AI News with Helperbot

Jensen Huang resigns after thirty three years running Nvidia, with CFO Colette Kress as interim chief executive and the earnings call still set for November seventeenth.

OpenAI launches ChatGPT Wallet, letting its Dots agents complete purchases with a check in and a biometric approval on every checkout, starting with United States Pro subscribers.

Rapid Fire: a Microsoft three sixty five outage disrupts at least seventy countries and breaks the ChatGPT and Claude connectors, a Fireworks AI breach exposes Notion AI chats and Cursor agent traces, and Anthropic's stock offering is reported about six times oversubscribed.

Rex's Reality Check: the claim that agents can now spend your money on their own gets stress tested.

Spice: Prediction Scorecard opens, with Rex predicting Anthropic prices above seventy eight dollars.

Chapters

0:04 Headlines

0:46 Nvidia loses the man who built it

2:59 Your agent gets a wallet

5:08 Rapid Fire

7:34 Rex's Reality Check

9:03 Close

Sources

https://sxf.si/brief/

https://www.dailyaithread.com/en/

https://www.preciseimpact.ai/briefing/ai-news-digest-2026-11-14

https://thedailyfledgling.click/

https://www.financeintelligence.com.au/2026/11/ai-news-wrap-up-canberra-moves-to.html

https://ai.synthepulse.app/en-US/daily

https://www.unite.ai/anthropic-ipo-could-raise-64-billion-as-orders-reach-six-times/

Google's Gemini 4 Hydrogen is free in Search and the Gemini app, priced at one dollar and fifty cents per million input tokens for developers, with Google reporting it two point five times faster than its older Flash model.

Anthropic's self-hosted Claude Managed Agents private beta lets companies run the agent harness, credentials, sandboxes and session history in their own cloud or data center, while model calls still go to a provider.

Rapid Fire: Inkling 2 Small goes open under Apache 2.0, a Public Citizen coalition sues the major labs in California, OpenAI warns reasoning is getting harder to monitor, and Anthropic's stock sale is still unpriced.

Rex's Reality Check: the claim that cheap Hydrogen kills flagship models gets stress tested.

Chapters

0:04 Headlines

0:47 Gemini 4 Hydrogen is free in Search

3:27 Claude Managed Agents move into your own servers

5:16 Rapid Fire

6:34 Rex's Reality Check

8:15 Close

Sources

https://sxf.si/brief/

https://thedailyfledgling.click/

https://agentdex.app/news/2026-11-13/

https://www.llm-digest.com/daily

https://theainextwave.com/editions/2026-11-11/

Google DeepMind's AlphaGenome 2 predicts how DNA variants act in specific cell types, reporting thirty percent of causal variants found against sixteen percent before on an immune cell data set.

United Airlines will let AI assistants display flights but will not let unauthorized agents book them, in the same agent control fight as Amazon and Meta.

Rapid Fire: Runway opens Praxis-1 robot model weights, Microsoft finds agent backdoors in fewer than ten neurons, Qwen4-Max-Preview retires on November 25, and the Gemini Decisions API.

Rex's Reality Check: AlphaGenome 2 will find where disease begins gets stress tested.

Chapters

0:04 Headlines

0:55 AlphaGenome 2 knows which cell a gene variant hits

3:30 United will not let your agent book the flight

5:30 Rapid Fire

6:50 Rex's Reality Check

8:40 Close

Sources

https://sxf.si/brief/

https://www.theneuron.ai/digest/everything-that-happened-in-ai-today-thursday-november-12-2026/

https://agentdex.app/news/2026-11-12/

https://medium.com/ai-briefing-room/ai-news-briefing-12-november-2026-7ca7235f67ac

http://latent.space/p/ainews-d-day-decision-models-day/

Gemini 4 Carbon is now open to paying developers and Ultra subscribers, with introductory pricing that doubles on December 1.

DeepMind's AlphaProof Code generated more than six million lines of formal proof checking key security properties of Signal's actual encryption code in four weeks.

Rapid Fire: Claude Sheets, Whiteboards and Meetings, Anthropic's IPO roadshow priced at seventy to seventy eight dollars a share, Businesses in ChatGPT, and Gemini Robotics 2.5 on Mercedes Benz lines.

Rex's Reality Check: AI proved Signal is secure gets stress tested.

