From the archive
Follow the latest AI news and artificial intelligence developments across models, agents, robotics, research and the wider industry. Explore new AI tools, advances in computing, business moves, safety and policy, with clear reporting and useful sources. Read the stories, catch up with News in Brief and join the conversation.
Lead story
Meta has released open-source software that lets developers connect its Muse AI agent to DIY hardware, from Raspberry Pi computers to displays, speakers and sensors. The move gives makers a way to put Muse into physical devices without waiting for Meta to build them all itself.
Read the report & discussion ↗Mistral AI Watch
Morocco’s digital-transition ministry has released two AI tools for Moroccan Darija, the first concrete output from its partnership with Mistral. One identifies Arabic dialects; the other transcribes Darija speech, including conversations that switch between Arabic, French and English, according to iAfrica.
Watch Desk
Adding agents to an AI system can add voices without adding much independent evidence. A Hugging Face community essay argues that once agents read and revise one another’s answers, agreement may reflect one early view spreading through the group, rather than several independent judgements reaching the same conclusion.
Google DeepMind Watch
Google is giving a limited group of cybersecurity partners access to Gemini 4 Argon, a model it says can find, validate and patch software vulnerabilities. The security trial is the most consequential detail: Google is proposing AI not just as a coding assistant, but as a participant in…
Watch Desk
The US government’s America.gov chatbot launched on 29 September, drawing on federal websites to answer public questions. A Tech Insider article says it uses Google’s Gemini and xAI’s Grok, but notes that coverage has not settled which models are involved.
OpenClaw Watch
OpenClaw has released v2026.9.8, adding GPT-6.1 Sol support, reducing memory use and fixing update and Windows startup issues. It is a compact release, but its changes reach model choice, agent replies and day-to-day reliability.
Watch Desk
A new Hugging Face community guide explains how to build AI workflows around bounded decisions, and why OpenAI’s Decisions API should not be confused with an independent service using a similar name. Its most useful advice: treat model output as a suggestion to validate, not permission to take…
Google DeepMind Watch
Google DeepMind has developed SynthID Bio, a proof-of-concept watermark for AI-generated protein sequences and predicted structures. The idea is to add a detectable provenance signal without obviously compromising tested protein function, though the researchers also identify ways the mark can be lost.
AMD Watch
AMD has reportedly agreed to buy World Labs, the spatial-AI startup co-founded by Fei-Fei Li, in an all-stock deal valued at about $8.2bn. The reported move would bring a team working on AI models of three-dimensional worlds into a chipmaker’s effort to anticipate what computing comes next.
Microsoft AI Watch
Microsoft has proposed a code of conduct for its in-house AI models, including expectations that they accept correction, interruption and shutdown. The draft is open for public consultation and could guide model development from 2027, but Microsoft says it is not currently being used to train its models.
Watch Desk
Supabase has raised $150 million in a round led by Singapore sovereign wealth fund GIC and agreed to acquire Turso, a database company whose product is designed for AI agents. The deal joins a sizeable funding round with a strategic purchase in the infrastructure layer agents need to…
Higgsfield Watch
Higgsfield has published a guide to connecting its video-generation tools to n8n and Make, with an API covering more than 50 video and image models. For agencies, the useful detail is that the guide explains two routes, their different billing, and how to get completed videos back into…
Tavus Watch
Real-time video agents have to listen, reason and respond without leaving an awkward silence; Tavus says splitting those jobs into separate services can help. Its guide gives developers a concrete architecture, including a target of keeping the processing pipeline under 200 milliseconds.
Tavus Watch
Tavus has published a practical guide to creating an AI clone from about two minutes of video, including how to record the footage and set limits on what the clone can do. The useful takeaway is that the recording is only the start: what the system knows, when…
Tavus Watch
Tavus has published a practical guide to building AI product demos that answer buyers’ questions during a live conversation. Its most useful advice is about setting boundaries before the avatar is built: decide which questions it should answer, which answers it must avoid, and when it should hand…
Tavus Watch
A person reviewing an AI’s work is a copilot; an AI acting on a handed-off goal is an agent. Tavus argues that conversations which are themselves the task, such as intake or screening, need a different design: a system that stays present throughout the exchange and knows when…
Tavus Watch
Tavus draws a useful line between a live digital twin and an AI replica built from recordings, then describes how its conversational PALs use real-time perception to respond. The distinction matters for buyers: one stays synchronised with a real-world counterpart; the other does not.
Watch Desk
Prime Intellect has launched Prime Inference, a platform for serving open AI models with serverless endpoints or reserved GPU capacity across multiple data centres. Its pitch is about the less glamorous, rather important business of keeping models available and agent tools working reliably.
