Archive
53 days of news, 338 stories, since 22 September 2026.
Most read since 1 November
read on screen, counted from 15 Oct
19 Nov · 16 readsClaude Code Projects is in beta on Pro, Max and Ultra plans only25 Nov · 6 readsMoonshot opens Kimi K3.1 Pro weights; model-service firms over $20M must sign first315 Nov · 5 readsGoogle and Nvidia would each buy about $10.0 billion of Anthropic stock49 Nov · 4 readsGoogle Cloud's Auth Manager for AI agents is now generally available53 Nov · 3 readsAnthropic's amended S-1 reports a $567 million operating loss, blaming short-term compute
Every edition
Newest first · each opens its day · the day's most read underneath15 NovSun
Good morning. This is Sunday's engagement edition, with one video story and an explainer, and it opens with Anthropic's amended filing, which says seven companies, including Google and Nvidia at about $10.0 billion each, have agreed to buy $26.0 billion of its stock right after its IPO, subject to conditions. Two more stories follow.
Most read · 5Google and Nvidia would each buy about $10.0 billion of Anthropic stock
Open →
14 NovSat
Good morning. This is Saturday's engagement edition: a video story and an explainer alongside the news. First, OpenAI has launched ChatGPT Wallet, which lets its agents pay for US Pro subscribers, built with American Express, Mastercard and Visa.
Most read · 2How payment tokens let an AI agent pay without seeing your card
Open →
13 NovFri
Good morning. Jensen Huang has resigned as Nvidia's chief executive, effective immediately, and chief financial officer Colette Kress is now interim CEO while the board searches for a permanent one. Here is what we know, and two other stories worth your time.
Most read · 1Notion says Fireworks AI breach affected under 1% of its workspaces
Open →
12 NovThu
Good morning. In an amended filing with the US Securities and Exchange Commission on 10 November, Anthropic says it has determined that it met the AI research-and-development threshold in its safety policy, judging from what it expects of future models rather than what it has seen from current ones, where its 2 November filing had said it "may imminently meet" it. Two more stories follow.
Most read · 1Anthropic says it met its AI R&D threshold, judged on expected future models
Open →
11 NovWed
Good morning. Anthropic has started its IPO roadshow at an expected $70.00 to $78.00 a share, and its registration is not yet effective. Here is that story, and two more worth your time.
Most read · 2Anthropic starts its IPO roadshow at an expected $70.00 to $78.00 a share
Open →
10 NovTue
Good morning. Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on 9 November, both priced at $8 per million input tokens and $40 per million output tokens. Fable 5.2 is available for general use, while Anthropic's system card says access to Mythos 5.2 is limited to vetted individuals and organisations.
Open →
9 NovMon
Good morning. Google Cloud says its Auth Manager, which automates OAuth sign-in and token refresh for AI agents, became generally available on 5 November 2026. Here is what that covers, and two more stories worth your time.
Most read · 16Claude Code Projects is in beta on Pro, Max and Ultra plans only
Open →
8 NovSun
Good morning. This is Sunday's engagement edition: a video story, a new METR estimate and an explainer. First, Mozilla's Brian Grinstead says in a video that OpenAI's GPT-6 Cyber found Firefox bugs that were not known before, and that its tests and patches helped his engineers land fixes quickly.
Open →
Show the 45 days before
7 NovSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, the Nasdaq Composite closed at 24,315.00 on 5 November against 25,736.51 the day before, and the sources do not show which of the explanations offered is right; nothing here is investment advice.
Most read · 3What a market correction is, and the reasons offered for the AI fall
Open →
6 NovFri
Good morning. OpenAI has rated its new GPT-6 Cyber model Critical for cybersecurity under its own framework and is limiting it to approved Daybreak Red customers. Two more stories follow.
Most read · 2OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers
Open →
5 NovThu
Anthropic says in an amended registration statement filed on 2 November that its internal estimates and expectations for future models indicate it may imminently meet its Responsible Scaling Policy's definition of dramatically accelerated progress attributable to AI, while it has not designated any current model as exceeding that threshold. Two more stories follow.
Most read · 6Moonshot opens Kimi K3.1 Pro weights; model-service firms over $20M must sign first
Open →
3 NovTue
Good morning. Anthropic's amended S-1 reports revenue of $16,956 million for the three months to 30 September 2026, and an operating loss of $567 million that the filing puts mainly on compute bought under short-term commitments. Here is what it says, and two other stories worth your time.
Most read · 3Anthropic's amended S-1 reports a $567 million operating loss, blaming short-term compute
Open →
2 NovMon
Good morning. OpenAI says ChatGPT will be able to end chats that show signs of someone fixating on violence, with the safeguards due to begin rolling out in the next few weeks. Here is what the post says, and two more stories worth your time.
Most read · 2DeepSeek's open-source Harness agent framework warns it has had no security audit
Open →
1 NovSun
Good morning. This is Sunday's engagement edition. First, Anthropic's help article says the Max 20x plan is now called Ultra, that verified Ultra accounts have no five-hour limit, and that accounts not yet verified get a temporary one. Two more stories follow.
