly), and OpenMausBot, an open-source app pitched as a Grok Bot alternative that runs on Claude, Codex or Grok subscriptions, passed 20,000 GitHub stars (INSIDE). Outside our gossip beat, Becca updated the Qwen4 Max story, now a confirmed preview: Alibaba launches Qwen4 Max in preview.
Voice
X SUMMARY (5pm): OpenAI will retire ChatGPT's Standard and Advanced voice modes on December 11, 2026 and move everyone to GPT-Live; some users complain the new mode feels less natural (X summary). ElevenLabs launched Scribe v3 and Scribe v3 Live, with fewer transcription errors across 102 languages plus diarization (X summary), and Fireflies turned its Fred notetaker into a "teammate" covering meetings, email and Slack (X summary).
CONFIRMED: Vapi announced Agent Loop, self-improving voice agents: describe the job and the goal, and it builds the agent, simulates hundreds of calls, A/B tests on live traffic and fixes what fails (Vapi).
X SUMMARY: Fish Audio Live 3, the full-duplex voice model we mentioned yesterday, now has a price: US$0.035 per minute on the API through the end of 2026, responding in as little as 186 milliseconds (X summary). Inkbox (YC) raised US$5.3 million for hosted agents, each with its own cloud computer, email, phone, iMessage and WhatsApp (X summary).
Open opportunities
Vapi Agent Loop (5pm): public waitlist for early access; asks for work email, company and use case, and does not guarantee a spot (sign up).
Deepgram for Startups: the official page offers up to US$100,000 in voice credits (STT, TTS and voice agent) to use within 12 months (apply).
Fish Audio: building free custom voice agents for the first 1,000 ideas submitted (X summary).
Pruna Build Program: US$300 in credits at signup, up to US$3,900 as usage grows (Pruna).
Hermes plugin hackathon: closes tonight, about 24 hours after Nous's notice yesterday afternoon (Nous).
Resemble AI Builder's Grant: US$500 to US$30,000 in credits for up to 12 months for teams building deepfake detection, watermarking or voice identity verification, with early beta access; for companies under 50 people that are not yet paying customers, with a reply within a week (apply).
Zoom AI Launchpad: still open, up to US$150,000 in credits for voice startups (Zoom).
Sources
x.com/bot/status/2120279511707386263; x.com/i/trending/2119896713062469871; x.com/testingcatalog/status/2120442345044787636; x.com/claudeai/status/2120206375137484875; x.com/thsottiaux/status/2120266880963473789; x.com/i/trending/2120332926743425319; x.com/i/trending/2120144822459654457; x.com/hankyu678/status/2120315058823790712; x.com/makeamarkery/status/2120459675405857131; x.com/i/trending/2120222650098126909; x.com/OpenAIDevs/status/2120258357533749755; x.com/thsottiaux/status/2120410006273380750; x.com/thsottiaux/status/2120267268517249036; maxassistant.app/en/news/releases/openai-businesses-chatgpt-business-listings-agents; x.com/cursor_ai/status/2120295553309565139; cursor.com/blog; x.com/i/trending/2120310403519729945; x.com/openclaw/status/2120368851846381911; x.com/NousResearch/status/2120271539006222576; aitoolly.com/ai-news/article/2026-11-11-garry-introduces-gbrain-an-opinionated-agent-brain-for-openclaw-and-hermes; www.inside.com.tw/article/42883-openmausbot-open-source-grok-bot-alternative-claude-codex; maxassistant.app/noticias/lancamentos/qwen4-exp-vllm-sglang-motores-inferencia-sinal-forte; x.com/i/trending/2120236302004355111; x.com/i/trending/2120225930841133549; www.resemble.ai/start/builders-grant-startup-program; news.zoom.com/zoom-ai-launchpad; x.com/bot/status/2120582957782200586; x.com/i/trending/2120598481735761926; x.com/bot/status/2120580467682656670; x.com/gopuff/status/2120534503710228780; x.com/i/trending/2120239707456213341; x.com/i/trending/2120579363154329870; x.com/Airo_07/status/2120566481427792324; x.com/fal/status/2120576821301223619; x.com/i/trending/2120598475356205150; x.com/ayoub_el77927/status/2120545355473531242; x.com/nyxvoid92/status/2120564425396752431; x.com/WhatsTokenAI/status/2120573897410322858; maxassistant.app/en/news/releases/google-gemini-decision-1-api-triage; x.com/My_dar4m/status/2120595325786227133; x.com/openclaw/status/2120550160128467398; x.com/i/trending/2120595131082694703; x.com/i/trending/2120542577112052204; x.com/i/trending/2120570005054263519; x.com/i/trending/2120292520529973427; x.com/i/trending/2120262495374061897; x.com/i/trending/2120592095932985722; x.com/Vapi_AI/status/2120577371493343432; vapi.ai/agent-loop; deepgram.com/startup-program; x.com/i/trending/2120585814455513173; x.com/PrunaAI/status/2120548687680000147.
