An expanded class-action lawsuit claims Anthropic failed to adequately disclose weekly usage caps on its Claude Max plans, which are marketed as offering 5x or 20x more usage than Pro. The case highlights how AI subscriptions bundle unpredictable compute and could force providers to be clearer about limits.

GPT-6 Astra from OpenAI is now generally available on Amazon Bedrock. It delivers deeper reasoning and sharper judgment for complex, high-stakes tasks such as financial analysis, contract review, codebase debugging, and software automation. Organizations can call the model via Amazon Bedrock APIs or configure ChatGPT Work and Codex to use it, with enterprise plugins extending browser-use capabilities across common business tools. Security is layered through OpenAI model safeguards and AWS controls like zero-operator access, encryption, IAM, VPC endpoints, and CloudTrail logging.

Frigade Assist API gives AI agents a way to show users interactive guides that point to where they should click next.

Mistral started as an open-weight AI startup, but its focus is now shifting toward sovereign AI in response to the European market. The new funding round is expected to help bridge that transition.

OpenAI claims it has solved the long-unresolved Navier-Stokes problem with an internal model and 10,000 concurrent agents, but a mathematician who published related research the day before has raised concerns that OpenAI may have used his Codex sessions.

Google DeepMind has launched the AlphaGenome Atlas, a free academic resource with precomputed molecular effect predictions for roughly 9 billion human single-nucleotide variants, plus AVI scores, feature attributions, and DNA motifs. Early collaborations show validated rare-disease findings and broader non-coding association discovery.

Discussion | Link

Anthropic is facing an expanded class action alleging that it deceptively marketed its Claude Max subscription tier. The lawsuit says the advertised 5x and 20x usage limits are misleading because they apply only to five-hour sessions and are capped by weekly limits. Two former FTC attorneys are representing the subscribers; Anthropic has not commented on the latest filing.

deepagents introduces context modes that decide whether subagents inherit a supervisor’s conversation (fork) or start clean (isolated). Fork can be faster and cheaper by reusing prompt caching, while isolated is ideal for independent review or parallel research. The post maps worker, verifier, researcher, and memory agents to the right mode.

An MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.

Catalyst Lab director Chris White talks with Weishung Liu about his path from DARPA fieldwork in Afghanistan and dark-web search tools against human trafficking to leading public-good technology research at Microsoft, and the humility and resilience that guide him.

HPE Zerto partnered with AWS to build an Amazon Bedrock-powered agentic troubleshooting system that runs on-premises as part of the Zerto product. Using the Strands Agents framework, Guardrails, Knowledge Bases, and multi-agent orchestration, it grounds AI answers in live disaster-recovery data. Since launch in Q2 2026, more than 20% of HPE Zerto customers have adopted it, and supported workflows saw a 10% reduction in support cases.

Databricks shows how to build cloud agents that survive worker restarts and long human reviews by pairing Temporal's durable execution with Lakebase Postgres as queryable operational state. The underwriting reference implementation records evidence, decisions, and metrics with deterministic IDs and guarded writes, while Unity Catalog supplies governed policy and Change Data Feed can publish audit history back to Delta tables.

Enterprises are increasingly looking beyond data residency when considering AI sovereignty.

In this article, you will learn the key differences between Chain of Thought and Tree of Thoughts prompting, and how each reasoning framework is applied...

Open models are more competitive than ever in 2026, but license trends are diverging: Western makers Google and Meta are moving toward Apache 2.0, while Chinese frontier labs are adopting restrictive custom terms. Zhipu GLM-5.3 introduces a $10B revenue threshold and security review requirement, and new releases such as Motif-3, GLM-5.3-Flash, and Tencent Hy4-preview show the breadth of the ecosystem.

This article presents five no-cost ways to access AI coding agents and models: free model access through OpenCode Zen, OpenAI Codex on the ChatGPT free plan, Kilo Code's Auto Free routing, OpenRouter's free model API, and Google Antigravity's $0 individual plan.

Google DeepMind has released the AlphaGenome Atlas, a freely accessible public database containing precomputed predictions for all 9 billion possible single-letter DNA variants in a reference human genome. Designed to help researchers study how genetic changes affect gene regulation without running the computationally intensive AlphaGenome model themselves, the Atlas adds an impact score but carries important limitations.

Google DeepMind has released AlphaGenome Atlas, an AI-powered predictive map it says covers every possible DNA single-letter change in the human genome. The platform aims to help researchers understand the molecular effects of mutations, prioritize the most promising variants, and speed up disease research and drug development.

As LLM applications grow more complex, inference cost and latency become increasingly important. This article breaks down four caching techniques used in LLM serving—KV caching, prefix caching, prompt caching, and semantic caching—and explains how each one avoids repeated computation at a different layer of the stack.

Discussion | Link

OpenAI has introduced ChatGPT Images 2.5, an updated image generation tool that transforms ideas, sketches, and reference photos into more personalized, refined images that better reflect your creative intent.

Discussion | Link

We are sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, featuring a writeup and a formal proof in Lean.

OpenAI invites applications for a $5 million grant program supporting independent research on how generative AI affects teen development, well-being, and safety.

Discussion | Link

Reducto has introduced r-1, a new document parsing model that folds OCR, layout detection, table extraction, formatting, and grounding into one full-page pass. It claims a 20% error reduction versus its legacy agentic pipelines and drops pricing from 3–6 cents per page to a flat 1 cent, available now in preview via the V3 Parse API.

Catenary is a spatial canvas IDE for AI coding agents, presenting a visual, canvas-based workspace for code and AI-driven development. It is currently listed on Product Hunt with a discussion thread and link.

We choose embedding models, dimensions, and indexes. The geometry usually comes with the package. But why use a flat space, and what else could we choose? What Is a Manifold, and Where Do Our Vectors Live? A manifold is the space our embeddings live in. For embeddings, we care about the geometry we give that space. It determines how we measure distance, what the shortest path looks like, and how much room there is as we move outward.

Engineers at 1Password use Codex to rapidly build new features and internal tools, reaching production-readiness while maintaining rigorous security policies.