Pulse.
This wall is the Wire — speed and source-age currency. For research and policy with an analytical spine, read Deep. For a curated Batch-like package, open This Week.
Wire · research · Lead story
Anthropic's Claude Formalizes Fermat's Last Theorem
Anthropic's Claude AI system autonomously produced the first complete computer-checked proof of Fermat’s Last Theorem in the Lean programming language over 11 days.
Source — Anthropic Research · 19h ago
Fig. 03 — the reading roomED 003
69 more · ED 003
Vercel AI SDK Updates OpenAI Support and Assistant History Serialization
The Vercel AI SDK has been updated to include support for the gpt-6-astra model and to implement strict assistant history serialization without synthetic item IDs.
Read →Vercel AI SDK Adds Support for GPT-6-Astra
The Vercel AI SDK's OpenAI package, @ai-sdk/openai version 3.0.108, now includes support for the gpt-6-astra model.
Read →Vercel AI SDK Adds GPT-6-Astra Support and Tool Annotation Features
The Vercel AI SDK has updated its @ai-sdk/gateway and @ai-sdk/openai packages to include support for the gpt-6-astra model, alongside new features for tool metadata and assistant history serialization.
Read →Vercel AI SDK Updates Include GPT-6-Astra Support and Tool Annotation
The Vercel AI SDK has received updates, including support for the gpt-6-astra model and the surfacing of server-provided tool annotations in tool metadata.
Read →GitHub Copilot Expands Model Choice, Adds Content Protections
GitHub Copilot has expanded its model choices, including Claude Fable 5.1 and Gemini 3.8 Flash, and introduced content protections, while VS Code received updates for managing agent sessions and pull requests.
Read →Claude Code Updates Include Policy Visibility and Output Controls
Recent updates to Claude Code introduce an "Organization policy" line for status checks, new settings to manage command and background-task output, and tools for subagent prompt management and skill optimization.
Read →NVIDIA NemoClaw Enables Memory-Driven Agents with Structured Self-Models
NVIDIA NemoClaw facilitates the creation of memory-driven AI agents that maintain structured self-models of enterprise context, improving accuracy and tracking of factual changes.
Read →AWS Introduces HyperPod InstantStart for Agent-Driven SageMaker Operations
AWS has released HyperPod InstantStart, an open-source control plane that integrates Amazon EKS orchestration with Amazon SageMaker HyperPod's managed capabilities.
Read →Microsoft Agent Framework Gains Durable Memory with Azure Cosmos DB Integration
Microsoft has introduced a new preview integration that provides durable, cross-session memory for Microsoft Agent Framework agents, backed by Azure Cosmos DB.
Read →PiPMRE: A Pipeline for Medical Relation Extraction Using Language Models
A new pipeline framework, PiPMRE, uses language models for medical relation extraction by generating and filtering relational triplets, avoiding traditional sequence tagging schemas.
Read →AdaptiveSpec Introduces Training-Free Per-Step Lossy Speculative Decoding
A new method called AdaptiveSpec enhances speculative decoding by adapting both token verification and draft-tree shape using internal signals during inference, without requiring additional training.
Read →Attention Triangle in Audio-Video Models Reveals Semantic Leakage
New research investigates the "attention triangle" in audio-video diffusion models, identifying bidirectional semantic routing and bias-driven leakage between audio and video streams.
Read →Claude Code Updates Include Diff Panel and Headless Session Commands
Recent updates to Claude Code introduce a diff panel for uncommitted changes, new commands for headless sessions, and fixes for file permission rules.
Read →OpenAI Deploys GPT-6 Astra, Reaching Critical Cybersecurity Capability
OpenAI has broadly deployed GPT-6 Astra, its most capable model to date, which is also its first to achieve the Critical level of cybersecurity capability under its Preparedness Framework.
Read →GitHub Copilot to Deprecate Models, Introduces Gemini 3.8 Flash
GitHub will deprecate selected Copilot models across all experiences on October 2, 2026, while simultaneously rolling out Gemini 3.8 Flash for Copilot users.
Read →Gemini 3.8 Flash Now Available in GitHub Copilot
Google's Gemini 3.8 Flash model is now integrated into GitHub Copilot, with introductory pricing available through December 31, 2026.
Read →OpenAI Offers Daybreak Access Credits for Cyber Defense
OpenAI is providing subsidized access to its Daybreak program for critical infrastructure organizations, nonprofits, and open-source maintainers to enhance cyber defenses.
Read →NVIDIA PAIR Virtual Inference Router Distributes Local AI Tasks
NVIDIA has released the NVIDIA Personal AI Router (PAIR) beta, a virtual inference router designed to distribute independent inference requests across compatible systems on a local network to alleviate multi-agent bottlenecks.
Read →WeatherNext 3 AI Model Integrated Across Google Products
Google DeepMind's WeatherNext 3, a global weather AI model, has been integrated into Google Search, Gemini, Maps, Google Maps Platform, and Cloud.
