Heya - I'm the author. Glass is a cross-platform MCP server that gives an agent a way to see and interact with GUI apps. I don't like building UI. Agents can code it, but can't verify or debug it. Glass exists to close that loop.

Glass is purpose-built for software development, and aims to enable headless, parallelized agent work while the user continues to use the machine unhindered[0]. It minimizes token usage, increases consistency, and just goes faster by preferring to interact with an app's accessibility tree, before taking the screenshot route (which glass also fully supports).

Glass is cross-platform and the code is Apache-2.0 licensed. We release builds for Linux, Windows, and macOS hosts. Glass supports development targeting the three above host platforms, Android (all hosts) and iOS[1] (macOS hosts). The internal API is backend-agnostic, and new platforms can be added in the future pretty easily.

Feedback is very welcome, either here or at github. If it breaks on your app, I am especially interested in knowing about that. I hope you build great things.

[0] Due to the license enforcement on Windows and macOS, isolated headless development targeting those platforms requires a VM. It will otherwise interact directly with the desktop. Android and iOS development on those platforms can still be done headlessly via the emulators.

[1] The iOS backend does not currently support multi-touch gestures.

When something can be treated as a score to compare with others, it will be. Everything else just follows.

I really enjoyed the recent Hackerbox[0] featuring an FPGA. I'd never worked with one prior to that.

https://hackerboxes.com/collections/past-hackerboxes/product...

collect $1,329,063

Having good error messages is one of the important priorities for me in my compiler, so I made the commitment early and am using codespan_reporting[0] to report the errors, which is going well so far.

0: https://github.com/brendanzab/codespan

Did that include bugs due to encountering unexpected nulls?

My understanding is that he was looking for instances of TypeError, AttributeError, undefined is not a function and so on. So yeah, nulls should've been included.

https://www.reddit.com/report to skip the subreddit mods and send it right to the admins.

I did try that - the response above is from reddit admins itself.

I have an old phone set up here, running Octo4a. It's working great.

https://github.com/feelfreelinux/octo4a

My big want here is containerized deployments with build + release steps to allow me to e.g. run migrations and after-deploy tasks (we use both). This prevented a move to Render for us previously.

Could you go into one more level of detail about your app? I think this will help me better understand content for some documentation.

Here's the rough bits of what Fly has:

1. There's a release command (https://fly.io/docs/reference/configuration/#run-one-off-com...) that runs after the container is built, but before its deployed. In Rails that's when a database migration would be run.

2. To run a task after the application is deployed, there's shell access. Here's what that looks like for running Rails tasks: https://fly.io/docs/rails/the-basics/run-tasks-and-consoles/

3. Pre-deployment/build commands can be run from the Dockerfile, like a Rails asset compilation. Here's a link to that https://github.com//superfly/flyctl/blob/master/scanner/temp...

I recognize that this is a lot for folks who aren't comfortable configuring stuff and want the "no-config ease" of Heroku, but it's at least possible on Fly.

To double tap the sibling's replies here - I'm a fly user and found it challenging to locate the release_command docs. Highlighting these in the docs somehow would be a good idea since I think _most_ apps will need something like this.

Oh, it looks like the release command would fit the `build` step I mention.

As far as the after-deployment tasks go, we automate those just like migrations - they're (occasionally very slow) one-offs that we don't want to hold up a restart for. Really, an analogue to the release command that can run after restart is all that I'm talking about here.

We're going to start working on a release phase at Render in the next month. Stay tuned! We'll also update feedback.render.com.

That explains the JVM, but why Java, specifically? JRuby offers everything you mention.