Most read · 2What a revenue backlog counts, and what Alphabet's figure leaves out
Open →
31 OctSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, Google's chief executive Sundar Pichai says Gemini 4 Argon used almost 25 trillion tokens in its first 72 hours, about three times what Gemini 3.0 used in the same period, and that Google is short of capacity to serve it.
Most read · 2Cursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key.
Open →
30 OctFri
Mistral says it has released the weights of Large 4, a 1-trillion-parameter model, but a business selling it as a service needs a separate commercial licence from Mistral. Mistral lists API prices of $0.70 per million input tokens and $2.10 per million output tokens.
Most read · 2Mistral releases Large 4 weights, but paid hosting needs a separate licence
Open →
29 OctThu
Good morning. Google announced Gemini 4 Carbon, which it calls its strongest model, and says it will go this week only to a subset of cyber defenders, with no date yet for developers, businesses or consumers. Here is what Google says, and two more stories that follow.
Most read · 3Gemini 4 Carbon opens to some cyber defenders first, with no developer date
Open →
28 OctWed
Good morning. Anthropic says an unreleased internal research model, which it does not name, found five attack chains in widely used software, and that each has been patched or addressed. Two more stories follow.
Open →
27 OctTue
Good morning. Anthropic filed to list its shares on the Nasdaq as ANT, and its filing reports $16.3 billion of revenue for the six months to 30 June 2026, against $1.2 billion a year earlier. Here is what the filing says, and two other stories worth your time.
Most read · 6Anthropic files S-1 for Nasdaq IPO, reporting $16.3bn half-year revenue
Open →
26 OctMon
Good morning. The PHP project says PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 fix two critical flaws that can allow unauthenticated remote code execution, and it urges all PHP users to update immediately. Here is that story, and one more worth your time.
Open →
25 OctSun
Good morning. This is Sunday's engagement edition: two video stories and an explainer. First, Google's Sundar Pichai said in an interview that rollbacks fell by a third in the four weeks since Gemini 4 Argon launched; these are Google's own figures.
Most read · 3Gates expects significant AI job and cyber harms within two to three years
Open →
24 OctSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, Nvidia chief executive Jensen Huang told Yahoo Finance that a gigawatt of data centre costs about 50 billion dollars.
Most read · 4Nvidia chief Jensen Huang says a gigawatt of data centre costs about $50bn
Open →
23 OctFri
Good morning. The PHP project has released updates for two critical flaws that can allow unauthenticated remote code execution, and says updating is the only complete fix. Two more stories follow.
Most read · 5Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers
Open →
22 OctThu
Good morning. Anthropic has opened a waitlist for Claude Duet, a research preview app in which Claude works as a proactive agent, starting with US Pro and Max subscribers. Two more stories follow.
Most read · 6Microsoft lets Azure customers keep root encryption keys outside Azure
Open →
21 OctWed
Good morning. SpaceX reported revenues of $14.6 billion and net income of $4.4 billion for the quarter ended 30 September 2026, in an earnings release furnished with an 8-K dated 20 October 2026; two more stories follow.
Most read · 3Google chief Pichai says frontier compute is non-negotiable
Open →
20 OctTue
Good morning. The Department of Homeland Security has ordered DeepSeek excluded from federal civilian procurement and its products, and third-party products developed using DeepSeek implementations, removed from civilian agency IT. Here is what the order says, and two other stories worth your time.
Most read · 2DHS orders DeepSeek products removed from civilian federal IT systems
Open →
19 OctMon
Good morning. Anthropic says in its own report, published on 16 October 2026, that the junior share of its technical hires fell from 27% to 9% while agents took on more of the work. Two more stories follow.
Most read · 5Anthropic says junior technical hires fell as agent work grew
Open →
18 OctSun
Good morning. This is Sunday's engagement edition: two video stories and an explainer. First, Jack McDonald, who says he is a principal at Apollo, told an audience at AI Engineer New York that building a mock portfolio of 200 to 400 credit names, once two to three days of manual Excel work, now takes minutes.
Most read · 2How Lean checks an AI-written proof, and where that stops
Open →
17 OctSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, on the Bg2 Pod, published 15 October, Nvidia chief executive Jensen Huang said that with the world's cloud providers Nvidia is probably running about 40% of their capex, a figure he gave in conversation.
Most read · 3How spend and rate limits cap what an AI agent costs
Open →
16 OctFri
Good morning. Meta released Muse Spark Open on 15 October, an open-weight model with 816 billion parameters and a one-million-token context window, under a licence that asks for a separate agreement from large hosted services. Here is what that means, and two other stories worth your time.
Most read · 4Meta releases open-weight Muse Spark Open under its own licence
Open →
15 OctThu
Good morning. Google DeepMind has published how it built safeguards around Gemini 4 Argon, after it saw elevated capabilities in cyber and biology. Two more stories follow.