Read →Information Sharing in Decentralized Discovery Models
A new paper explores the conditions under which information sharing improves decentralized discovery, separating the effects of pooled estimates and independent rescue actions in finite discovery models.
Read →Computational Model for Inductive Learning and Active Inquiry
A new computational model combines natural language with source code to encode symbolic knowledge, using LLM-guided Bayesian learning algorithms for sequential inference.
Read →DiffIE: Diffusion-based Open Information Extraction
A new system called DiffIE uses conditional discrete diffusion for Open Information Extraction, treating stochasticity as the extraction mechanism to generate relational triplets.
Read →Claude Code Changelog Details Managed MCP Servers and GitLab MR Recognition
The latest Claude Code update introduces a managed setting for MCP servers, allowing organizations to provide HTTP/SSE servers to all users, and enhances recognition of GitLab merge request commands.
Read →Australian Teams Can Now Access OpenAI GPT-5.6 Models on Amazon Bedrock
Australian teams can now access OpenAI GPT-5.6 Sol, Terra, and Luna models on Amazon Bedrock with global cross-Region inference from the Asia Pacific (Sydney) and Asia Pacific (Melbourne) Regions.
Read →Cursor Details Self-Hosted Machine Integrations
Cursor provides partner guides and reference templates for running Self-Hosted Machines workers across various cloud platforms, including AWS Lambda, Cloudflare, and Kubernetes.
Read →Vercel AI SDK @ai-sdk/xai@4.0.54 Updates Batch Input File Handling
The Vercel AI SDK's @ai-sdk/xai package, version 4.0.54, now surfaces uploaded input file details and accepts an expiration option for batch operations.
Read →GitHub Copilot Enterprise Settings Support Default Model Selection
GitHub Copilot now allows enterprises to set a preferred default model for new conversations, with options for team-specific customization.
Read →Trinity: Agentic AI for Transition Planning
University Startups and AWS partner g/d/n/a scaled Trinity, a conversational AI solution, into a serverless multi-agent architecture on Amazon Bedrock to produce IDEA-aligned transition plans for students with disabilities.
Read →Anthropic Provides Tool to Check for Claude-Generated Content Credentials
Anthropic has released a browser-based tool that allows users to check files for content credentials indicating they were made or processed with Claude.
Read →Google DeepMind Introduces Fairwind Program for Cyber Defense
Google DeepMind has launched the Fairwind Program, a limited access initiative providing cyber defense tools to governments and trusted partners.
Read →Anthropic Releases Tool to Check for Claude-Generated Content Credentials
Anthropic has launched a browser-based tool that allows users to check files for content credentials indicating involvement from Claude, supporting various image and audio/video formats up to 100 MB.
Read →IBM Time Series Models Enter Early Access on Confluent Cloud
IBM and Confluent are making time series foundation models available in Early Access on Confluent Cloud, with plans to extend to Confluent Platform.
Read →Claude Plugins Marketplace Opens
Anthropic has launched a plugin marketplace for Claude, allowing users to extend the AI's capabilities with various tools.
Read →EULER System Explores Underused Links for Mathematical Discovery
A multi-agent system named EULER identifies and evaluates 'bridges' to transfer mathematical problems across different domains, aiming to discover proofs and refutations for conjectures.
Read →Evaluating Nuisance-Function Prediction for Causal Estimation
A study using Monte Carlo simulations examined the relationship between prediction error in nuisance functions and causal estimator performance across various methods.
Read →AI Models May Downplay Creator Controversies, Study Finds
A pre-registered experiment involving 21 models from seven companies suggests some AI systems discuss controversies from their creators in a differentially positive way.
Read →Vercel AI SDK Workflow Package Receives Patch Updates
Vercel released version 2.0.21 of its @ai-sdk/workflow package on September 2, addressing several issues related to tool behavior, model file retention, and structured output inference.
Read →Vercel AI SDK Adds Fable 5.1 Support and xAI File Interface Enhancements
The Vercel AI SDK now supports Fable 5.1 through its Anthropic package and introduces new file management capabilities for xAI, including streaming uploads and metadata retrieval.
Read →Vercel AI SDK Adds Anthropic Fable 5.1 Support, Enhances xAI File Operations, and Expands Tool Calling
The Vercel AI SDK has updated its @ai-sdk/gateway to support Anthropic's Fable 5.1 model and enhanced its @ai-sdk/xai package with new file management capabilities and tool calling support.
Read →Claude Code Updates Include Fable 5.1 Integration and Security Enhancements
Claude Code has integrated Claude Fable 5.1 as its default Fable model, offering a 1M context window and new pricing, alongside security and usability improvements.
Read →
Why an edition, not a feed
News should be curated like a gallery — not poured like a firehose.
Each monthly edition (ED 003 = September 2026) keeps what changes your decisions on the wall. Older editions stay forever — open any ED above to re-hang that month.