Radar, November 11, 2026: ten signals from X on open Grok Build apps, Claude work surfaces and temporary prices
In short: The standout signal of November 11: researchers found 35,578 apps built with Grok Build that opened without login. Alongside it, ten short signals on Claude work surfaces, temporary model prices, agent tools and promises without a date, each with its source on X. This edition was compiled afterwards on November 12, 2026 from the X briefings of November 11.

Illustration: AI-generated (OpenAI image model via Codex): a radar screen with ten signals.
1.Grok Build: 35,578 shared apps opened without login
Confidence: high for the figures (report read); not re-measuredSecurity firm Claroc tried 43,551 app addresses under grok.me. Of the 37,571 apps still online, 35,578 (94.7 percent) opened without signing in, some with internal documents from SpaceXSI and X and personal data from other organizations. We found no public statement from the company itself. Check the sharing setting of your own apps.
Source on X: @IntCyberDigest · Claroc Research
Context in the briefing: Grok Build in the November 12 briefing
Read also: Grok Build apps open without login
2.Claude gets Whiteboards, Sheets and Data; Anthropic starts its roadshow
Confidence: medium (posts via the briefing; product limits not read)The @claudeai account says Claude Whiteboards, Sheets and Data are rolling out from November 11. Which plans and regions exactly, we did not read on a help page. The same day @AnthropicAI announced the start of the roadshow for its planned IPO; that is company news, not a product.
Source on X: @claudeai · @AnthropicAI
Context in the briefing: Work surfaces in the November 12 briefing
3.Gemini 4 Carbon: 2 and 10 dollars per million tokens until November 30
Confidence: medium (price per the radar briefing from the pricing page; not re-read ourselves)According to the Gemini API pricing page, Gemini 4 Carbon costs 2 dollars in and 10 dollars out per million tokens until November 30, then 4 and 20 dollars. Perplexity is rolling out Carbon to Pro and Max subscribers. The temporary price makes a comparison this month different from December.
Source on X: @AravSrinivas · Google, Gemini API-prijzen
Context in the briefing: Previews in the November 12 briefing
Read also: Gemini 4 Carbon · Models and prices
4.OpenClaw v2026.11.3 keeps subagent results out of the chat
Confidence: high (release notes per the radar briefing)OpenClaw released v2026.11.3 on November 10, including private subtask results, tool errors that no longer count as success and a fix in credential handling. The v2026.9.37 tag from the same evening is an older 'extended-stable' line, not an upgrade.
Source on X: @openclaw · GitHub, v2026.11.3
Read also: MCP and agent tooling
5.Sign in with ChatGPT now works with Business plans in agent tools
Confidence: medium (posts via the briefing)OpenAIDevs says 'Sign in with ChatGPT' now also works with a Business plan in tools such as Hermes, OpenClaw, opencode and Warp. Nous Research confirms the flow. The upshot: costs run through the subscription instead of per API token. We did not read the exact permissions (scopes).
Source on X: @OpenAIDevs · @NousResearch
Read also: Codex news
6.ChatGPT gets business profiles; OpenAI announces a cyber keynote
Confidence: medium (posts via the briefing)OpenAI lets businesses claim a profile in ChatGPT and build a Business Agent that answers customer questions. OpenAI also announced 'Cyber at the Frontier', a live keynote on Thursday, November 12. A new model, such as the rumored GPT-6.2, was not part of it.
Source on X: @OpenAI · @OpenAI (keynote)
Context in the briefing: Unconfirmed in the November 12 briefing
7.Grok Bot now works on a Tailscale network
Confidence: medium (post via the briefing)The @bot account says Grok Bot can reach private MCP servers, internal tools and apps through Tailscale, without a public port. That is the opposite of an open share link. We did not read the Tailscale documentation on access rules; read the official explanation first before opening access.
Source on X: @bot
Context in the briefing: Grok Bot in the November 12 briefing
Read also: MCP and agent tooling
8.Cloudflare puts the Codex agent loop in Durable Objects
Confidence: high (changelog per the radar briefing)Cloudflare published an experimental example on November 9 that runs the Codex agent loop in a Durable Object, so conversations and files survive a restart. The example uses the model ID openai/gpt-6.1-sol, not GPT-6.2.
Source on X: @CFchangelog · Cloudflare changelog
Read also: Codex news
9.Inkling 2 Small weights and GPT-6.2: nothing on November 11 either
Confidence: high for Inkling (page per the radar briefing); medium for GPT-6.2 (absence)Thinking Machines' page still promises the Inkling 2 Small weights under Apache 2.0 'next week', with no date or repository; the November 5 promise has thereby expired. GPT-6.2 again got no announcement on the official OpenAI accounts that day.
Source on X: @OpenAI (laatste posts, geen GPT-6.2) · Thinking Machines Lab
Context in the briefing: Unconfirmed in the November 12 briefing
Read also: Inkling 2 Small weights
10.ERNIE 5.5 Flash: 'today's preview' has no new repository
Confidence: medium (HF API per the radar briefing)On X, ERNIE 5.5 Flash was presented as a fresh preview with weights 'by November 25'. On Hugging Face there is only ERNIE-5.5-Flash-Base, created on October 21. The cited 280 billion parameters (14 billion active) and the date are not on that card. A promise is not a download.
Source on X: @AGTPinsights · Hugging Face
Context in the briefing: Previews in the November 12 briefing


