Hello WKND AI Warriors!

Anthropic gave Claude control of your computer, letting it click, type, and run apps while you trigger tasks remotely from your phone.

Also, OpenAI is shutting down its standalone Sora app and folding video generation into ChatGPT, with Disney quickly walking away from its deal.

Plus, Claude Code introduced Auto Mode, letting the AI decide which actions are safe to run without asking every time.

Oh yeah, and Nvidia’s Jensen Huang says we may have already hit AGI, before partially walking the claim back.

Today’s newsletter includes:

•  AI NEWS RECAP

• 🤿 AI DEEP DIVE

📰 AI NEWS RECAP

Claude Computer Use, Cowork, & Dispatch...OH MY!

Claude just did it…AGAIN.

You can now send it a task from your phone
and it will use your computer to complete it.

Not suggest.
Not guide.
Actually do it.

Anthropic introduced two pieces:

Dispatch
Send tasks from anywhere

Computer use
Claude can click, type, open apps, and run workflows on your machine

Put together, it changes the interaction.

You are no longer using AI in a session.
You are assigning work and walking away.

What stands out:

• No chat window required
• Runs in the background
• Works across apps, even without APIs

It is still early.
It makes mistakes.
It needs guardrails.

But the direction is clear.

AI is moving from something you use
to something you manage.

And that changes the question.

Not “How do I use AI?”
But

“What am I willing to hand off?”

OpenAI is in early talks for a deal where Sam Altman's fusion startup Helion would supply OpenAI with up to 5 gigawatts of power by 2030 and 50 gigawatts by 2035. Altman has stepped down as Helion's board chair to avoid a conflict of interest, and Microsoft already has a similar deal in place, pointing to Big Tech's growing hunger for dedicated clean energy to power AI.

ChatGPT now has a built-in visual shopping experience powered by the Agentic Commerce Protocol, letting you browse products with rich images, compare them side by side, and hand off a purchase to an AI agent without leaving the chat. OpenAI says it has no affiliate deals with retailers, so results are not paid placements, though it is clearly positioning ChatGPT as a direct competitor to Google Shopping.

Claude Code's new Auto Mode replaces the risky "dangerously skip permissions" flag with a smarter system where Claude evaluates the risk of each action before running it. Low-risk steps proceed automatically, higher-risk ones pause for confirmation, so you get fewer interruptions on routine tasks without handing the agent a blank check over your system.

Nvidia CEO Jensen Huang told Lex Fridman "I think we've achieved AGI," pointing to AI's ability to initiate and run a billion-dollar tech company as the bar being met. He then slightly walked back the claim, acknowledging that definitions vary, but the statement landed as one of the boldest public declarations yet from a top chip industry leader that the milestone is already behind us.

OpenAI is shutting down its standalone Sora video app, and Disney immediately exited its licensing deal as a result. OpenAI says Sora's video generation will be folded into the broader ChatGPT and API ecosystem rather than living as a separate product, but the swift Disney exit signals the entertainment industry's limited tolerance for AI video platforms that can't sustain themselves.

Google published TurboQuant, a new set of quantization algorithms that compress large AI models to near 3-bit precision while keeping accuracy close to full-size models. It also dramatically speeds up vector search, and Google says TurboQuant makes it possible to run frontier-class AI at a fraction of the memory and compute cost, which could meaningfully lower inference costs across the industry.

🤿 AI DEEP DIVE

I watched all the Claude videos so you don’t have to.

See all of Claude’s updates here!

How can you help?

Refer my newsletter to help others learn AI.

Missed last week’s edition?

Find all of my newsletters here!