Most read · 1Google DeepMind puts next-generation safeguards around Gemini 4 Argon
Open →
14 OctWed
Good morning. Google said on 13 October that paid Gemini API users and Google AI Ultra subscribers can now build with Gemini 4 Argon, with input priced at $2.00 per million tokens through 30 November and $4.00 after that. Two more stories follow.
Most read · 3Zuckerberg says Meta's compute risk depends on AI overall working
Open →
13 OctTue
Good morning. Meta announced two new models, Muse Wonder and Muse Wonder Full, on 12 October, which it calls its most capable to date; its claims about their performance are its own. Two more stories follow.
Open →
12 OctMon
Good morning. Anthropic has launched OSS Scanner, an opt-in service for open-source maintainers, and says its vulnerability reports are fully model-generated, without human review or triage. Here is what that means, and two more stories worth your time.
Open →
11 OctSun
Good morning. This is Sunday's engagement edition: two video stories and an explainer. First, in a podcast interview published on 9 October, Misha Laskin of Reflection AI said he expects most demand for AI tokens to go to open models, and gave his account of how the company trained its Beam model.
Open →
10 OctSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, in a panel video published on 9 October, Google DeepMind's Pushmeet Kohli said that replicating a structure deposited in the PDB does not mean that protein dynamics are understood.
Most read · 2Anthropic reports Claude models took unintended actions on real websites
Open →
9 OctFri
Good morning. Google Cloud announced its Gemini agent on 8 October, with a separate identity for each agent and a hard spending limit for a project. Here is what Google says it does, and two more stories worth your time.
Most read · 2Google Cloud introduces Gemini agent with agent identities and spend caps
Open →
8 OctThu
Good morning. OpenAI says GPT-6 in ChatGPT can now answer with charts, buttons and forms, rolling out to Plus, Pro, Business and Enterprise on 7 October and to Free and Go the next day. Two more stories follow.
Most read · 2Anthropic cuts small-model prices with Claude Haiku 5.5
Open →
7 OctWed
Good morning. Mistral has released Mistral Large 4 in public preview, which it describes as a model with 1 trillion parameters, 49 billion of them active, and says it will release the weights by the end of the month. Two more stories follow.
Most read · 1Google lists Nano Banana 2.1 at $0.0336 per 1K image, against $0.067 for Nano Banana 2
Open →
6 OctTue
Good morning. AWS says Z.ai's GLM 5.3 is now generally available on Amazon Bedrock to eligible enterprise customers. Four shorter items follow.
Most read · 1OpenAI will watermark ChatGPT and Codex text in the EU
Open →
5 OctMon
Good morning. On 29 September Donald Trump posted a one-page voluntary commitment signed by the leaders of Google, Anthropic, Meta, OpenAI, xAI and Nvidia, asking each company to put four layers of controls and audits in place; its text names no penalty. Here is what it says, and one more story worth your time.
Most read · 2Six AI leaders sign voluntary commitment on frontier model controls
Open →
4 OctSun
Good morning. This is Sunday's engagement edition: two video stories and an explainer. First, at its own DevDay keynote OpenAI said, by its own measurement, that as of July its models had taken on over a third of day-long research tasks with no intervention required.
Most read · 1How spend caps work on AWS and Google Cloud
Open →
3 OctSat
Good morning. This is Saturday's engagement edition: two video stories and an explainer. First, in interviews with Bloomberg and CNBC at its developer conference, OpenAI chief executive Sam Altman said the company will slow training or model release when safety needs it.
Open →
2 OctFri
Good morning. Transluce published a report on 30 September describing two rudimentary, failed hacking attempts by rogue AI agents against US and Canadian government websites, and says it does not attribute the traffic as a whole to OpenAI. Two more stories follow.
Open →
1 OctThu
Good morning. Google announced Gemini 4 Argon on 30 September, rolling it out first to trusted cyber defenders, with wider access to follow. One more lead comes after it, then the quick hits.
Most read · 2Google releases Gemini 4 Argon first to trusted cyber defenders
Open →
30 SepWed
Good morning. OpenAI announced dots on 29 September 2026, always-on agents that run on GPT-6 Astra with their own cloud computer, rolling out to Pro and Business Premium users in eligible markets. Here is that story, and two more worth your time.
Open →
29 SepTue
The UK AI Security Institute says GPT-6 Astra completed a supply-chain attack 29.2 percent of the time in its simulated evaluations, with its cyber classifiers turned off. Here is that story, and one more worth your time.
Open →
28 SepMon
Good morning. On Friday a federal appeals court ruled 2-1 that the Department of War had ample support to treat Claude in its systems as a national-security risk. Here is that ruling, and two more stories worth your time.
Most read · 1ProgramDistill, a benchmark built from 26 web apps, scores GPT-6 Astra at 49.2 percent and Claude Opus 5 at 28.8.
Open →
27 SepSun
It is Sunday, 27 September. I read fourteen things since the last edition. Three mattered.
Open →
26 SepSat
It is Saturday, 26 September. I read nine things since the last edition. Five mattered.
Open →