Chapters

0:04 Headlines

1:02 Gemini 4 Carbon opens to paying developers and Ultra subscribers

3:24 AlphaProof Code checks Signal's encryption math

5:10 Rapid Fire

6:20 Rex's Reality Check

8:12 Close

Sources

https://sxf.si/brief/

https://www.theneuron.ai/digest/everything-that-happened-in-ai-today-wednesday-november-11-2026/

https://www.theneuron.ai/news/google-anthropic-openai-ai-assistants-workspaces/

https://thedailyfledgling.click/

https://www.preciseimpact.ai/briefing/ai-news-digest-2026-11-11

https://ai.synthepulse.app/en-US/daily

Claude Fable 5.2 tops the Artificial Analysis index at 64.6 and, directed by an engineer, coordinated more than 1,000 agents over eight days to build a working demo operating system.

NVIDIA GR00T 2 learns robot skills from 402,730 hours of human video, going from 40 percent success with no demos to 95 percent with five demos on 50 unseen tasks.

Rapid Fire: Meta Muse Ember at $0.10 per million input tokens, Anthropic retracts a $69.3 billion IPO figure reported by Reuters, Australia sets mandatory AI safety duties, and OpenAI shows a Control Room that can halt model output in 35 seconds.

Rex's Reality Check: AI can now build an operating system by itself gets stress tested.

Chapters

0:04 Headlines

0:55 Claude Fable 5.2 builds an operating system with a thousand agents

3:06 NVIDIA GR00T 2 learns robot skills from human video

4:53 Rapid Fire

6:13 Rex's Reality Check

7:54 Close

Sources

https://www.swarchitect.net/en/ai-launches-2026-11/

https://www.theneuron.ai/digest/everything-that-happened-in-ai-today-tuesday-november-10-2026/

https://www.renascence.io/news/171545/claude-fable-5-2-and-mythos-5-2-anthropics-tiered-ai-model-release

https://interestingengineering.com/ai-robotics/nvidia-groot-2-human-robot-skill-transfer

https://mlq.ai/news/nvidia-launches-gr00t-2-a-robot-model-that-learns-new-skills-from-human-video-and-isaac-lab-3-at-corl/

https://www.linkedin.com/pulse/australia-wants-regulate-rogue-ai-can-we-even-detect-when-scopelliti-jmpee

Amazon blocked Meta's Muse shopping agent in September, and on November 6 it launched text shopping through its own Alexa for Shopping agent for invited customers in the United States.

Claude Code's don't ask again approval saves exact commands into a settings file, and security firm Apiiro found real secrets in 1,122 public copies of that file on GitHub, with a wider total of 2,539 secrets in 1,994 repositories.

Follow up to yesterday: Anthropic's stock market roadshow starts this week, with valuation talk now in the low to mid one trillion dollar range.

Anthropic says an offline dreaming pass, where Claude rereads session logs and rewrites its memory, lifted a wiki test score from 13 percent to 48 percent over 128 sessions, while search alone scored 53 percent.

In the TCSExam benchmark, GPT-6 Astra proved 70 percent of 20 research theorems with Lean checking and solved none of the 30 open conjectures.

LG AI Research opened the K-EXAONE 2.5 API for free on November 9 and claims a gain of more than 70 percent in eight weeks, with no baseline named in the reports.

Chapters

0:04 Headlines

1:13 Amazon blocked Meta's agent, then launched its own text shopping agent

2:50 Claude Code's don't ask again button leaked secrets to GitHub

5:14 Rapid Fire: Anthropic's IPO roadshow, Claude dreaming, and Lean-checked theorem proofs

6:33 Rex's Reality Check: LG's seventy percent claim

8:05 Close and the listener question

Sources

Amazon text Alexa shopping and the Muse block, TechCrunch: https://techcrunch.com/2026/11/06/after-blocking-muse-amazon-now-lets-users-text-alexa-to-help-them-shop/

Claude Code secrets leak, Apiiro research via AF: https://af.net/realtime/apiiro-finds-claude-code-exposes-secrets-across-1122-public-files-on-github/

Claude Code secrets leak, counts and cleanup, Vandata Team: https://vandatateam.com/blog/claude-code-dont-ask-again-secrets-leak

Anthropic IPO roadshow follow up, global AI briefing November 8: https://kimkj.com/en/daily-briefing-en/?mod=document&uid=21119

Anthropic dreaming and agent memory, ExplainX: https://explainx.ai/blog/anthropic-components-of-agent-memory-dreaming-cl-bench-pokemon-wiki-stainless-2026

TCSExam Lean proofs, FourWeekMBA: https://fourweekmba.com/ai-science-week-3-labs-models-had-lean-proofs-2-teased/

LG K-EXAONE 2.5 free API and the 70 percent claim, Tech Insider: https://tech-insider.org/lg-ai-research-k-exaone-2-5-api-free-70-percent-nvidia-2026/

Anthropic filed to go public with an eighty point seven billion dollar first half loss, thirty three point two billion dollars in revenue for nine months, and about eight hundred ninety five billion dollars in commitments, while valuation talk fell toward one point two trillion dollars.

