AI news for builders and product teamsUpdated Nov 17, 2026, 20:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Databricks argues that agents built quickly with AI tools need persistent memory designed for production, not just prompt-and-forget scaffolding, to survive as maintainable enterprise systems.

Meta has paused Muse's conversation-sharing feature after Claroc researchers found over 6,800 publicly reachable Muse links via public DNS records. Meta says the feature created link-only pages and is under review for changes.

Smartsheet has released a generally available Model Context Protocol server that lets AI assistants read, write and manage Smartsheet data through more than 40 tools, with setup paths for Amazon Quick and the Kiro IDE.

Bloomberg reports that former Google CEO Eric Schmidt owns about 1% of Anthropic, roughly half the 2% stake held by co-founder and CEO Dario Amodei, after investing tens of millions in the lab's 2021 Series A.

AWS published an end-to-end pattern for conversational supply chain intelligence that uses Amazon Bedrock AgentCore as the tool and orchestration layer behind Amazon Quick, with answers pulled from live systems of record instead of model recall. An open-source sample walks through four business MCP servers, a discovery registry, and human-approved actions.

A federal magistrate ordered Meta to identify by name the supervisors, managers and executives who authorized pirating books from shadow libraries such as LibGen and Anna's Archive for AI training, rejecting the company's claim that nobody with authority decided anything.

Octave built a shared multi-tenant AI foundation on Amazon Bedrock for its Reliance quality management platform, using Cohere Embed v4 and Bedrock AgentCore to embed 2 billion documents and power a Complaint Triage Assistant now running in production. The company says the foundation cut deployment time per use case by 50 percent.

Databricks is adding an agent memory capability to Agent Bricks, its agent development platform, aimed at letting agents retain context across interactions.

Mistral AI has released Voxtral 3, a voice agent stack combining a speech-to-text model, a text-to-speech model, and a LiveKit-based SDK, with transcription starting at $0.0035 per minute.

Meta's AI companion Muse reached 10.2 million U.S. downloads in its first 40 days, a milestone Sensor Tower says only 11 other apps have hit, and it retains users better than ChatGPT, Claude and Grok.

DeepSeek released DeepInfer, its internal inference stack, alongside a new 552B-parameter V4.1 Flash model under an MIT license, while the US restricted DeepSeek in federal civilian procurement and China's MOFCOM proposed pre-approval for sharing frontier models internationally.

Sam Altman reportedly presented OpenAI staff with three long-term governance scenarios for advanced AI, ranging from an IAEA-style international watchdog to dominance by a single company with enough power to shut off rivals' compute, an option he called "terrible."

Amazon Bedrock now serves OpenAI's GPT-5.6, GPT-6 and GPT-6.1 Sol models to applications running in the Canada (Central) and Canada West (Calgary) regions via cross-Region inference, which routes requests to US or global capacity pools depending on the chosen profile.

OpenAI says its GPT-6 models handle low-resource legacy languages such as COBOL, Fortran and Visual Basic far better, with GPT-6 Astra averaging an 87.4 percent test pass rate across 20 languages on a new MultiLangBench, up from about 38 percent for GPT-5.6 Terra.

AWS Agent Registry, now generally available, is a managed catalog of MCP servers, agents and agent skills that an orchestrator can search at runtime so new capabilities are usable after registration and approval rather than a code change and redeploy.

Amulya Vishwanath, founder of Techable Ventures and a former Nvidia product marketer, says physical AI is in a "LEGO era" of basic building blocks rather than awaiting a ChatGPT-style breakthrough, and that talent access is her firm's main edge.

Allen Institute for AI (Ai2) released FairStress, a benchmark that tests whether LLMs resist persuasive pressure equally regardless of which demographic group a challenge favors, not just whether they pick the same candidate across identity swaps. It has 300 scenarios expanded to over 13 million test cases, plus a 50K-item FairStress-Core subset, and is built into Olmo-eval.

A GitHub blog post profiles three Java Champions—Vlad Mihalcea, Patrick Baumgartner and Frank Delporte—describing how they use GitHub Copilot in production and open source, from cutting a web app build from weeks to two days to migrating roughly 900 teaching slides in eight to ten hours.

AWS and Keyfactor published a joint reference architecture that gives AI agents verifiable X.509 identities, using signed request envelopes, user-signed delegation, RFC 8693 audience-scoped tokens and real-time OCSP revocation checks. It targets the authentication gap in MCP, which defines no built-in agent identity standard.

The PyTorch Foundation added nine members on Oct. 20, 2026, including Platinum-tier Nebius, Nscale and Shopify, and detailed new backends bringing PyTorch to Arm Cortex-M CPUs, AWS Trainium, TPUs and IBM Spyre.