howdy, it’s Barsee again.

happy tuesday, AI family, and welcome back to AI Valley.

here are the biggest things worth knowing today:

• Anthropic’s new Claude just built an operating system

• Build a self-improving agent system with Fable 5

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

ADOBE

Creative AI is helping creators think bigger. Here’s what they want next

Creative AI is helping creators grow their businesses, build new skills and take on bigger projects. But bringing their ideas to life isn't always easy. Adobe’s latest Creators’ Toolkit Report, based on a survey of more than 2K+ U.S. creators, reveals where creative AI is opening new opportunities, fueling creative ambitions and giving creators more ways to bring their ideas to life while staying in control.

A few key findings:

• 78% of creators using creative AI say it has accelerated their business or following.

• 65% say making more video is important to achieving their goals as creators.

• 36% of creators who’ve used AI agents for creative work say agents helped them complete a project that was previously too complex.

• 77% say the final creative decision should always remain with the human, even when using agents

Explore how creators are using creative AI to grow, tackle bigger projects and stay in control of their work.

_{*This is sponsored}

THROUGH THE VALLEY

Anthropic’s new Claude just built an operating system. Claude Fable 5.2 and Mythos 5.2 are out, and Fable has gone straight to the top of Artificial Analysis’s Intelligence Index. They’re the same underlying model, but Mythos is only available to vetted cybersecurity and life sciences organizations.

The coding demo is a bit ridiculous: One Anthropic engineer gave Fable an empty folder and asked it to build an operating system. It coordinated more than 1,000 agents for eight days and came back with something that actually boots on a laptop: its own kernel, drivers, file storage, windows, a login screen, even an installer. Not quite macOS. Still, eight days.

Somehow that isn’t the most interesting part. From one-paragraph prompts, Fable designed three electronic devices: a voice-assistant hub, a USB speakerphone and a handheld virtual pet. It did the circuit boards, the enclosures and the software. Diode built them, and all three worked without a single electrical fix.

Then there’s the science. Mythos came up with a new way to model the protein shells of viruses used to deliver gene therapies, and got 28 of the 34 it tried right. Fable solved two open problems in black hole physics, including a calculation that was laid out back in 2003 but was too laborious for anyone to finish.

And the maths: 1,000 Fable agents formalized the existing proof of Fermat’s Last Theorem in 30 hours. That means turning the proof into something a computer can check line by line, not discovering a new one. Anthropic’s first attempt at the same job took 11 days.

People outside Anthropic are already pushing it too. Felix Rieseberg wanted to test Fable with one-shot prompts and no human feedback, and ended up with a 3D game where everything was made from scratch: the models, the sounds, the textures. It took eight hours and cost about $550 at API prices.

Fable 5.2 costs $8/$40 per million input/output tokens, 20% cheaper than 5.1. Cheaper per token doesn’t mean cheap, though. Artificial Analysis ranked it the most expensive model per task on its index, because it thinks for a long time.

Why it matters: I honestly don’t think anyone knows what the next few years look like. Not the labs, not governments, definitely not me. One model just took on problems in physics, biology, hardware and robotics that used to need teams of specialists and years of work. And it dropped on a random Monday.

Picture thousands of agents working on science around the clock, testing ideas, catching each other’s mistakes and building on each other’s results. New medicines, cheaper energy, better materials. Things we’ve accepted as expensive, difficult or impossible might not stay that way.

I think the next decade could bring more scientific progress than we’re ready for, along with a lot of disruption.

TRENDING TOOLS

• ChatGPT Images 2.6 - it’s faster, better and now with GIFs

• Muse Ember - Meta’s smallest and most efficient model, built for complex reasoning execution and everyday agent work

• Patch - build and play games

• Voyager - it puts AI agents inside your Mac's file browser, so you can find files and use them as working context without leaving the folder

• Reverse Engineer Anything - it's one MCP server that lets your AI agent (Claude Code, Codex, Antigravity) take apart an app and explain how it works, down to the binary level

WHAT I'M CONSUMING

• Build a self-improving agent system with Fable 5 in 14 steps

• Fable 5.2 and jagged AGI 

• How to reverse engineer with the most dangerous repo & Opus 5.5

• I cloned myself into Claude

• AI is causing signal collapse everywhere all at once

• The future of software

• Graph Engineering with Claude: 14-step roadmap from 0 to graph architect (full course)

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.