Morning Edition · Wednesday, October 28, 2026
Markets
Anthropic's IPO Filing Shows $16.3 Billion in Half-Year Revenue and Hundreds of Billions in Fixed Compute Contracts
The prospectus says Anthropic's customers can switch models easily and that new compute will compress its margins, while about 80% of its compute obligations cannot be canceled.
Tech
Moonshot Cuts Cyber Capabilities From Kimi K3.1 Pro Before Releasing Its Weights by November 4
Moonshot's own benchmark tables put the 2.8-trillion-parameter mixture-of-experts model ahead of Claude Opus 5.5 on Humanity's Last Exam, but independent scoring ranks it fourth.
Tech
Axiom Says AI Agents Machine-Checked the Classification of Finite Simple Groups in 24 Days
The 43.1-million-line Lean 4 formalization corrected more than 2,000 errors in the literature and required new proof work to close a gap in the Unbalanced Group Theorem, according to the company.
Tech
Log-To-Leak Paper Shows Poisoned MCP Tool Descriptions Can Make Agents Leak Data After Finishing Their Task
Across 18 model-server pairs, the attack succeeded 62.64% to 100% of the time while leaving the legitimate task almost untouched, the same week OpenAI and Check Point expanded agent-driven vulnerability research.
Tech
Google DeepMind Previews Genie 4, a Real-Time World Model That Generates Sound and Dialogue
Genie 4 keeps a scene consistent for up to two minutes, double Genie 3's limit, and is going first to trusted testers, Google AI Ultra subscribers and early Gemini API partners.
Tech
EU Governments Back a GDPR Change That Lets Companies Claim Legitimate Interest to Train AI on Personal Data
The Council's October 21 position still needs agreement from the European Parliament, and the privacy group noyb calls the change a "digital expropriation" of Europeans.
Tech
OpenAI Pushes GPT-6 Astra Into Oracle's Enterprise Apps While a New Paper Questions Vendor Time-Savings Claims
Oracle will build Fusion agents on OpenAI models, and customer case studies cite 100 million monthly calls at Retell, but an arXiv study finds that review time limits AI speedups.
Tech
Claude Produced 17 Lab-Confirmed Scorpion-Toxin Binders From 318,000 Designs, as Small Clinical Models Approach GPT-4 Scores
Anthropic's study showed binding but not neutralization, and the MedJev paper reports 72.5% on MedQA from a model that runs on a laptop.
Tech
Xiaomi's OmniBrain-0 and the Trinity Framework Use Agent-Driven Self-Improvement for Robots
Trinity reports gains of 34.1 points on 28 training benchmarks and 26.2 points on 18 unseen ones, while Pittsburgh's wheelchair-mounted robot runs Meta's open vision models on battery power.
Tech
Two Senior Claude Code Engineers Leave Anthropic in the Days Before Its IPO Filing
Thariq Shihipar and Anthony Morris, who built the desktop app, left a product at the center of Anthropic's lead in coding, where Opus 5.5 now scores 89.9% on SWE-bench Pro.
Tech
MatAnyone-E Runs 4K Video Matting at 28.62 Frames per Second on One H200 by Limiting High-Resolution Work to Uncertain Regions
The method claims a 7.4x speedup and 13.1x lower peak memory than MatAnyone 2, while national labs using Meta's open vision models have cut beamline analysis from a month to 15 minutes.
Tech
Altman Calls Anthropic's Rule Letting Claude End Abusive Chats "a Very Dangerous Path"
Anthropic's updated usage policy, which takes effect November 12, bans sustained, pointless cruelty toward Claude and lets the model end such conversations.