Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},Meta built an AI model that runs on a Mac, and the catch is that it wants a full terabyte of RAM to do it. After that, we’ve got a couple of thoughts on picking AI tools that last, some handy new tools, and one AI experiment where reality slowly falls apart. Let’s dig in.
🤿 DEEP DIVE
Meta's AI Has a Mac Problem
Powerful enough to edit videos. Hungry enough for 1TB of RAM.
Meta has built an AI model that runs locally on a Mac. Which sounds pretty great, until you find out how much memory it needs.
A whopping 1 terabyte of RAM.
The model is called Muse Spark Open, and Meta describes it as the most capable open-weight AI model in the US. It has 816 billion parameters and can handle up to 1 million tokens of context.
But the more interesting part is what Meta is actually doing with it.
This isn't just another AI chatbot. Muse Spark Open is designed to handle complex tasks on your computer, including writing code and working with creative software.
Meta tested it alongside its Muse Code terminal, where it fixed bugs, added features, and completed a full build of a Mastodon client. It also demonstrated the model reviewing and cropping footage in DaVinci Resolve.
So we're talking about AI that can actually work inside desktop applications, not just tell you how to use them.
And because it's an open-weight model, developers get more flexibility and transparency than they would with a fully proprietary system.
Of course, there's still that small issue of needing 1TB of RAM. Meta says it's best suited to an M5 Ultra Mac Studio, so this isn't exactly something most people will be running on their everyday MacBook.
The model is also available through API platforms, meaning local installation isn't the only way to access it.
The quick take
816 billion parameters and a 1 million token context window.
Built for coding, desktop automation, and multimodal tasks.
Demonstrated fixing software and editing video footage.
Running it locally requires at least 1TB of RAM.
Available through multiple API platforms.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Everyone’s Buying AI Stocks.
Few Understand What Powers Them.
The AI boom isn’t just about chips.
Every model needs cooling.
Every server needs memory.
Every system depends on high-speed networks.
That’s where the real bottlenecks are forming — and where a different group of companies is quietly winning.
This briefing breaks down the infrastructure behind AI — and 3 names positioned to benefit from every dollar flowing into the space.
By clicking, you’ll receive this and other Elite Trade Media LLC financial newsletters, which may include advertiser offers. Privacy · Unsubscribe anytime.
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about
It is a useful reminder that interesting software does not always come from adding more features. Sometimes the better idea is noticing something distinctive about the hardware and designing around it. The fact that an AI-assisted hackathon produced the prototype makes the speed from idea to working interaction interesting too.
This has a practical consequence for anyone trying to choose an AI tool at work: “Which model is best?” may increasingly be the wrong question. The useful skill becomes building workflows that can survive the model underneath them changing every few weeks.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
📬 Talanoa
Make your email inbox feel more like a chat app.
Why it’s useful: Handy if you're tired of digging through messy email threads and want to manage replies, tasks, and calendars in one place, without sending your data to the cloud.
⚖️ GLM-Legal
An AI model built to handle the more tedious parts of legal work.
Why it’s useful: Worth a look for legal teams building tools to review contracts, analyze evidence, or research compliance requirements, though API access is still limited to beta applicants.
Control your smart home without reaching for your phone.
Why it’s useful: Could be a handy central screen for managing connected devices, making FaceTime calls, and using Siri around the house.
👀 ON OUR RADAR
One more thing we thought you'd like
Your Search Bar Is Haunted
Shoppers type exactly what they want and get “0 results found” back. They don’t complain. They just vanish to a competitor.
Gladly understands what shoppers mean, typos and all, and turns dead-end searches into product recommendations.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
Give an AI the same fruit photo 300 times and watch reality fall apart
Kenan Tang asked Google's Nano Banana Pro to copy the same fruit photo 300 times, feeding each new copy back into the AI. Tiny mistakes slowly compound until the fruit platter completely collapses into black-and-white noise while another AI agent keeps confidently declaring the copies accurate. It's both a strangely satisfying visual experiment and a clever demonstration of how autonomous AI errors can snowball.
AI used
Google Nano Banana Pro, operating as an agentic image-editing system
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
