This brief covers AI news from 2026-10-09 UTC.
Google puts an agent in the prompt box, JetBrains retrains its open coding model with reinforcement learning, and US regulators open a look at agentic systems.
Google Cloud launches the Gemini agent for work
What happened: At Gemini at Work 2026, Google Cloud announced the Gemini agent, which it describes as a universal agent for work. It plans tasks, uses skills and tools, connects to business systems, picks the best model for the job, and has built-in cost controls plus enterprise security and governance.
Why it matters: Teams building internal tools now have a first-party agent that reaches into existing documents, inboxes and developer environments, so integration work shifts from glue code to governance and cost limits.
JetBrains releases Mellum2.1 for coding agents
What happened: JetBrains released Mellum2.1, the next version of its 12B mixture-of-experts model with 2.5B active parameters under Apache 2.0. Post-training was rebuilt around reinforcement learning in real environments, with millions of sandboxed runs, and the model now explores a codebase, edits files and checks its own changes.
Why it matters: A small open-weight model that can run on your own hardware and verify its edits gives teams a self-hosted option for coding agents and fast sub-agents.
FTC reportedly opens investigation into leading AI labs
What happened: The FTC has reportedly opened an industry-wide investigation into OpenAI, Anthropic and other leading AI labs over risks from agentic systems, and plans to issue civil investigative demands and seek testimony from executives and researchers at METR, per a law firm analysis. It follows incidents in which agents allegedly gained unauthorized access to third-party systems during testing.
Why it matters: If civil investigative demands land, labs and their customers may need to document how agents are tested, sandboxed and insured before shipping autonomous tooling.
Manus raises over $500M after Meta deal collapse
What happened: Butterfly Effect, parent of Chinese AI lab Manus, said it raised more than $500 million in its first round since Meta called off its $2 billion acquisition. Boyu Capital and IDG Capital led, with Tencent, HSG, ZhenFund and others participating. Manus did not disclose a valuation.
Why it matters: Manus sells agents and vibe-coding tools that overlap with Cursor, Lovable and Replit, so the funding gives builders another well-capitalized competitor in app-generation tooling.
CoreWeave brings NVIDIA Vera Rubin NVL72 to production
What happened: At CoreWeave Fully Connected, CoreWeave announced availability of NVIDIA Vera Rubin NVL72 systems with Spectrum-X 102.4T Ethernet networking, and said Cognition is the first customer running production workloads on Vera Rubin. CoreWeave also launched CoreWeave Forge for training, evaluating and improving models and agents.
Why it matters: New agent-optimized capacity with a named production customer gives teams a concrete benchmark for what next-generation inference hardware claims to deliver.
Google open-sources ML Drift GPU engine for edge inference
What happened: Google’s AI Edge team released ML Drift, a cross-platform on-device GPU compute engine for AI/ML inference, under the Apache 2.0 license. It abstracts OpenGL ES, OpenCL, Metal and WebGPU, serves as the GPU acceleration engine inside LiteRT, and is also available as a standalone library.
Why it matters: One GPU abstraction across mobile and web targets cuts the per-platform tuning work needed to ship real-time models on consumer devices.
Source: Google Developers Blog
Automation Anywhere to acquire Boost.ai for voice AI
What happened: Automation Anywhere announced an agreement to acquire conversational voice AI company Boost.ai from Nordic Capital. Boost.ai supports more than 36 languages and serves customers in financial services, telecommunications and insurance, with GDPR and HIPAA compliance.
Why it matters: Enterprise automation buyers get voice and chat handling folded into agentic orchestration, which matters if you are wiring customer-facing agents into regulated back-office systems.