I was annoyed at how Claude Code ignores my CLAUDE.md and skills, so I was looking for ways to expand type checking to them. So I wrote a wrapper on top of claude-agents-sdk that reads my CLAUDE.md and skills, and compiles them into rules - could be linter rules or custom checking scripts. Then it hooks up to all tools and runs the checks. The self improving part comes if some rule doesn't work: I run the tool with the session id in review mode, it proposes the fixes and improves the rule checkers. (not the md files) So it's kinda like vibe coding rules, definitely lowers the bar for me to maintain them. Repo: https://github.com/chebykinn/agent-ruler

It says that x86 is highly standardised - even with different combinations of chips, peripherals and motherboards you know it will work just fine. It's not the case for ARM systems - can you even have something similar to IBM PC with ARM?

I personally know that adding support for ARM devices on Linux is a huge and manual task - e.g. look at devicetree, it's a mess. There is no standard like ACPI for ARM devices, so even powering off the computer is a problem, everything is proprietary and custom.

I don't agree with the article though, x86 is dying and my worry is that ARM devices will bring an end to such an open platform like modern PCs are.

I'm a long time linux user - now I have more time to debug issues, submit them, and even do pull requests that I considered too time consuming in the past. I want and I can now spend more time on debugging Firefox issues that I see, instead of just dropping it.

I'm still learning to use AI well - and I don't want to submit unverified slop. It's my responsibility to provide a good PR. I'm creating my own projects to get the hang of my setup and very soon I can start contributing to existing projects. Maintainers on the other hand need to figure out how to pick good contributors on scale.

So Bash and Edit tools are a bit weird, Bash tool is essentially JS execution, and Edit tool automatically generates a script that performs the edits on the page. These tools are needed for the model to explore the page, whatever it does at the end it creates a separate script that will be applied on the page load.

At a high level, the agent generates and maintains userscripts and CSS that are re-applied on page load. Rather than just editing DOM via JS in console the agent is treating the page, and the DOM as a file.

The models are often trained in RL sandboxes with full access to the filesystem and bash, so they are really good at using it. So to make the agent behave well, I've simulated this environment.

The whole state of a page and scripts is implemented as a virtual filesystem hacked on top of browser.local storage. URL is mapped to directories, and the agent starts inside this directory. It has the tools to read/edit files, grep around and a fake bash command that is just used for running scripts and executing JS code.

I've tested only with Opus 4.5 so far, and it works pretty reliably. The state of the file system can be synced to the real filesystem, although because Firefox doesn't support Filesystem API, you need to manually import the fs contents first.

This agent is really useful for extracting things to CSV, but it's also can be used for fun.

Demo: https://x.com/ichebykin/status/2015686974439608607

Comments URL: https://news.ycombinator.com/item?id=46800484

Points: 56

# Comments: 15

Comments URL: https://news.ycombinator.com/item?id=46797168

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=46737865

Points: 56

# Comments: 13

Comments URL: https://news.ycombinator.com/item?id=46737120

Points: 3

# Comments: 1

With it you can pretty much have a coding agent for the currently opened website. You can ask it things like:

- Extract all links from this page and save them to CSV - Switch this site to dark mode - Copy the page content into a Google Sheet - Remove ads

The agent writes JS script that automatically loads every time you visit the page. It is heavily using the userScripts API so you need to enable a lot of permissions to run the extension, and I'm not sure it can be published anywhere.

Under the hood, scripts and styles live in a virtual filesystem where each website is a directory. The agent can search and edit the DOM as a file, which makes it work more or less reliably similar to claude code. Currently it only support Claude models, and I've tested it on Opus 4.5.

Demo video: https://x.com/ichebykin/status/2014258108500177255

Comments URL: https://news.ycombinator.com/item?id=46725118

Points: 2

# Comments: 1