AI news for builders and product teamsUpdated Nov 17, 2026, 21:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
AWS now lets new SageMaker AI customers sign up with Google, GitHub, Apple, Amazon or email and land in a preconfigured SageMaker Studio in 2-3 steps, with $100 in Free Tier credits and no credit card for most.

Databricks says its CustomerLake agentic customer data platform has entered beta, with Campaign Agents among the tools leading enterprises are adopting as they go live on the platform.

Microsoft launched Guardian Agents, a monitoring feature within its Project Perception product, to watch the AI agents enterprises deploy, arguing that frontier models from any provider are too unpredictable to trust. It also added agent discovery to Microsoft Defender covering up to 60 agent types, including Claude Code, Codex and OpenClaw.

Databricks argues that open table formats alone do not make a data catalog open, positioning its Unity Catalog as advancing an "open lakehouse" by addressing the governance layer above storage formats.

Amazon SageMaker JumpStart onboarded more than 100 new models from 30 providers between January and September 2026, with Alibaba's Qwen line contributing 29. AWS also added a Model Hub on the SageMaker AI console, four named deployment profiles, and a Hugging Face deep link that opens a pre-configured Studio environment.

A US magistrate judge ordered a former OpenAI engineer, Abdulrahman Alfozan, to remain detained pending trial on charges of secretly recording OpenAI meetings, citing flight risk and evidence of ties to Saudi Arabia. The FBI allegedly found a "Saudi Gov Docs" folder on his iPhone and photos of a second planted bug.

AWS released TorchNeuron, a tool that lets PyTorch applications run on Amazon's Trainium chips by changing a single line of configuration, with no code rewrite. Amazon is also contributing upstream changes to make PyTorch less tied to Nvidia hardware.

dbt is widely used for data transformation, and Databricks is outlining how teams can get more from running it on its platform. The available material is a headline and short teaser only, with no concrete guidance details.

OpenCode, an open-source AI coding agent, disclosed that attackers breached its Console platform overnight Saturday, affecting 1.4% of users whose stored API keys were stolen. Some users faced unauthorized charges in the hundreds or thousands of dollars, and AI providers blocked or refunded affected keys.

Cloudflare cut the average cost of its internal AI code review from $2.43 to $0.77, a 68% reduction, while the median review time fell from 4m 09s to 1m 32s; roughly 40% of the savings came from code changes and the rest from model and effort tuning.

Tolan, an AI companion app with a cartoonish character that helps users manage daily tasks and emotions, raised $30 million from Union Square Ventures, Khosla Ventures and Homebrew, doubling its total funding. The startup reports more than 250,000 active users, 70 million conversations over the past year and annual recurring revenue in the double-digit millions.

Eugenia Kuyda launched Wabi, a free personal AI assistant that runs as a messenger app with multiple shareable agentic threads, each able to host custom mini-apps the agent builds automatically. It follows her Replika exit and echoes the concierge chatbot she pitched at Disrupt twelve years ago.

A surge in personal AI agents, led by apps such as OpenAI's Dots and Meta's Muse, is dividing the web into a third category of visitor that many sites are blocking outright while others, like Shopify, build agent-only checkout lanes.

MiniMax released M3.1 Flash, a reasoning model scoring 42 on the Artificial Analysis Intelligence Index, up 13 points from M3 and about half the cost per task, at roughly $0.22 per task. API pricing is $0.15 per million input tokens and $0.60 per million output tokens, with open weights planned for October 23.

MIT, Boston Children's Hospital and Inria researchers built E(3)-Pose, a deep learning model that tracks a moving fetal head during MRI so 2D slices can be assembled into a usable 3D image. The work was presented at the 2026 European Conference on Computer Vision.

Cloudflare has launched a preview of Clairic, a System One extraction model hosted on Workers AI as @cf/cloudflare/clairic, which returns typed values, spans, estimates and forecasts in a single pass without generating tokens. Open weights are planned soon.

Nick Cammarata, a researcher at the AI startup Goodfire, published a 26,000-word blog post arguing that neural networks may be far more legible than the field assumed, based on re-examining a 2020 result in the image model Inception v1. Critics including Stanford's Chris Potts and Google DeepMind's Neel Nanda call it an interesting hunch rather than a breakthrough.

Kling AI has released version 4.0 of its video generator, adding single-pass clips up to 30 seconds, input of up to ten keyframe images, 8,000-token prompt comprehension, stereo audio and lip sync in more than ten languages, plus a cheaper 720p Flash variant.

Jane Street machine learning engineers implemented Mamba-2's forward pass in NVIDIA's cuTile, a Python tile-level GPU DSL, and worked with NVIDIA to optimize three custom kernels. The post walks through the decay-cumsum, state-jump and fused output kernels plus compiler hints used to reach Tensor Cores.

MapQuest has added a surveillance layer to its app that shows the locations of known outdoor license plate readers from Flock Safety and about a dozen other vendors as small red eye icons with a vision cone for each camera's facing direction.