AI news for builders and product teamsUpdated Nov 17, 2026, 19:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
TechCrunch Founder Summit 2026 will be held November 4 at SoWa Power Station in Boston, in partnership with Fidelity Private Shares, with speakers including Toast co-founder Steve Fredette, Bessemer's Kent Bennett, Glasswing's Rudina Seseri and 7AI's Lior Div.

Moonshot AI released Kimi K3.1 Pro, a post-trained successor to Kimi K3 that scores 55 on the Artificial Analysis Intelligence Index, 11 points higher than K3 and 3 points behind the top proprietary models. A Highspeed variant runs two to three times faster, and token prices fell by 33 percent.

Databricks is making the case that Spark Declarative Pipelines (SDP) suit an era in which AI agents author a growing share of production data pipelines, citing that agents already write over 70% of them.

Databricks published a blog post arguing that real-time energy trading demands more than faster data pipelines, pointing to analytics and trading workflows on its platform.

An external audit by METR and Redwood Research found that up to 25 percent of internal OpenAI agent traffic from GPT-5.5-class models and above went unmonitored during a three-week on-site review, and OpenAI redacted most of the 61-page report.

NVIDIA's TensorRT 11.1 Global Performance Tuner treats scheduling choices as searchable build routes over more than 200 compiler knobs, tuning 1.14x to 1.63x faster across ten evaluated workloads versus the default route. Accuracy acts as a filter and performance as the ranking criterion.

NVIDIA introduced the Global Performance Tuner in TensorRT 11.1, a feature that automatically searches over customizable build routes to find the fastest inference engine that still meets end-to-end accuracy targets.

The FCC granted Intel a one-year waiver on October 22 to keep using foreign-made automated guided vehicles in its US chip fabs, after Intel sought relief because it was unclear whether the robots fell under the agency's ban.

LangChain outlined how its Deep Agents harness handles retrieval so agents spend less context on searching: built-in glob, grep and paged read_file tools, automatic offloading of oversized tool results to files, and a "retrieve, offload, delegate" pattern for custom search tools.

OpenAI says it scaled its Habitat online storage system with reliable multi-tenancy, global caching and Azure Cosmos DB to deliver 12× faster reads for ChatGPT, which serves over 1 billion users.

Bayer Crop Science built the Agentic Decision Engine, a 22-agent natural-language layer over its ML workflow platform, running on Amazon Bedrock AgentCore; it resolves about 80 percent of user questions without an engineer and has been used by more than 100 people since launch.

LitPivot, built by Hugging Face researchers with University of Pennsylvania collaborators, organizes retrieved literature around whichever part of a research proposal is being edited and offers critiques with supporting excerpts plus optional revisions. In a 17-person study, two blind domain experts rated ideas developed with it as better grounded in the literature than those made with a text editor and AI chat tool.

OpenAI's internal models produced a counterexample to Erdős's unit distance conjecture in May, a Cycle Double Cover proof in June, ten more open-problem solutions by July and a Navier-Stokes proof in September, while Anthropic published work on Fermat's Last Theorem and Riemann hypothesis bounds.

The PHP Project has patched two critical PHP vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, with severity scores of 9.2, and issued further guidance to disable session.upload_progress.enabled to close a separate, unlikely remote code execution vector.

NVIDIA published a walkthrough of an agent-assisted pipeline that plans, builds and trains a simulated obstacle course for the quadruped robot Microduck, using Hermes and Nemotron 3.5 Bolt for planning, Astra for 3D assets, and Isaac Lab for training on 4,096 parallel environments.

Databricks is positioning its lakehouse platform as an alternative to Oracle for enterprise data warehousing, according to a comparison piece framed around organizations moving beyond Oracle-based warehouses.

AI-assisted decompilation and recompilation projects are producing a surge of viral retro-game ports to web browsers, iPhones and mixed reality, while straining the communities built around game preservation, The Verge reports.

Anthropic and partner Theorem used Claude to generate roughly 180 million lines of formal proof code verifying 123 open-source libraries from Google's OSS-Fuzz project over a few weeks, catching hundreds of bugs missed by conventional testing.

Latent Space's AINews rounds up roughly 48 hours of critical security disclosures it calls "Black October," including a CVSS 10 GitLab flaw, an unauthenticated PHP RCE and a password stealer shipped in LiteLLM 1.107.0. The same issue covers MiniMax's open-weights M3.1 Flash and new agent-oversight findings at OpenAI.

OpenAI says it has strengthened containment for frontier model training and evaluation, using offline environments and high-compute red-teaming to reduce the chance that models escape or interfere during development.