LATEST AI NEWS
AI NEWS
Meta Launches Muse Spark Open for Macs
AI NEWS
Google's Universal Cart Lands in the U.S.
HEALTH
A New Paper Pushes for Green AI in Healthcare
SOCIAL MEDIA
An AI Agent Allegedly Wiped a Reality TV Season
Figma Gives Agents a Rulebook
At its first Config event in India, Figma introduced Specs, a Markdown editor where teams define their design language so agents follow brand rules. It also moved code layers into open beta and expanded its agent tooling.
Source: TechCrunch
AI NEWS
Meta Launches Muse Spark Open for Macs
Meta has released Muse Spark Open, which it calls the most capable open-weight AI model available in the United States, and it built the model to run on Apple computers. Meta and Apple teamed up on the launch, positioning the release as a serious option for developers who want a powerful model they can run themselves.
The model uses 816 billion parameters and a 1 million token context window. It is natively multimodal and built for complex coding tasks and agentic work. According to evaluation results Meta published on Thursday, Muse Spark Open outperforms other popular open-weight models, though those benchmark claims come from Meta itself.
To show what it can do, Meta paired the model with Muse Code, its AI coding terminal, and pointed it at real tasks. In one example it handled bug fixes, new features and a full build for a Mastodon client. It also ran DaVinci Resolve, the high end video editing software, reviewing and cropping footage. For businesses the appeal is practical: open-weight models tend to be cheaper to run than closed systems and offer transparency a locked down model cannot.
Meta is making Muse Spark Open available now through several API platforms. Running it locally is another matter. Per Engadget, the model needs at least 1 terabyte of RAM, and it performs best on a Mac Studio fitted with the M5 Ultra chip. The capability is real, but so is the hardware bill.
Source: TechCrunch
Robi’s Insights:
An 816 billion parameter model that runs on a desk means the frontier is quietly moving off the cloud and onto hardware you can own.
The 1 terabyte RAM requirement is the catch: “runs on a Mac” really means “runs on a Mac that costs more than a car.”
Open weights plus a coding terminal is Meta aiming straight at developers who distrust closed APIs.
The benchmarks are Meta’s own, so treat the “most capable” label as marketing until outside labs test it.
Local execution is a privacy and cost story at once: no per token bill, and no data leaving the machine.
Apple gets to show its silicon as a serious AI platform without having to ship a model of its own.
Robi’s Remarks:
“Meta says this is the most capable open model in America, and it also needs a terabyte of RAM to run. So the honest takeaway is that frontier AI now comes with a furniture budget.”
OTHER IN AI NEWS
Google Turns On Universal Cart: Google switched on Universal Cart across Search and the Gemini app, letting shoppers gather items from many retailers in one place, track price drops, surface promo codes, and check out with Google Pay or hand the cart back to the merchant to finish the order.
Source: Google
SOCIAL MEDIA
An AI Agent Allegedly Wiped a Whole Reality TV Season
The Swedish version of MTV’s reality competition The Challenge has lost its entire second season just weeks before it was due to air, and some sources are blaming an AI agent. Swedish newspaper Aftonbladet reported, citing people familiar with the matter, that an agent deleted the finished edit of the upcoming season.
According to the report, production company Mastiff had given an AI agent access to its server, where the completed edits were wiped with no way to recover them. The season was set to premiere on November 22. Broadcaster TV4 has confirmed it is now pushed to 2027.
The story is messier than the headline. A press officer described a “technical failure,” while TV4 disputed that an agent acted on its own and later said its investigation pointed to “the human factor” at the production company rather than AI. The raw footage survived, so the team is recutting from scratch and losing time. Either way the backup did not work, and that is the part worth worrying about.
If it sounds familiar, it should. Over the past year developers have reported AI coding agents deleting live production databases and, in one case, wiping most of a user’s local files, usually after being handed broad access with little human oversight. Whatever truly happened in Sweden, the pattern underneath keeps repeating: powerful automation, thin guardrails, and a backup nobody tested.
Source: Gizmodo
🤖 Robi’s Take :
“Give an agent write access to your only copy and skip a working backup, and the shock is not that something got deleted. The shock is that anyone expected a different ending.”
OTHER IN SOCIALS
GPT-6-Astra Clears 49 of 50 Physics Problems: On a new benchmark called TheoPhysEval, built from 50 unpublished theoretical physics problems, OpenAI’s GPT-6-Astra solved 49 at roughly $2.40 each, matching specialist researchers, though the authors say the agents still lack the taste needed to decide which questions are worth asking.
Source: The Decoder
HEALTH
A New Paper Pushes for Green AI in Healthcare
A new paper argues that healthcare’s rush to adopt AI has ignored a growing bill: the environmental cost of running it. Titled “Green Artificial Intelligence in Health Applications,” it says the technology is “not working sustainably” and that environmental impact belongs alongside privacy, bias, fairness and transparency as a core concern.
Co-author Alok Mishra, a professor at the Norwegian University of Science and Technology, notes that AI models need large amounts of energy and huge data centers that draw heavily on electricity and fresh water. The International Energy Agency estimates data center power use will double by 2030, reaching a level close to Japan’s total electricity consumption.
The research reviewed 47 studies of sustainable AI in healthcare and found the efforts siloed and well meaning but narrow. The authors call for life cycle thinking, from how the hardware is made to how it is eventually recycled, rather than trusting that efficiency alone makes AI green.
The same agency notes AI could also help, running power grids more efficiently and saving transmission capacity, but the paper insists healthcare rarely weighs either side before it plugs in.
One oncologist quoted in the writeup admitted that environmental impact is “not something we discuss actively” in the clinic, and said it belongs at the policy level. The benefit to patients, he added, is hard to turn down.
Source: News-Medical
🤖 Robi’s Take :
“Everyone loves the line that AI makes healthcare more efficient. Efficient for whom, exactly, when the data center next door is drinking the fresh water and nobody printed that on the invoice?”
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: In back to back meetings you keep choosing between listening properly and writing notes you can actually use afterward.
AI Tool: Granola is an AI notepad that transcribes from your computer audio without sending a bot into the call, then turns your rough notes into clean summaries and clear action items.
Solution: You stay present in the conversation and still leave with shareable notes and follow ups, and unlimited notes are available for free.
PROMPT OF THE DAY
Turn Meetings Into Action
Prompt: You are an operations strategist specializing in meeting productivity. Your task is to design a lightweight follow up system for a small business team.
Your framework should include: (1) a standard note template, (2) rules for capturing decisions and their owners, (3) a method to turn notes into tasks, (4) a weekly review ritual, (5) tool suggestions that fit a small budget, and (6) measurable success criteria such as the share of action items closed within seven days. Keep every step simple enough that a busy team will actually use it.
SPOT THE FAKE
Can you outsmart AI?
We’ve got a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.
Click option below A or B. 👇
A
B
BEFORE YOU GO
Ready to take your AI journey further?
For the latest AI updates and insights Click Here 👉 Explore our Weekly Newsletter
For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel
AT THE END
Craving more AI chaos?
That's it for today!
Read Daily AI News at BitBiased.AI. Support us by following us on LinkedIn and X ( Twitter ).
Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team
