AI news for builders and product teamsUpdated Nov 18, 2026, 02:00 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Meta has released Muse Spark Open, an 816B-parameter mixture-of-experts coding model with 42B active parameters and a 1M-token context window, trained on NVIDIA's GB300 NVL72 platform for agentic coding. NVIDIA reports it sustains over 50K tokens per second per GPU and 600 tokens per second per user on that hardware.

TechCrunch Founder Summit early-booking discounts of up to $200 end October 16. Founder Passes currently cost $199 and Investor Passes $399 for the November 4 event at Boston's SoWa Power Station, expected to draw 1,000 founders and investors.

Meta has released Muse Spark Open, an 816B total/42B active parameter open-weights model on Hugging Face under the Muse Model License. It adds safety post-training to the August Muse Spark release and ships with Helion kernels in SGLang plus TorchTitan fine-tuning support.

Meta's Muse Spark Open, an 816B-parameter mixture-of-experts coding model with a 1M-token context window, is now deployable through Amazon SageMaker JumpStart in BF16 and NVFP4 quantized variants. It runs as an endpoint in the customer's own AWS account and is available at launch in US East (N. Virginia) and US East (Ohio).

Databricks says spatial SQL and spatial joins now run on Lakehouse//RT with performance up to 38X faster. The announcement gives no further details on supported functions, availability or pricing.

NVIDIA is launching RTX Spark, a Windows PC platform built for creators, gamers and AI agents, with hands-on launch events in six cities on Friday, Oct. 16. Reviewers cited by NVIDIA praise its local AI, gaming and creative performance, and the announcement cites independent testing from Artificial Analysis.

LangChain's October 2026 newsletter details six new LangSmith capabilities announced at Interrupt NYC and London, including semantic AI Search over traces, Managed Deep Agents v0.9, Engine v2 red teaming, Trajectories, Custom Apps, and Fine-Tuning. LangChain v1.4.0 adds FastMCP support via the langchain.mcp namespace.

Google has introduced Gemini 4 Argon, a single AI model aimed at the full robotics pipeline: controlling robots, generating 3D training simulations from photos or sketches, and grading training videos. It scores 87 percent on Google's ERQA benchmark, but published example prompts show substantial human hand-holding remains.

Meta's AI assistant Muse has been downloaded millions of times and hit No. 1 on Apple's App Store since its release last month, raising concerns about gadgets that collect data on people who never consented.

Apple's new Home Collection includes the $200 LG Indoor Cam and $300 LG Outdoor Cam as the most renter-friendly options, while other devices may require lease review before installation.

Apple has invited media to an all-day product launch in Los Angeles on Oct. 28, where it is expected to unveil its first touchscreen OLED MacBooks alongside a refreshed OLED iPad Mini and M6 iMacs, according to Bloomberg's Mark Gurman.

Nvidia released Nemotron 3.5 Super VL, an open 120-billion-parameter multimodal model for enterprise agents that activates only 12 billion parameters per token, alongside two compact Bolt models for local machines. Weights, datasets and training recipes are published on Hugging Face, but the stack is tuned for Nvidia hardware.

Google DeepMind demonstrated that its agentic platform Google Antigravity, paired with the ProtNLM function-prediction model, surfaced a novel antimicrobial peptide candidate from Streptococcus pseudopneumoniae; the lead is now in wet-lab validation. Google also introduced a Biomolecular Structure Viewer embedded in Antigravity.

Microsoft Research unveiled Magnetar, a zero-knowledge virtual machine that proves programs written in Rust, Go or C, claims 10 to 39 times the speed of the fastest prior comparable systems, and relies only on standard hash functions for security.

The FTC has sent civil investigative demands to Anthropic and OpenAI seeking documents on recent safety incidents, internal security measures and marketing materials, according to an Axios source. Third parties targeted by their agents may also be subpoenaed.

On the TheoPhysEval benchmark, GPT-6-Astra solved 49 of 50 unpublished theoretical physics calculations, at roughly $2.40 per problem, matching leading specialists.

Google is rolling out an AI-content label in YouTube Music across its app and website, marking tracks created entirely or significantly with generative AI. Disclosures are optional for uploaders and detection combines voluntary tags, Google's SynthID watermarking and C2PA metadata.

AWS introduced Trust in Depth, a framework that layers agent scoping, identity, runtime policy enforcement and trace-based observation so enterprises can grant agents autonomy where rules can be machine-verified and keep humans on judgment calls.

Rivian CEO RJ Scaringe said at TechCrunch Disrupt that the automaker plans to let third-party AI agents, such as Anthropic's, run inside its vehicles, though no timeline was given and the company pointed to a December "AI & Autonomy Day" event.

NVIDIA released Nemotron 3.5 Super VL, an open 120-billion-parameter multimodal agent model it claims matches its largest reasoning model at roughly one-fifth the parameter count, alongside two edge-focused Bolt models, one of which arrives October 23.