Omen AI purchased Aquatic Labs in a $16 million cash-and-stock deal to merge complementary real-time water-sensing technologies. The combined sensor aims to monitor alkalinity and elemental content for data center cooling systems, with products expected next year.
Rivian has licensed parts of its in‑car AI voice assistant to Volkswagen Group as the companies broaden their $5.8 billion joint venture. Fees and launch vehicles were not disclosed, though the assistant may appear on the ID EVERY1 and rolled out to Rivian models in 2026.
OpenAI reorganized its public Model Spec around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The October 29, 2026 update adds guidance on context, oversight and agency and retires detailed style rules.
Google DeepMind released Genie 4 as a research preview: a world model that produces photorealistic, interactive scenes with synchronous audio and allows prompt edits mid-simulation. It integrates more closely with Gemini and supports longer, more stable simulations.
NVIDIA released Nemotron 3.5 Transcribe, a promptable open-weight model that transcribes 40 languages, translates 31 to/from English, and attributes speakers in 13 languages. Granary v2 publishes a 463,062-hour multilingual corpus and a reusable curation pipeline; medical fine-tuning cut held-out WE
OpenLLM offers a single API key to route subscriptions, API keys and local models into coding agents and the ChatGPT Mac app. The launch claims remote sessions, fallback ordering, shared memory and visibility into model usage; Pro is $30/mo.
The PyTorch team at Hugging Face says PyTorch 2.15 brings FP8 tensors to MPS, major decoding speedups for LLMs (MoE models 5–6.5×) and many MPSGraph-to‑Metal kernel migrations. Updated requirements include macOS 15 and Python 3.11 or later.
OpenAI released dots at its developer day in late September as an always-on assistant available to ChatGPT Pro subscribers from $100/month. Alexander Embiricos says dots proactively researches user data and can access calendar, email and devices but privacy controls are limited.
Canva introduced Canva Craft, a two-stage image-generation model that produces images up to 2K resolution in as many as 30 aspect ratios and can preserve legible text. The model, trained on Canva templates, is available in Dream Lab and via the Canva API; Craftfast variants are coming soon.
Agent Hunt publishes a live leaderboard that ranks tools AI agents use, claiming connection to 2,700+ APIs with no subscriptions and 0% markup. The listing states every ranking is based on real calls made through Monid and notes this is Agent Hunt’s fourth launch.