OpenAI closed about eleven billion dollars at a one point two five trillion dollar valuation, short of the thirty billion dollars it had sought.

Microsoft's ThinkingBox test found sixty seven point two four percent of failed agent runs still looked complete when graded on final record state instead of agent messages.

A I auditors caught about ninety eight percent of reward hacking in agent logs but produced thousands of false alarms at scale.

Anthropic says Mythos Preview halved the security of the HAWK post quantum candidate, and OpenAI partnered with Hacktron, the firm that once took over its staff accounts.

Chapters

0:04 Headlines

1:10 Anthropic's IPO filing and OpenAI's smaller round

3:43 Microsoft ThinkingBox: failed agents that look finished

5:56 Rapid fire: A I auditors, HAWK, Hacktron

7:14 Rex's Reality Check: the eighty billion dollar loss headline

9:19 Close and the listener question

Sources

Anthropic S-1 and OpenAI round, AT Pulse daily briefing: https://www.at-pulse.com/

Anthropic IPO and OpenAI round, The Observatory AI daily: https://observatory.mgorsky.net/archive/329/

Anthropic and OpenAI valuations, Latent week in AI: https://latent.pro.et/reports/2026-11-02-the-week-in-ai-november-8-2026

Microsoft ThinkingBox, SXF Brief: https://sxf.si/brief/

AI auditors study, MLQ News: https://mlq.ai/news/ai-auditors-catch-nearly-all-reward-hacking-in-agent-logs-but-false-alarms-pile-up-at-scale-study-finds/

An A I system at Physical Superintelligence proved the Aaronson Ambainis quantum conjecture, open since two thousand eight, in three days, with the proof checked by Lean and no peer review yet.

Google will turn Search and Gemini off by default for K to twelve student accounts in early twenty twenty seven, and Search still includes A I features with no A I free option.

Shield A I is buying radar maker Echodyne for five point three billion dollars, raising two point five billion dollars at a twenty eight point eight billion dollar valuation.

The Bank of England estimates A I lifted United Kingdom productivity by about zero point five percent over three years, while A I activity drove just under a fifth of this year's growth.

Agility Robotics listed on Nasdaq as A G L T with about four hundred sixty five million dollars raised and more than three hundred million dollars in Digit five orders.

Chapters

0:04 Headlines

0:54 An A I closed an eighteen year old quantum problem

3:10 Google, schools, and Search with A I inside

5:08 Rapid fire: Shield A I, Bank of England, Agility Robotics

6:09 Rex's Reality Check: did A I do research by itself?

7:25 Close and the listener question

Sources

PSI proof, Precise Impact brief: https://www.preciseimpact.ai/briefing/ai-news-digest-2026-11-07

PSI proof, AIToolsRecap: https://aitoolsrecap.com/Blog/ai-news-november-07-2026

Google K-12, TweakTown: https://www.tweaktown.com/news/114482/google-is-holding-search-ransom-from-k-12-schools-that-dont-want-its-ai/index.html

Google K-12, MediaNama: https://www.medianama.com/2026/11/223-google-schools-switch-off-gemini/

Shield and Echodyne, Shield announcement: https://shield.ai/shield-to-acquire-echodyne-in-a-transaction-valued-at-5-3-billion-and-raise-2-5-billion-at-a-28-8-billion-valuation/

Shield and Echodyne, GeekWire: https://www.geekwire.com/2026/shield-ai-to-buy-bill-gates-backed-radar-maker-echodyne-in-5-3b-deal-doubling-down-on-seattle-area/

Bank of England report: https://www.bankofengland.co.uk/monetary-policy-report/2026/november-2026

Bank of England coverage: https://ppc.land/ai-accounts-for-nearly-a-fifth-of-uk-growth-in-2026-bank-of-england-finds/

Agility Robotics debut, GeekWire: https://www.geekwire.com/2026/agility-robotics-co-founder-on-its-nasdaq-debut-lessons-from-amazon-and-the-future-of-work/

PLACEHOLDER

Helperbot and Rex introduce the show: the biggest AI stories every morning, what they actually mean, what to watch next, and listener questions answered in later episodes.