Running legacy educational Java applets, especially around math and physics, has been a longstanding popular use case of our CheerpJ Applet Runner extension, running Java bytecode in the browser via WebAssembly.
I am not sure how to feel about agents solving the problem via proper modernization. It's certainly positive that students will be able to interact with this content in a modern and more accessible way, but the educational use case for our product, although not commercially important, has always been a source of pride.
As another commenter pointed out this is similar to our WebVM (https://webvm.io) , although this one is based on v86 so it will be less performant.
Although the x86 virtualization angle is powerful we think that native WebAssembly executables on top of a Linux-compatible kernel are more viable for the execution of full workloads in the browser.
As a matter of fact WebVM and BrowserPod share the same kernel, the difference is all on the performance side.
WebVM uses x86 virtualization and hence has a significant performance penalty, with the upside of running any existing software without needing the source code.
BrowserPod on the other hand runs WebAssembly binaries at almost native speed. Source code is required, but that is a fair compromise in the world of sandboxing. Most language runtimes and CLI tools are FOSS anyway, and many closed-source tools (such as Claude Code) are written in scripting languages and run on top of FOSS engines.
The architecture is a fairly straightforward WebAssembly-native monolithic kernel. Most of the complexities come from making things work well within the browser constraints for real world, large apps.
We have quite a bit of experience on the topic however, these are previous projects of ours:
WebVM (https://webvm.io): x86 Debian shell running client-side in the browser via x86 -> WebAssembly JIT compilation
Our technology is much more general that WebContainers, and it's based on a Linux-compatible WebAssembly kernel. It also supports real command line tools, including git, bash and the complete set of busybox utilities.
The version of Claude Code you see running is completely unmodified.
I can confirm non-Java languages work as expected. I've personally tried Kotlin, and IIRC some user from the community reported Clojure to be working as well.
Thanks :-) We have been building WebAssembly-based products for a while now, so for us it's second nature.
But I think you are right, most developers, even experienced ones, have not yet come to grasp the fully capabilities of the Web platform in conjunction with WebAssembly.
You might find previous projects from us also interesting:
We are thrilled to share with the HN community the second preview of BrowserCode: A FOSS web app to run TUI agents (such as Claude Code, OpenCode, Gemini CLI and the like) fully in the browser. This release focuses on Claude Code and initial support for mobile. BrowserCode is released under the Apache License, version 2.0.
BrowserCode is based on BrowserPod (https://browserpod.io), a in-browser WebAssembly-based code sandboxing technology that can currently run Node.js, python, git, bash and many other command line tools. This will further expand to Ruby / Rails, Go, Rust and eventually x64 Linux binaries.
BrowserCode is free to use and unlimited. You'll need to login to each CLI with the corresponding login, i.e. with your Anthropic account or API key. All the data and execution stays completely local to the browser and it's persistent across sessions thanks to a disk backend based on the Origin Private File System API or IndexedDB.
This is a preview release, so please try and break it! Please report issues on GitHub and star the repo if you like our work. Your support will help us push this project forward.
Hello HN! I am happy to share to release with the community.
Thanks to BrowserPod it is now possible to run node.js developer workflow completely in the browser. An example of what can be done today.
- Clone a repo (via git clone)
- npm install
- npm run dev
- Connect to the dev server from anywhere on the internet (via Portals)
- Commit changes and push them to the repo
This is achieved by native WebAssembly builds of git, bash, node and many other utilities. A WebAssembly kernel, designed from scratch by us, provides a consistent virtual machine view to all the processes in the sandbox.
Python is also available in preview, with Ruby, Rust and Go coming soon.
The main use case is to sandbox AI-generated code, but the technology is generic and we can envision many other use cases: Web-based IDEs, educational platforms, live docs, ...
Happy to answer any questions, let us know what you think.
Thanks for sharing. I see how this could impact heavy user, but there is always the option of authenticating via a non-ephemeral auth key as a potential workaround. It's not integrated in the public WebVM UI, but it is supported by the underlying engine (CheerpX).
Hi, lead dev of WebVM here. Can you link to the specific change that affects ephemeral devices? I cannot find this reference in the article.
Keep in mind that we also plan to offer builtin networking in the near future, we are developing the infrastructure to do so as part of our newest product (In-browser sandboxes): https://browserpod.io
Do you have any specific test case that you would consider "very challenging" on the compatibility side? I'd be curious to check if BrowserPod can support that already.
This is not correct. You are using WebVM here, not BrowserPod.
WebVM is based on x86 emulation and JIT compilation, which at this time lowers vector instructions as scalar. This explains the slowdowns you observe. WebVM is still much faster than v86 in most cases.
BrowserPod is based on a pure WebAssembly kernel and WebAssembly payload. Performance is close to native speed.
We test performance internally constantly, and we plan to soon publish a benchmarking repository.
Depending on the workload the performance is actually very close to native, I'll give a few highlights.
* Node.js C++ code is compiled to WebAssembly and as such is very close to native speed
* Node.js JavaScript payloads are executed directly by the browser engine. There is some additional overhead in some scenarios, but it's not large and we expect to further improve this down the line.
* There is proper multi-process support via WebWorkers. Multiple node processes can actually run in parallel on top of the same WebAssembly kernel.
There are a few areas that we need to improve still, to make an example we are hitting some of the limitations of IndexedDB for data persistent and we are experimenting with the OPFS API to squeeze some more performance. This is tricky since performance with OPFS is very dependent on the browser at this tie.
Our kernel itself still need some work, as concurrency in some scenarios is limited by overly extensive locking. We expect to make significant progress on these problems as well over the next weeks and months.
Moreover, the most direct alternative to BrowserPod is not a local sandbox, but a cloud provisioned one. Of course cloud performance depends on how much the platform owner is willing to spend, but we expect BrowserPod to be actually _much_ faster than common cloud options especially on the cheaper tiers.
We plan to publish benchmark against the main cloud sandbox providers as well in the future.
[Edited to mention the comparison against cloud VMs]