AI news for builders and product teamsUpdated Nov 18, 2026, 05:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Biohub, backed by Mark Zuckerberg and Priscilla Chan, is coordinating a $1.8 billion initiative to train AI models that predict cell behavior, with a first dataset expected in about a year. Meta, Google DeepMind, Isomorphic Labs, and the US Department of Energy are among the funders.

A North Carolina man was sentenced to 18 months in prison for using AI-generated songs and bots to steal over $8 million in streaming royalties, marking the first US criminal prosecution for AI-assisted streaming fraud.

Google has made its SynthID Detector public, letting anyone check whether an image, video or audio file carries a SynthID watermark. Google says more than 180 billion images and videos now carry the watermark.

Microsoft's Nvidia-powered Surface RTX Spark Dev Box is up for preorder at $5,999 and is slated to ship in November, priced above Nvidia's DGX Spark mini PC.

GitHub reports that one in three pull requests now involves an AI agent, up from fewer than one in 10 a year ago, and has built a fine-tuned classifier with Microsoft Applied Sciences that extends push protection to unstructured secrets, assessing candidates in under two milliseconds.

Gary Marcus criticized OpenAI's report on a new math result as too vague to assess, noting unknown procedures and failure rates, and said the system's generalizability beyond math is unclear. The essay promises a second section with Terence Tao's take.

Liquid AI released two open-weight decision models, d1-3B and d1-omni-600M (experimental), built on its Liquid Foundation Models. d1-3B scores 48.57 on the Decision Index 0.2.1 and answers questions in under 50 ms on measured edge devices.

Meta announced it took action against 33.2 million pieces of child sexual exploitation content on Facebook and Instagram in the first half of 2026, with over 97% found by its systems before users reported it. It is also rolling out new AI tools, including a large language model system to detect ad "signposting" that may direct users to illegal content elsewhere.

Databricks announced the general availability of building permission-aware Databricks Apps with on-behalf-of-user authorization. Databricks Apps lets developers build and deploy data and AI applications directly on the Databricks platform.

Microsoft Research Asia has open-sourced Agent Lightning v1.0, a Harnessed Agentic RL framework of roughly 3,500 lines that trains agents using their real production harness via an OpenAI-compatible LLM proxy. Its coding agent example lifted Qwen3.5-9B from 41.8% to 56.4% Pass@1 on SWE-bench Verified with about 6,000 training samples.

NVIDIA published a developer blog describing how its DSX Air node-based digital twin and Brev GPU compute can be combined with AI agents to validate AI factory infrastructure and software changes before production. The post also cites the NVIDIA AI Blueprint for Video Search and Summarization as a demonstration of the pattern.

OpenAI is adding college planning tools to ChatGPT for Teens, a mode introduced in August with safeguards and break reminders. The College Planner feature will gather application requirements, deadlines, tasks, and financial-aid steps, and is set to arrive soon.

An AWS Machine Learning Blog post introduces the Agentic Value Model, a framework for building business cases for agentic automation that captures four value dimensions the RPA-era ROI model misses. It also presents a prioritization matrix and cites three Amazon Quick Automate deployments at Kitsa, dLocal, and Genpact.

Qlik built Qlik Answers, a grounded enterprise AI assistant, on Amazon Bedrock for its 40,000+ customers, with general availability in February 2026. Qlik's Discovery Agent has surfaced more than 100,000 discoveries since launch.

AWS published a workflow combining AWS DevOps Agent, Amazon EventBridge, AWS Lambda Durable Functions, and Amazon Bedrock to turn incident investigation summaries into pre-validated remediation actions requiring a single human approval. Read-only actions run autonomously; mutating changes suspend until approved. A demo used a Lambda timeout incident, raising the timeout from 3 to 30 seconds.

NVIDIA introduced the mPDLP solver in cuOpt, distributing large linear programming problems across NVLink-connected GPUs with up to 6x lower peak memory per GPU versus single-GPU PDLP and LP problems capped at 2.1B nonzeros. Kinaxis and PSR reported speedups on large models.

An AWS Machine Learning Blog post describes a six-week internal program that paired non-engineering business professionals with mentors and production-grade AWS tools to build working AI agent prototypes. It reports before-and-after self-assessment gains and details one team's WealthWise multi-agent financial advisory prototype.

Cornerstone OnDemand built a multi-agent AI system called Orion AI on Amazon Bedrock and Strands Agents that cut database diagnosis time from 45 minutes to 10 minutes, a 78% reduction. A three-person team delivered the system in six months.

Meta's Muse iOS app, which launched nearly a month ago and became the top free app in Apple's App Store, now adds native iPad support in its latest update. A Mac version of the agentic AI tool was released about a week after the original mobile debut.

OpenAI announced public beta availability of a new Decisions API that classifies text and images about ten times faster than the Responses API, and reduced its paid API tiers from five to three.