Google Will Now Let Any AI Agent Run Your Smart Home

September 16, 2026
Artificial Intelligence
Comment off

Google began rolling out early access to Home MCP on 16 September 2026, a Model Context Protocol server that lets any MCP-capable AI agent enumerate devices, read live state, search event history and execute actions across the Google Home ecosystem. Google named Antigravity, Claude, ChatGPT, Hermes and OpenClaw. Access is limited to Google Home Premium Advanced subscribers in the US at $20 a month, setup requires a Google Cloud project with the Home APIs enabled, and the server enforces rate limits plus a block on sensitive actions such as unlocking doors.

anthropic drug discovery novo nordisk claude a microscope with angled eyepiece tube

Novo Partners With Anthropic to Speed Up Drug Development With Claude

September 16, 2026
AI News
Comment off

Novo Nordisk and Anthropic announced a collaboration on 16 September 2026 to speed up medicine development with Claude. Novo will first test Claude Science on specific R&D workflows and use Anthropic frontier models for AI-driven software development, with data governance and human oversight written into the arrangement. No financial terms or duration were disclosed. It is Novo’s third AI partnership of 2026 after OpenAI in April and AWS in August, and Anthropic’s second big pharma deal after Bristol Myers Squibb.

deepmind institute shane legg ai safety controls a escapement wheel with anchor lever lying flat v2

DeepMind Co-Founder Warns AI Capabilities Must Not Outrun Safety Controls

September 16, 2026
AI News
Comment off

Google DeepMind co-founder Shane Legg used the launch of the DeepMind Institute on 16 September 2026 to warn that rapidly advancing AI must not outpace its safety controls. The platform shipped five essays on day one, including a technical case for keeping chain-of-thought reasoning readable, an evaluation of eleven economic policies for an AGI labour shock, and a philosophical argument for a new utopianism. Legg also said it is premature to declare AGI achieved, while holding to his 50% by 2028 forecast.

jev model typesafe ai programmatic logic a sorting block with square recesses

ChatGPT Pioneer Launches Jev Model for Programmatic Logic

September 16, 2026
AI News
Comment off

TypeSafe AI has left stealth with $40 million led by DCVC and Jev, a model that does not chat, write code or generate a single word of prose. Founded by InstructGPT co-author Diogo Almeida, it returns typed decisions with calibrated probabilities in 70 to 500 milliseconds, at $0.042 per million input tokens with output billed at nothing. The speed and cost claims survive outside testing; the accuracy question does not get a clean answer.

hang ten systems vishal sikka 53m seed funding a surfboard lying flat with three fins

Former Infosys Chief’s AI Startup Hang Ten Systems Nabs Another $53M

September 16, 2026
AI News
Comment off

Vishal Sikka’s Hang Ten Systems has closed a second seed round of $53 million led by Xora, a Temasek-backed fund, taking total funding to $85 million just five weeks after the first round closed. Mayfield and Aramco Ventures both returned, and Yahoo co-founder Jerry Yang has joined the board. The four-month-old company claims multiple seven-figure contracts, a pipeline spanning 21 major enterprises and a 25-day deal cycle, on a headcount of roughly 20 to 25 people.

open source software ai strain maintainers a sawhorse trestle with plank

AI Is Reshaping Open Source Software and Straining the Systems That Sustain It

September 16, 2026
Artificial Intelligence
Comment off

A new ACM TechBrief argues that AI is improving open source software and overwhelming the volunteers who sustain it at the same time. The economic stake is enormous – firms would spend 3.5 times more on software without open source, a demand-side value of $8.8 trillion – yet 60% of maintainers are unpaid. curl closed a seven-year bug bounty after its confirmation rate fell from above 15% to below 5%, tldraw shut external pull requests, and Kubernetes chose disclosure plus personal accountability instead of a ban.

ssd failure prediction multiple instance learning a duffel bag lying flat with carry handle

New AI Method Improves SSD Failure Prediction From Imperfect Maintenance Data

September 16, 2026
Artificial Intelligence
Comment off

Data centres report whole racks as failed rather than isolating one drive, so failure-prediction models train on labels that are systematically wrong. A team at SEOULTECH, working with Samsung Electronics on real Alibaba Cloud drive data, grouped drives into failure bags and let a temporal convolutional network score them individually – holding F1 at 0.717 where a conventional model collapsed to 0.261.

firefox smart window mistral small 4 beta a casement window frame with two panes

Mistral AI Models Now Power Mozilla Firefox Smart Window Beta

September 16, 2026
Artificial Intelligence
Comment off

Mozilla and Mistral announced their browser partnership on the same morning in two posts that do not say the same thing. We read both against the Mistral Small 4 model card, Mozilla’s support documentation, published list prices and StatCounter’s browser share – including what “zero data retention” actually commits each company to.

Apodex 1.1 mini - apodex 1 1 mini frontieragent framework release a relay baton lying flat on two rests

Apodex Releases the FrontierAgent Framework and the 35B Apodex 1.1 mini

September 16, 2026
AI News
Comment off

Apodex released a 35B open-weight sibling to its 397B flagship, plus FrontierAgent, an Apache 2.0 agent runtime with ReAct and Agent Team modes. We work from the Hugging Face config, the GitHub repository and the arXiv report rather than the announcement — including the real parameter count, which benchmark scores belong to which model, and why the quantised builds are outrunning the originals.

$deep think mathematica google internal math model a slide rule lying flat with centre slider$

Google Tests a New Deep Think Mathematica Model Internally

September 16, 2026
AI News
Comment off

A leaked Gemini API descriptor names an internal Google build called Mathematica (DeepThinkV3) Teamfood Raw Thoughts, with a one-million-token window, an UNSTABLE_EXPERIMENTAL stage and unsummarised reasoning. We decode the identifier field by field, identify the open number-theory problem in the leaked trace, compare it with the public Gemini 3.1 Deep Think, and separate what the screenshots show from what has been read into them.

Google Will Now Let Any AI Agent Run Your Smart Home

Novo Partners With Anthropic to Speed Up Drug Development With Claude

DeepMind Co-Founder Warns AI Capabilities Must Not Outrun Safety Controls

ChatGPT Pioneer Launches Jev Model for Programmatic Logic

Former Infosys Chief’s AI Startup Hang Ten Systems Nabs Another $53M

AI Is Reshaping Open Source Software and Straining the Systems That Sustain It

New AI Method Improves SSD Failure Prediction From Imperfect Maintenance Data

Mistral AI Models Now Power Mozilla Firefox Smart Window Beta

Apodex Releases the FrontierAgent Framework and the 35B Apodex 1.1 mini

Google Tests a New Deep Think Mathematica Model Internally

Links

Newsletter

Contact

AI News

Google Will Now Let Any AI Agent Run Your Smart Home

Novo Partners With Anthropic to Speed Up Drug Development With Claude

DeepMind Co-Founder Warns AI Capabilities Must Not Outrun Safety Controls

ChatGPT Pioneer Launches Jev Model for Programmatic Logic

Former Infosys Chief’s AI Startup Hang Ten Systems Nabs Another $53M

AI Is Reshaping Open Source Software and Straining the Systems That Sustain It

New AI Method Improves SSD Failure Prediction From Imperfect Maintenance Data

Mistral AI Models Now Power Mozilla Firefox Smart Window Beta

Apodex Releases the FrontierAgent Framework and the 35B Apodex 1.1 mini

Google Tests a New Deep Think Mathematica Model Internally

Tags

Links

Newsletter

Contact