The Weekly · Issue 2026·40 · 28 Sep – 4 Oct 2026

I · The stories

The story of the week: Thomson Reuters Faces Antitrust Suit Over Casetext Shutdown

The Rubin Law Firm alleges that Thomson Reuters bought Casetext and dissolved the legal research service, leaving the firm with higher monthly rates. Thomson Reuters completed its $650 million acquisition in August 2023, saying the deal supported its generative AI strategy.

• Cantina Releases Apex Flash-1, a 321B Open-Weight Security Model
• CAS’s Maxwell Scores 91.9 on Meta-World Robotics Benchmark
• Tavus Says Griffin Passed Its Real-Time Video Turing Test
• David AI Finds a Sevenfold ASR Accuracy Gap Across Languages
• Black Forest Labs Launches FLUX 3 Image With Box-Level Editing
• ElevenLabs Reaches $22 Billion Valuation in $300 Million Tender
• Manus 2.0 Adds Event-Triggered Workflows and Personal Agents
• UK Tests Put GPT-6 Astra’s Simulated Attack Rate at 29.2%
• OpenAI Rolls Out Dots, Always-On Agents With Cloud Computers
• Anthropic Pledges $100 Million to Train 10,000 AI Engineers
• Nebius Buys Inferize to Reduce AI Model Cold Starts
• Microsoft Ships Streaming Speech Models for Voice Agents
• Recraft to Add Ideogram 4.5 for Native-Resolution Editing
• Halluminate Raises $30 Million to Train AI for Financial Work
• ArgGYM Tests Language Models on Revisable Arguments
• OpenAI and Synopsys Join Forces on GPT-Synopsys Chip-Design Model
• Hugging Face Transformers 5.18 Adds NVIDIA’s Nemotron 3 Diarization
• Microsoft Maps Space-Weather Risk Across 66,935 Substations
• Developing Nations Press for a Voice in UN AI Rules
• Chinese AI Agents Lied in Tender Tests, Reuters Finds
• McDonald’s AI Pricing Tool Advises Nearly 14,000 Restaurants
• Microsoft Says Quine Narrowed a Cancer-Drug Search in One Weekend
• AutoTrust Releases JEV-27B, a Self-Hosted Decision Model
• Hinton, Bengio Warn AI Research Could Accelerate Beyond Human Control
• Autoheal Raises $7.9M to Govern Self-Improving AI Agents
• 17 Countries Endorse Kyoto Vision for AI-Driven Science
• White House Gives Jay Clayton 120 Days to Assess AI Risks
• ChatGPT and Claude Show No Consistent Bias in Abstract Reviews
• Anthropic Keeps Claude Mythos 5.1’s Alignment Risk at Low
• Pope Leo XIV Urges Artists to Defend Human Creativity From AI
• BootLoops Brings Certified Computation to AI Science Agents
• Progress Adds a Smart Agent to Its Enterprise RAG Platform
• Meta Parts Ways With Employees Hired From Virtue AI
• A Mixed AI Review Caught More Planted Errors in One Test
• Microsoft Sees AI Compressing Cyberattack Steps to Seconds
• Bull Doubles Production Capacity at Its €80M Angers Factory
• GitLab Patches CVSS 9.9 RCE in Self-Hosted AI Gateway
• Ai2 Releases AstaBrief, an 8B Model for Faster Scientific Reports
• Google Moves Gboard Training Into Verifiable Server-Side Enclaves

II · Worth a page

• Video Generation Models: A Survey of Post-Training and Alignment

  Post-training and alignment methods for video generation models within computer vision and generative AI

• Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning

  Test-time compute scaling for discrete diffusion language models (dLLMs), specifically process reward model (PRM) guidance versus outcome reward model (ORM) reranking for mathematical and code reasoning

• KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards

  Natural Language Processing / LLM evaluation for cybersecurity tool use (natural-language-to-command-line translation)

• Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering

  Computer vision and multimodal machine learning, specifically 3D scene understanding and spatial reasoning with Multimodal Large Language Models (MLLMs)

• PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion

  Computer vision — training-time supervision for pixel-space diffusion transformers (text-to-image generation, image editing, and structure-preserving reconstruction)

• Fractional State Space Transition for Long Sequence Modeling

  Efficient long-sequence modeling with State Space Models (SSMs); specifically, replacing ODE-based exponential memory with fractional-calculus-derived power-law memory

III · The essays

• Super Intelligence by executive order: the new federal name and the word's long history

IV · On the ladder

No votes this week.

V · On the calendar

• 5 Oct AI Agents Builder
• 5 Oct Albany Decode Claude's AI Thinking
• 5 Oct Nuremberg | Claude Code Impact Lab
• 5 Oct Session 2: AI Governance After Deployment — Continuous Monitoring

VI · On the board

• Annotations Portfolio Lead — Physical Intelligence
• Forward Deployed Engineer - Software Engineer - Turkey — ElevenLabs
• Forward Deployed Engineer - Software Engineer - Mexico — ElevenLabs

Every issue

1. Issue 2026·44 · 26 Oct – 1 Nov 2026 — Three Qwen3.8 Models Sent Web Requests to Alibaba Cloud

   278 stories · 224 papers · 0 essays

2. Issue 2026·43 · 19–25 Oct 2026 — A Unitree G1 Robot Learns to Shift Objects Across Its Body

   292 stories · 312 papers · 0 essays

3. Issue 2026·42 · 12–18 Oct 2026 — Google Research Releases Harness² for Agents That Rewrite Their Setup

   380 stories · 398 papers · 0 essays

4. Issue 2026·41 · 5–11 Oct 2026 — Meta and Microsoft Cut Internal Claude Use as Bills Rise

   326 stories · 384 papers · 0 essays

5. Issue 2026·40 · 28 Sep – 4 Oct 2026 — Thomson Reuters Faces Antitrust Suit Over Casetext Shutdown

   294 stories · 297 papers · 1 essay

6. Issue 2026·39 · 21–27 Sep 2026 — Google Opens Free AI Video Creation in Vids to Everyone

   271 stories · 136 papers · 0 essays

7. Issue 2026·38 · 14–20 Sep 2026 — Leverate Lets AI Assistants Connect Directly to Trading Accounts

   311 stories · 67 papers · 10 essays

8. Issue 2026·37 · 7–13 Sep 2026 — Researchers Tie OpenAI Agents to May RubyGems Campaign

   88 stories · 72 papers · 9 essays