In partnership with
6-min read
Hello AI Enthusiasts,
Explaining the same project to AI again gets old.
Meta’s latest models aim to remember useful context and keep working through longer tasks. Gumloop now lets you describe an agent and have its assistant build the setup.
Today: what changed, the infrastructure behind longer AI tasks, and a small project to practice giving AI better instructions.
🧠Meta’s new AI remembers context across conversations
Meta introduced Muse Wonder and Muse Wonder Full on October 12, designed for tasks such as understanding documents and building working apps.
Image source: AI-generated
Two features to know:
Active Memory: carries relevant information across conversations.
Active Focus: aims to keep pursuing a goal, recover from errors and try another approach.
Why it matters: For a small team, remembering project details could mean less time repeating instructions. The useful test is whether the assistant also follows them correctly.
Try this: Give it a fictional customer brief. Ask it to identify missing information, suggest next steps and flag decisions that need your approval.
Check access: Muse Wonder is available through the Meta Model API and Muse Code. Wonder Full is available to Muse Code subscribers. Broader availability in Muse is still upcoming, starting with subscribers.
These are Meta’s claims—not results from our own testing.
🤎 Sponsored
Yes, that Arnold Schwarzenegger.
He has a newsletter on beehiiv. So do Codie Sanchez, Jay Shetty, David Begnaud, Colin & Samir, and Joanna Stern.
They could publish just about anywhere. They chose beehiiv to build a direct connection with the people who want to hear from them most. And you can do the same, whether you’re world-famous or just at the start of your plan for world domination.
🔥 Two updates worth your attention
Image source: https://blogs.microsoft.com
🔹Microsoft previews chips built for longer AI tasks
Microsoft says Maia 200 already runs production workloads including Copilot. Its next system, Maia 300, is scheduled to begin production deployments in 2027.
Microsoft expects Maia 300 to deliver 40% more tokens per megawatt than leading GPU-based systems
Why it matters: More efficient infrastructure could help Microsoft support more AI work at scale.
The honest part: This is an infrastructure announcement, not a feature you can switch on today. It does not promise cheaper Copilot subscriptions.
🔹Gumloop now builds AI agents from a plain-English request
Gumloop’s October 7 update lets you tell Gumball, its assistant, what you want an agent to do. It can build or update the agent’s instructions, connected tools, knowledge sources and triggers.
The business idea: Describe a repeatable job instead of starting with a blank automation setup.
Try this: “Build an agent that sorts customer inquiries by topic and drafts replies for me to review. Do not send anything.” Test it with fictional messages first.
Check access: Review available features, connected-app permissions and usage costs. Creating an agent doesn’t establish that its workflow works correctly.
🤎 Sponsored
100+ ChatGPT Prompts to Revolutionize Your Day
Supercharge your productivity with HubSpot's comprehensive guide to Chat GPT. This free resource is your fast track to AI mastery:
Industry-Specific Use Cases: 15+ real-world applications across various sectors
Productivity Guide: 21 best practices to 10x your efficiency with AI
Prompt Powerhouse: 100+ ready-to-use prompts for immediate implementation
Challenge Buster: Overcome common AI hurdles with expert strategies
Plus, in-depth sections on email composition, content creation, customer support, and data analysis.
🛠️ Build a 3D delivery game from one prompt
Already tried “Pack the Order”? Our Google Playground walkthrough also includes “Parcel Hop 3D.”
A tiny robot carries a parcel across rooftops. It can put the parcel down, use it as a stepping stone and collect it again.
Inside the guide:
The complete prompt to copy and paste.
Keyboard and mobile controls.
Rules for checkpoints, falls and restarting.
Three things to test: Can the robot retrieve its parcel? Does falling reset both correctly? Can someone play without your explanation?
The useful skill goes beyond games: give AI clear rules, find a failure and request one specific correction.
Member-only Medium article. Playground launch access was limited to U.S. adults 18+, with creation rolling out by subscription tier.
🤎 Sponsored
Take the prompts our creative team actually uses
The key to maintaining brand quality at scale? Mastering how to communicate with AI models and embedding them into your creative strategy.
Join our honest discussion with Ari Murray, Chief Digital Officer at Salt and Stone, and get 5 tips for expert LLM prompting with ready to run prompts.
📬 From our other newsletters
💼 Call-center jobs shrink. These related roles are growing.
Sunday Special #69 examines the workforce shift—and why moving into a growing role isn’t automatic.
Also inside: remote career picks, Office inside Copilot and two online hackathons with cash prizes.
Want it every Sunday?
🩺 Google’s medical AI meets real patients—with doctors watching
AIHealthTech Insider #122 explains what doctors found useful in an AI chat before appointments—and what needed checking.
Also inside: bone-health clues in chest X-rays, infant movement screening and an experimental fat-targeting injection.
Want it weekly?
One useful move this week
Choose one task you repeat. Write down what a correct result looks like before asking AI to do it.
Then track the fixes it needs. That tells you more than a polished first draft.
Know someone building a business with AI? Forward this issue.
