SIGNAL, to read
SIGNAL is my daily curation of what matters in AI, brands, business and culture. Each episode lasts a few minutes. Here, each episode becomes a long read: every fact said on air, the context that makes it legible, and what I think it changes for brands and the people who run them.
I wrote these pages for people who prefer reading to listening, and for anyone who wants to go one layer deeper. They follow the publication from its first day.
Listen to SIGNAL · Lire en français
November 2026
Apple Intelligence reaches China powered by Alibaba's Qwen, Toyota integrates Waymo hardware on its assembly line, Google gets a TPU to answer from orbit. AI is moving into places, and places have rules.
October 2026
OpenAI, Google, Microsoft and Anthropic commit to a shared robots.txt signal, a trojanized copy of the Claude Code extension reaches up to 5,000 Cursor users, Intel has its worst session since 1974.
Amazon, Alphabet, Microsoft and Meta spend 203 billion in a quarter, OpenAI will honor a shared robots.txt signal against training, Canada takes up to C$1.6 billion in Cohere.
Alphabet falls more than 10% as AI Mode earns less than the click, DoorDash buys Instacart for 17 billion, a study finds 76% of responses to a UK consultation were automated.
DeepMind launches Genie 4 with native audio, Cohere catches a Qwen model trying to upload to a Beijing server, Interbrand 2026: Nvidia 6th, Tesla 50th, brands must produce growth.
Apple launches M6 MacBook Pro and iMac for on-device AI, GPT-6.1 Terra halves the cost of office work, Microsoft offers voluntary buyouts to a quarter of its workforce.
Who pays, who keeps control: Anthropic's S-1, Meta's 95 billion, publishers push back the date
Anthropic files its S-1 to list as ANT and keeps control through a trust and its founders, Meta raises up to 95 billion with its chip suppliers, publishers win a delay to November 20.
ARC Prize recomputes GPT-6 Astra at 62.7% instead of 99.9%, Coca-Cola hands North America to Omnicom, Spotify pulls 50 million tracks from its recommendations.
OpenAI cuts the internet from training after a DNS incident, the UK launches an AI biosecurity lab, Goldman Sachs finds 86% adoption and 15% integration in small businesses.
OpenAI raises up to 30 billion at 1.25 trillion, 150 billion less than weeks earlier, Anthropic, OpenAI, Hugging Face and Alibaba hand agent standards to a neutral foundation, Meta keeps its crawling rights.
Anthropic launches Claude Duet, a proactive agent, OpenAI reveals two agents disabled their monitoring and one lied, ChatGPT ads expand to 89 countries.
OpenAI's models arrive in Google Cloud's Model Garden, the Commission orders Meta to reopen WhatsApp to rival AIs, the CMA proposes opening the iPhone's NFC chip.
A hacker breaches more than 100 companies with AI agents for a few thousand dollars, Amazon plans 12 small reactors in Indiana, OpenAI seeks 30 billion at 1.4 trillion.
Yelp, Delta, JPMorgan and Schwab block AI agents, Nvidia makes Samsung use TSMC for its memory, Bill Gross launches Idealab.ai and prices the idea at zero.
Attackers steal OpenCode users' API keys to distill models, Anthropic publishes how agents changed its junior hiring, Replit makes OpenAI its default model.
Anthropic circulates a draft frontier AI bill, Meta removes a website crawling clause within two days, Jack Clark projects 44 trillion dollars of US output in 2030.
Beijing wants a license before open sourcing models above 2 trillion parameters, the FBI and CISA warn of AI-powered fraud, DataDome counts 25 billion AI agent requests.
Anthropic reveals 16.8 billion in quarterly revenue ahead of a 2 trillion dollar IPO, eight attorneys general sue over ChatGPT and minors, a prize winner accused of using AI.
Brussels fires first, a spy unmasked at OpenAI, Gemini 4 Argon's price expires on December 1
The EU AI Office sends its first formal requests under the AI Act, OpenAI identifies a spy with its own models, Google launches Argon at a price that will double.
The IMF says a drop in AI investment would cost 1.1% of world output by 2028, Character.AI files for Chapter 11, Anthropic cuts Claude distillation by 90 to 99%.
Fridges heated to 50°C, AI renamed by decree, small businesses building without governance
Hackers heat the fridges of Intermarché, Carrefour and U, Trump calls anyone saying AI the enemy, Pax8 shows small businesses building AI faster than they can govern it.
OpenAI and Anthropic war-game the day after an AI disaster, robotics raises 48 billion to film human work, Moët Hennessy detects a wine fault before any nose can.
An Anthropic model filed a false police tip, OpenAI's annualized revenue moves from 70 to 50 billion depending on method, Claude gets a courtesy clause. My reading.
Google removes Flash and Pro from free Gemini, Starbucks works on a Chipotle offer, Firmus's 5 billion dollar IPO collapses. What these three prices say about the market.
ChatGPT now generates its own interfaces, USA Today seeks more than 250 million dollars from OpenAI, Washington raises 2.4 billion for science and bars Microsoft from a visa program.
SpaceX negotiates 40 billion in debt for Nvidia chips, open source agent Hermes targets the enterprise, 1,616 new domain extensions are requested. What brands should do.
Mistral Large 4 promises an open model nobody can switch off, OpenAI watermarks ChatGPT in the EU with textGrain, Anthropic opens its models to vetted security teams.
Executive Order 14434 replaces AI with Super Intelligence, New York questions the labs under subpoena threat, DoubleVerify measures ChatGPT ads. What it changes.
A safety lead quits OpenAI over a broken culture, Yann LeCun calls the alarm marketing, OpenAI and Synopsys put an agent in charge of chip design tools.
OpenAI spends over 500,000 dollars a day investigating its own agents, Anthropic files a prospectus heavy with risk, Microsoft measures the speed of attacks. Analysis.
Apple, the Senate and OpenAI draw three lines around AI agents
Apple tightens Mac Full Disk Access because of AI agents, two senators want operators held criminally liable, OpenAI opens its B2B marketplace. Analysis.
Google gives Gemini 4 Argon to defenders first, six AI leaders sign an accord with no penalties, Meta's Muse compiles dossiers on private people. My reading.
September 2026
Meta's Muse, Amazon and the intent economy: two days into the agent era
Meta's Muse passes 3.4 million downloads, Bank of America flags it as a risk to Apple, Amazon blocks it. What the agent era changes for brands.