Good morning! Here's what's happening in AI today:
Claude Artifacts now supports public sharing and multiplayer editing
Manus launches auto-publish so your builds go live automatically
Google Gemma 4 runs live on Cerebras at over 1,500 tokens per second
How to install your own identity into an AI agent
And 4 new AI tools worth trying today
AI TOOL
Anthropic announced that Artifacts now support public sharing and multiplayer editing in Claude Code, and can be created directly with Claude Tag.
Artifacts can now be shared publicly with a link, letting anyone view what you built without needing their own Claude account to see it.
Multiplayer editing means multiple people can work on the same Artifact at once, with changes syncing in real time across everyone involved.
Claude Tag lets you create an Artifact directly by tagging Claude in a conversation, skipping extra steps to get from idea to finished output.
This works inside Claude Code, meaning the collaborative editing extends to the same environment developers already use for building.
Artifacts have always been useful for one person building something quickly. Public sharing and multiplayer editing turn that into something a team can actually work on together. If you have ever built something in Claude and wanted to hand it off or collaborate on it live, this closes that gap directly.
AI TOOL
Manus introduced auto-publish, deploying every successful build straight to your live URL and removing the last manual step between building and shipping.
Manus now deploys each successful build directly to your live URL, so your site updates without you needing to click publish separately.
Auto-publish is off by default, so you stay in control of exactly what goes public and when, rather than everything shipping automatically.
Queue your changes and turn on auto-publish together, and you can wake up to a finished site without touching it again.
This removes the final manual step between finishing a build and having it live, which previously required a separate action every time.
The gap between finishing a website and actually publishing it has always been a small but real friction point. Auto-publish closes that gap entirely while still letting you control what goes live. For anyone building sites regularly with Manus, this means less clicking and more building.
AI MODEL
Google confirmed that Gemma 4 is now live on Cerebras, running the 31B open-weight model at over 1,500 tokens per second.
The 31B open-weight Gemma 4 model now runs on Cerebras hardware at speeds exceeding 1,500 tokens per second, a significant jump over standard GPU inference.
This is described as a 15x speedup, which unlocks real-time visual and agentic workflows that previously felt sluggish on standard hardware.
Faster inference means agents and multimodal tasks that require multiple rapid steps can run without the lag that usually comes with heavier models.
Because Gemma 4 is open-weight, developers can access this speed improvement without being locked into a specific closed platform.
Speed is often the difference between an AI feature feeling instant or feeling like a wait. Running Gemma 4 at this speed on Cerebras removes the lag that holds back real-time and agentic use cases. For anyone building on open-weight models, this is a meaningful jump in what feels usable in production.
HOW TO AI
I've watched more people set up an AI agent with everything technically perfect and still feel nothing from it. Every hook registered, dashboard running, all green lights. And the system still didn't know them.
That's not a setup problem. That's a missing ingredient problem.
The agent doesn't become useful because it can think. It becomes useful because it finally knows what you want it to think toward. Here's how to actually put that in.
Step 1: Write the one-page file that makes it yours
Most people skip straight to configuring tools and automations. But a technically perfect agent with no identity behind it is just a chat tab with extra steps.
Before anything else, write a single file describing who you are, how you work, and what you're aspiring toward. Career arc, worldview, the way you think through problems. This file loads into every session, so the agent is working from an actual person instead of a blank template.
Step 2: Give it a goal layer it can actually steer by
Above your identity file sits a layer of missions, goals, and strategies the agent can reference when it's deciding what to do next. This is what turns a system that responds into one that steers.
You're not scripting the agent's behavior line by line. You're programming it with yourself, the same way you'd feed a model data instead of hardcoded logic.
Step 3: Set boundaries around what it can read, propose, and write
Give the agent wide read access to your notes so it has real context to work from. Let it propose changes to your identity or goal files, but only behind a gate you personally approve. And give it exactly one place it can write to on its own, like a single inbox folder, not free rein across your entire system.
Read wide, propose gated, write narrow. You stay the author. The agent stays a guide.
Step 4: Make automations fail loud, not silent
Any automation you set up should log every run and have a deterministic fallback if something breaks. A silent failure that looks like success is worse than an obvious one, because you won't know to fix it.
If an automation needs stored credentials, keep in mind that on a Mac, background jobs cannot see what your logged in session can. Set those up to run through the proper system service instead of a basic scheduled task.
The machine handles the logic. You get to be the person the whole system exists to serve.
ChatGPT confirmed it is available again on WhatsApp in the EEA, along with new availability on Kakao in South Korea and Viber in supported markets.
GitHub improved Copilot CLI with smarter subagent delegation, cutting tool failures 23%, search failures 27%, and edit failures 18%, with no regression in quality.
Notion Calendar rolled out new agent tools with rich, interactive UI, letting you take quick actions like jumping into a meeting link without ever leaving the chat.
🎨 Claude: Share Artifacts publicly and edit them together in real time.
🚀 Manus: Deploy your builds automatically straight to a live URL.
🖌️ Reve: Edit image colors precisely using hex codes or on-image selection.
📅 Notion Calendar: Take quick calendar actions without leaving the chat.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial