Is that even a sandbox?

I thought it was just a wrapper around an (old) existing tool that has been infinitely rebranded. Their old "remote desktop" program and some web listing capabilities to launch it in "rootless" mode.

Yes, there is a sandbox.

https://simonwillison.net/2026/Jan/12/claude-cowork/

That’s the point of this gist, and the related blog post.

Also, it’s a bit of a stretch to call Claude Code, which isn’t even a year old…old.

sorry, thought you meant the “Windows App” from microsoft.

https://apps.apple.com/us/app/windows-app/id1295203466

Cowork is only available on macOS for now I think.

Its great, i have been using it . Two requests though 1. iOS app 2. API option to use against meeting transcription or route audio from Mic .

+1 on the meeting tranecription

Too many options here - keep it simple https://coderemote.dev/ https://www.tonkotsu.ai/ https://www.terragonlabs.com/ https://kisuke.dev/ https://opcode.sh/ https://yolocode.ai/

Claude is coming up in 6th or 7th place and below in most countries including US, but in 2nd place in the world, how is it possible, what am i missing.

The data gathering method is 1.1.1.1 Cloudflare DNS resolver. It won't have the information about people who use any other DNS.

No i meant when you change country on top of cloudflare report, it comes up like 6th and 7th for most of countries i selected, but it comes to 2nd place suddenly when you select world.

Yeah, a more meaningful study next would be the market share of DNS resolvers.

Not every metric published here can be used, because the observers are from the PoV of Cloudflare and cloudflare alone.

Got the same use case, i tried last month with gemini app and it was awful, most rendering was messed up.

Cool work, How much did it cost to train ? Will the source for training be open source ?

We won't know how much it cost until after the court case.

This looks like a model RLed on top of Qwen3-Coder or GLM 4.6 as per their graph and foot note.

What local models will it work with ? Also what will be the pricing for local llms?

Good question, I imagine you’d need to set up an ngrok endpoint to tunnel to local LLMs.

In those cases perhaps an open source (maybe even local) version would make more sense. For our hosted version we’d need to charge something, given storage requirements to run such a service, but especially for local models that feels wrong. I’ve been considering open source for this reason.

Can it cook ?

Looks like it could pass butter, at least.

From their FAQs:

> Initially, NEOs cooking capabilities will be restricted from use. NEO can provide you with great recipes or help with the cleaning up instead.

Huge risk having it operate in a kitchen: open flame, hot liquids, spills, sharps, etc.

The main risk I was thinking of was poisoning you.

Look at it, clearly only serves tomatoe sauces, red wine, and beets.

Let him cook.

What models are used ? and whether user data is used to train ? Not seeing privacy policy or TOS page.

Good questions! Will add these and put a data processors page tonight. We use Anthropic/OpenAI models.