Picking up where last week left off, a frontier lab became a public company in the largest listing on record, Nvidia posted its biggest quarter yet, and agents learned to work in teams and to pay.

A frontier lab became a public company, in the largest listing on record. On November 18, Anthropic priced its initial public offering at $83 a share, above its $70 to $78 range, selling 980 million shares to raise $81.3 billion and passing the $75 billion SpaceX raised in June as the largest offering ever. A separate $26 billion private placement with seven strategic partners, including Google, Microsoft and Nvidia, took total proceeds to $107.3 billion. Shares began trading on the Nasdaq as ANT on November 19, opened at $111.50 and closed at $97.17, up 17 percent on the day. Anthropic is a public benefit corporation, and its prospectus tells buyers in plain language that the company may choose safety over their returns.

Why it matters for how we live and build: super intelligence now has a public price, updated every trading day, and a public set of books, starting with $33.24 billion of revenue in the first nine months of 2026, so anyone can judge what this technology earns, what it costs and what its builders choose to protect.

Nvidia answered the succession question with a record quarter. On November 17, in its first earnings report since Jensen Huang resigned, Nvidia posted revenue of $115.8 billion for the quarter ended October 25, up 103 percent from a year ago, with Data Center revenue of $107.6 billion, up 110 percent, and earnings of $2.68 a share. Its new Vera Rubin NVL72 systems shipped in the quarter and brought in nearly $20 billion, the fastest product ramp in company history, and guidance for the current quarter rose to $130 billion, plus or minus 3 percent, assuming no Data Center compute revenue from China. Interim chief executive Colette Kress led the report.

Why it matters for how we live and build: the company at the centre of the buildout changed leaders and did not miss a beat in public, which steadies the supply and pricing of the computing power that almost every serious builder ultimately rents.

OpenAI shipped a stronger flagship, and cut its price. On November 17, OpenAI released GPT-6.2 Astra, built for demanding professional work, and GPT-6.2 Sol, its everyday model, in ChatGPT Work and Codex and on the API. Astra costs $8 per million input tokens and $40 per million output tokens, 20 percent below GPT-6 Astra, with cached input 60 percent cheaper, while Sol holds at $2 and $10. OpenAI reports Astra scoring 5.2 points higher than its predecessor on Agents’ Last Exam at less than a third of the cost per task, and calls it the first model it has released at its High threshold for AI self improvement, alongside a Critical rating for cybersecurity capability and tighter safeguards to match. Early power users are split, with some reporting frequent mistakes and little real world gain, so we treat the benchmarks as a claim to test, not a verdict.

Why it matters for how we live and build: capability rising while cost per task falls is the combination that moves super intelligence from special projects into daily work, and the honest test is always our own workload, not the launch chart.

Google filled the middle of its lineup, at speed. On November 18, Google released Gemini 4 Neon, its fourth Gemini 4 model in a month, priced at $3 per million input tokens and $15 per million output tokens with a one million token context window and four thinking levels. Google reports a score of 61 on the Artificial Analysis Intelligence Index, third overall, first place on its Coding Agent Index, and task completion three times faster than its flagship Carbon at 60 percent of the cost per task. Neon is live in the Gemini app for AI Pro and Ultra subscribers and through the API.

Why it matters for how we live and build: the race is no longer only about the smartest model, it is about the smartest model you can afford to run all day, and a fast middle tier is where most real work will actually get done.

Agents got a wallet, with a human check on every purchase. OpenAI launched ChatGPT Wallet with American Express, Mastercard and Visa, letting its dots agents pay on a user’s behalf. Before each purchase the user reviews the items, the total and the card, then approves with a biometric check such as Face ID or a fingerprint. The payment network then issues a tokenized credential, so neither the agent nor the merchant ever sees the real card number, and card rewards and dispute protections stay intact. It is rolling out to ChatGPT Pro subscribers aged 18 and over in the United States with eligible cards, and OpenAI takes no share of purchases at launch.

Why it matters for how we live and build: an agent that can find, compare and now pay closes the loop on delegated shopping, and the design worth copying is the control, a named human approval and a clear record on every single transaction.

Agents started working in mixed teams, and checking each other. On November 17, Perplexity added Multiagent to Computer, a new effort level above Ultra in which a lead agent running Claude Fable 5.2 plans a task and assigns parts to up to 16 agents running Claude Opus 5.5, GPT 6.1 Sol and Gemini 4 Carbon. The agents work in parallel, share findings directly and check each other’s work before the lead assembles one answer. It is available to Perplexity Max subscribers on the web. Perplexity published no benchmarks for it, so its value is a design idea for now, not a measured result.

Why it matters for how we live and build: super intelligence is moving deeper into the tools we already use as a team sport across rival models, and cross checking between different systems is a sensible way to catch errors that any single model would wave through.

China’s flagship answered with scale, and a lower price. On November 16, Alibaba released Qwen4-Max, the most capable model in the Qwen family, built on 2.4 trillion total parameters with 71 billion active per token, native vision and a one million token context window. It is priced at $1.50 per million input tokens and $4.50 per million output tokens, 25 percent below its predecessor, and Alibaba reports a score of 75.7 on the DeepSWE coding benchmark. The flagship stays closed, while Alibaba says smaller open weight models from the series will follow.

Why it matters for how we live and build: strong frontier class models now arrive from several countries at falling prices, which keeps real choice alive for builders and keeps any single lab from setting the terms for everyone.