I use a Mac, and wanted to be able to run MacOS programs like Xcode and iOS simulator, so I wrote a couple of different sandbox projects:

- SandVault (https://github.com/webcoyote/sandvault) runs the AI agent in a low-privilege account

- ClodPod (https://github.com/webcoyote/clodpod) runs the AI agent inside a MacOS VM

In both cases I map my code directories using shares/mounts.

I find that I use the low-privilege account solution more because it's easier to setup and doesn't require the overhead of a full VM

do you have a write up on your setup?

After reading an article about doing 10,000 pushups in a year (https://wjgilmore.com/articles/10000-pushups), I created "push10k", an iOS app to help me keep track and stay motivated. It's free (no money, no ads) in the iOS app store: https://apps.apple.com/us/app/push10k/id6754811078.

The Vorkosigan Series, by Louise McMasters Bujold. She’s won six (!!!) Hugo awards for her writing, and as Anne McCaffery says, “Boy, can she write”.

Space opera with warfare, intrigue, politics, drama, and world building.

Here are a couple of (open-source Apache license) projects I wrote to sandbox on Mac, which I use to run my agents, while still being able to build/run macOS apps:

Limited user account: https://github.com/webcoyote/sandvault

Virtual machine: https://github.com/webcoyote/clodpod

Hey, I got to see this code!

Back when Blizzard was still Silicon & Synapse, we got Rebecca's source code to Another World SNES from Interplay to use for a game we would develop, and they would publish, and I was the engine programmer.

I remember reading the source code, which was ... sparsely documented, and wondering what was going on. Like "you're writing to the DMA registers?!?"

The code was amazing, because it has has to draw polygons into 8x8 pixels cells that are stored in planar format at 60FPS. On a 3.5 Mhz processor. Blew my mind.

Incidentally, the game was called "Nightmare", and later became "Blackthorne", which was released for SNES, Genesis, and PC.

Yeah, Another World was an incredible feat with the hardware we had at the time.

> is there a concise theory of game design that properly explains why cutscenes are fucking stupid?

Yes. In general it's because they're made by a different team, with different incentives, working to a different schedule.

They're often made using an earlier version of the game lore and story. Due to the massive effort required to make changes and render frames, they often don't match up with late-breaking changes made by the game team.

But sometimes you get lucky and the cinematics team excels. I worked with Blizzard's cinematics team in the '90s, and those spectacular folks produced an amazing body of work.

There's also an app, MenuWhere, that enables you to configure different keys to walk the menu bar. It's free (but nagware). https://manytricks.com/menuwhere/

I'm playing around with sandboxing techniques on Mac so I can isolate AI tools and prevent them from interacting with files they shouldn't have access to -- like all my dotfiles, AWS credentials, and such.

I've created two open-source solutions, one which uses a VM (https://github.com/webcoyote/clodpod) and another which creates a limited-user account with access to a shared directory (https://github.com/webcoyote/sandvault).

Along the way I rolled my own git-multi-hook solution (https://github.com/webcoyote/git-multi-hook) to use git hooks for shellcheck-ing, ending files with blank lines, and avoid committing things that shouldn't be in source control.

Have you seen tart https://tart.run/ ?

Yes; the ClodPod project uses tart to build & run the VM. My project is a lot of scripts to make the whole thing turnkey.

Have you considered using docker? Seems possibly more lightweight than a VM with more isolation than a user account based method.

Yes, I've used docker and podman. They're great. But I wanted to be able to run Xcode and IOS simulator, which requires macOS, so developed these solutions.

My gripe with docker vs native code is docker is just slow to build. or maybe im just not using it right.

on macOS Docker is just a QEMU VM underneath, to my limited understanding, so not a big difference I think

> When you use Claude Code, we collect feedback

When they ask "How is Claude doing this session?", that appears to be a sneaky way for them to harvest the current conversation based on the terms-of-service clause you pointed out.

I have this same suspicion. Worse, there’s no way to opt out of giving a response.

If you turn off "Help improve Claude" you will never get this prompt (I never do).

https://claude.ai/settings/data-privacy-controls

That should be how this works, but unfortunately not. I have that toggle switched off, but I still regularly get this prompt.

i have that option switched off but still got the score card 1 to 5 yesterday while working on some code.

I joke about this with other folks:

- artist: look at my pictures

- sound engineer: listen to this intro

- UI engineer: check out this screenshot

Me (a backend engineer): look at these numbers. See how the one one the left goes up faster than the one on the right? That was a year of my life.

I could only imagine how many times numbers have gone underappreciated.

I personally find optimizations interesting.

I'm actually going to make major optimizations to my site I mentioned in my original post soon. Hopefully it'll be appreciated by my users and not go unnoticed