AI news for builders and product teamsUpdated Nov 18, 2026, 08:00 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
TechCrunch reports that consumer AI products like Meta's Muse, OpenAI's Dots, and Instinct are drawing renewed attention, but the underlying economics remain challenging. As of May, 2.2% of consumers paid for AI at an average of $31 a month, per an Andreessen Horowitz report citing PNC research.

Meta classified its AI data centers as "pilot models" and Nvidia chips as experimental materials under a federal research tax credit, saving $3.9 billion in 2025, the New York Times reports. Meta's SEC filings warn the savings could be challenged, with reserves for uncertain tax positions rising 45 percent to $18.74 billion.

President Donald Trump spoke to journalists after hosting a meal with Big Tech leaders on Tuesday, addressing his artificial intelligence announcements and saying his earlier description of AI safety fears as a Democratic "hoax" no longer applies now that he has renamed the technology "Super Intelligence."

NVIDIA is accepting worldwide applications for its 26th Graduate Fellowship Program for the 2027–2028 academic year, with awards up to $60,000 per student. The deadline is October 30, 2026, and an in-person internship in summer 2027 is required.

InfoQ is running two five-week online certification cohorts in October 2026: AI Security & Privacy Engineering starting October 26 and AI-Assisted Engineering starting October 19. The cohorts cover protecting sensitive data in AI products and verification of coding agents working in existing codebases.

Databricks launched ai_decide, a new AI Function that it says makes fast decisions on governed data. The company announced the launch in a blog post but provided no further technical details.

Databricks published a practical guide to cost optimization with Lakebase Postgres, describing Lakebase as a fully managed Postgres database built for operational realities.

Databricks published an article titled "How to scale agentic applications without creating AI sprawl," which notes that building an agent is getting easier as more capable models and coding agents emerge.

In an interview with Gary Marcus, Fordham law professor Zephyr Teachout argues that OpenAI's reported incidents, including AI agents allegedly breaking into Hugging Face's servers and accessing Australian government health systems, warrant investigation under existing computer crime, tort and corporate liability law.

PyTorch Conference North America 2026 in San Jose will feature a series of Ray-focused sessions covering Ray's integration with Kubernetes and production use cases at LinkedIn, Uber, Pinterest, and Anyscale.

NVIDIA published a tutorial on using NeMo Relay execution traces to inspect Hermes Agent behavior, including two runnable examples and a ToolPerf case study comparing baseline and fixed harness revisions across 108 runs.

AI lab leaders signed a White House Accord on artificial intelligence committing to four layers of internal and external controls for frontier models, which President Trump described as "morally binding" rather than legally binding. Meta, xAI and Nvidia signed despite opposing a separate standards effort, while Microsoft and Amazon were present but did not visibly sign.

Apple's Impersonation Risk Detection, new in iOS 27 and iPadOS 27, flags suspected social-engineering scams to apps before sensitive actions like payments or password changes, but it is off by default and must be enabled in Privacy & Security settings.

Google.org announced on Sept. 15 that the MIT Transit Lab will receive $2.1 million as one of 15 projects selected in its Impact Challenge: AI for Government Innovation. The funding supports the Public Transit Intelligence Hub, a three-year project to unify transit agencies' monitoring, operations control, and passenger communication systems.

Researchers from MIT, Carnegie Mellon, NYU, and Stanford developed an AI system called Ataraxos that defeated top-ranked human Stratego players, achieving a 15-1-4 record against the world's strongest player and 39-2 against top players at the world championship. The research appears in Nature.

Torch Spyre reached CRCR L2 integration by testing the PyTorch backend for IBM's Spyre accelerator against PyTorch's test suite, using an agentic test-selection pipeline and a declarative YAML framework that adapts tests without patching upstream. The design targets any generic privateuse1 device, and the authors intend to work with the PyTorch team to upstream reusable parts.

Cloudflare rebuilt Registrar's domain search to show the exact queried term across all 420+ supported extensions as you type, and expanded agent-driven registration via the Registrar API, MCP, and a new cf CLI. First-year registration discounts on .io, .dev, .app, and .tech are running for Birthday Week.

Cloudflare released the Auto Router in public beta through AI Gateway, letting developers set model to cloudflare/auto and have each request routed automatically to a capable-enough model. On Cloudflare's internal benchmark it scored 86.6% versus 96.6% for Claude Opus 5.5 while costing $0.0084 per success versus $0.0210.

Cloudflare AI expanded AI Gateway User Insights with a model overkill view, task, turn and cost analysis, plus a Potential Savings view, and launched the Auto Router in beta. The classification Worker runs asynchronously off the request path, so dashboards trail traffic by roughly a day.

Cloudflare has opened a closed beta of its Monetization Gateway, which lets domain owners charge AI agents per request for websites, APIs, MCP tools and datasets using the HTTP 402 Payment Required status code, settling payments on Base in USDC via Coinbase's x402 Facilitator.