3 November 2026
Anthropic reveals nearly $900B in spending commitments ahead of its IPO
Anthropic's IPO filing took most of the attention today. It shows a very large revenue figure and even larger spending promises. There was also news on AI security, Claude Code rules, lawsuits over AI scraping and new ways to test AI agents.
Enterprise
Anthropic reports $17B Q3 revenue in revised S-1 (opens the source in a new tab)
Anthropic's revised S-1 filing reports $17B in revenue for the third quarter. It also lists commitments of $895B.
Why it matters: If Claude is part of your stack, watch how these commitments affect pricing and supply over the next year.
cnbc.com via Hacker News · 2 Nov
Anthropic reveals $900B AI spending commitments ahead of IPO (opens the source in a new tab)
The Financial Times reports that Anthropic has disclosed spending commitments of about $900B before its IPO.
Why it matters: A supplier this size matters to your planning, so keep a backup model in case terms change.
ft.com via Hacker News · 2 Nov
Oracle files to sell $20B in new stock and bonds (opens the source in a new tab)
Oracle has filed to sell $20B in new stock and bonds to pay for its AI buildout.
Why it matters: Cloud providers are borrowing heavily for AI capacity, so expect that cost to shape what you pay for AI services.
cnbc.com via Hacker News · 2 Nov
Garry Tan spends up to $200K a year on AI tokens (opens the source in a new tab)
Y Combinator's Garry Tan says on a podcast that he spends up to $200K a year on tokens for his AI agents.
Why it matters: Heavy agent use can run up large bills, so set a usage budget before you scale an agent workflow.
TechCrunch · 3 Nov
OpenAI pitches data center power plants to protect nearby areas (opens the source in a new tab)
OpenAI will pilot its Grid Protectors Program at a data center in Effingham County, Georgia. It promises to free up to 1,000 megawatts during extreme weather or emergencies.
Why it matters: Power supply and local opposition limit AI capacity, which can affect future availability and cost of AI tools.
The Verge · 2 Nov
Security and safety
Agentic backdoors via pretraining poisoning (opens the source in a new tab)
A research paper looks at how backdoors can be planted in AI models by poisoning their training data, so that agents misbehave later.
Why it matters: If you let agents act on your systems, use models from trusted vendors and limit what an agent is allowed to do.
arxiv.org via Hacker News · 2 Nov
OpenAI raises security bug bounty rewards (opens the source in a new tab)
OpenAI is raising Security Bug Bounty rewards. It is adding bonuses for report quality, loyalty and priority campaigns for independent researchers.
Why it matters: More outside researchers hunting for flaws should help the tools you rely on, but still keep your own access controls tight.
OpenAI · 2 Nov
OpenAI prepares next security sprint with Cathedral Systems (opens the source in a new tab)
OpenAI is preparing an internal security sprint with Cathedral Systems. It will use AI to test and strengthen its defenses.
Why it matters: Vendors are using AI to defend as well as build, so ask your own providers how they protect your data.
OpenAI · 2 Nov
Tools and products
Anthropic stops Claude Code from re-creating existing video games (opens the source in a new tab)
Anthropic has updated Claude Code's system prompt so it will not be used to re-create existing video games. Making new games is still allowed.
Why it matters: If you build with Claude Code, expect it to refuse some copying tasks, so check your work does not copy others.
The Verge · 2 Nov
Claude's game cloning rules won't stop AI copying (opens the source in a new tab)
Ars Technica says the wave of AI imitators is far bigger than any one model maker can stop.
Why it matters: Copying is easy with AI, so protect what makes your own offer hard to imitate.
Ars Technica · 2 Nov
Apple Watch AI conversation recap features enter beta (opens the source in a new tab)
Apple's new Audio Intelligence features are in beta. They can help you remember what someone said or take notes on conversations.
Why it matters: If you take notes in meetings or calls, check consent and privacy rules before recording people.
Agents
Stop counting tokens, start counting what ships (opens the source in a new tab)
MIT Technology Review says a single agent coding task can use 10 to 50 times more tokens than a basic chat prompt.
Why it matters: Measure agent cost against finished work, not tokens, so you can see if it pays off.
MIT Technology Review · 2 Nov
Raising the quality of insights in agentic analytics (opens the source in a new tab)
MIT Technology Review says agentic AI promises to automate whole workflows. It also says the quality of insights agents currently generate is a problem.
Why it matters: Check an agent's reports by hand before you make business decisions from them.
MIT Technology Review · 2 Nov
The agentic web is here (opens the source in a new tab)
MIT Technology Review looks at the idea of an internet built for autonomous software agents, going back to Tim Berners-Lee's 2001 vision.
Why it matters: Think about whether your website and services can be understood and used by AI agents.
MIT Technology Review
Also worth knowing
OpenAI introduces LifeVal benchmark for everyday tasks (opens the source in a new tab)
OpenAI introduced LifeVal, a benchmark that measures how well AI helps people finish everyday tasks. Examples include finding childcare, managing finances and handling government paperwork.
Why it matters: It is a more practical test than most, so it may help you judge which assistant suits real admin work.
OpenAI · 2 Nov
Anthropic warns it is very close to self-building AI (opens the source in a new tab)
TechCrunch reports that Anthropic warns it is very close to RSI, where AI gets so good at training itself that capabilities speed up sharply.
Why it matters: Model abilities may change fast, so review your AI tool choices more often than once a year.
TechCrunch · 2 Nov
Apple hints at new revenue from third-party AI apps (opens the source in a new tab)
On its earnings call, Apple hinted at ways to earn revenue from third-party opportunities involving AI apps. It gave few details.
Why it matters: If you sell an AI app on Apple devices, watch for fee or rule changes.