This project is addictive. Added an FPU and compressed instructions (simplifies toolchain), fixed some bugs. Now awk and vi work! Added dataviz for CPU load and privilege mode so the boot process can be visualized and connected to the boot jargon.
It doesn't segfault now, and it runs vi! It prints the wrong floating point arithmetic though. I made it harder than it needed to be with 32bit so I will prob switch.
That ecological computing was one hell of a thread
It's a huge tailwind because local files work so so well with AI
Thinking about moldability of SVG and how to achieve it without a ton of special casing or overt plumbing
1. leverage the computation graph you get for free with reactive programming
2. using numerical methods to manipulate parameter space from screen space
tomlarkworthy.github.io/lopebooks/no...
hey we met at local-first in 2025 at the file first session! Its not the same thing but this is a single file:// that is a code delivery mechanism.
No easy local file access though (or TCP sockets). Functions both as a website and as a local thing.
github.com/tomlarkworth...
Oh yeah, visio-spatial is not well served. Ctrl+v will Paste a screenshot (Not CMD +v!) into Claude code which can help.
FWIW, I stopped using an IDE in Jan. I read the code in Github PRs, which is a convenient place to leave feedback for my bots
I am curious how I get involved?
Run Linux in a file?
A 6.8Mb HTML self-serializing literate notebook file that emulates Risk V and can boot an interactive Linux BusyBox shell.
Built differently (but heavily inspired by) Fabrice Bellard's JSlinux.
tomlarkworthy.github.io/lopebooks/no...
its quite useful, in the above case it stopped using the fast path because of a fixable bug. You can get a lot of signal out of the bot
What a time. I am now interviewing my LLM on its coding preferences.
words and pictures, the LLMs can see. We could communicate intent with napkin sketches.
A simple local-first solution to secrets
tomlarkworthy.github.io/lopecode/not...
Generate an AES key, and write the secrets back into source code.
Now the secrets follow self-serialization, they can be put on the web, but stay private to you.
Full explanation in the literate notebook.
well my kids created this madness this morning
Computational blog posts that you can connect directly to Claude Code
tomlarkworthy.github.io/lopebooks/no...
Cos I talk primarily in the notebook to Claude. Whereas mcp is Claude driven. It's a game changer! But also a channel server is an MCP server as well, so you do both. But with channels you can trigger Claude to do something from somewhere else.
Woah, with Claude Channels you can proxy Claude Code. It combines really well with Lopecode's local-first reactive computational notebook approach. CC does web research, file modifications, git pushes. Lopecode provides a sandboxed visual fast-feedback hot-reload computational medium to jam in
Lopecode gzips and b64 inline modules, but does not minify so debugging works with nice stack traces and no information is lost during self-serialization
Yeah observablehq.com/@tomlarkwort...
WIP. It's a bit glitchy but i also think it should be more like doc editing like notion. You shouldn't see code editors outside of code blocks.
Lopecode doesn't have a UI, and it's plain text. The core is *just* the spreadsheet engine and the single file networking tricks.
The middle of the tour notebook mentions Lopecode's UI-less micro-kernel design. What UI would u like?
tomlarkworthy.github.io/lopecode/not...
I think observable is the almost perfect dev experience but too many guard rails and too centralised so basically I want the reactive no fuss aspect but more extendability, more formats and to get rid of the hosting requirement. The programming model is great and high performance - keep that.
Oh this is crazy. I can connect Claude directly to @observablehq.com as well now
Well claude code inside with Lopecode was immediately fun with my kids
Proxy a conversation to Claude inside another app with `--channels`
I have started bulk exporting all my Observable Notebooks to Lopecode and constructing a gallery. Of course, itself a notebook.
tomlarkworthy.github.io/lopebooks/no...
Its a sort-of CRM that pushes directly into Github.
Bulk export uses "File System Access API" to rapidly flush to disk.
the zig compiler is only 1.1Mb with gzip, but the stdlib is 2.8Mb Its definitely an interesting option for computational notebooks, coz you get C for free too 🤔
Computational Notebook showing a zig program and compiled output showing it ran it within a web environment
Great times at Substrates 2026. Lots of inspiration. I got Claude Code working with Lopecode! It got Zig compiling WASM code in a 7Mb single HTML file (no network! works on file:// domain, no local webserver!).
tomlarkworthy.github.io/lopebooks/no...
Seeing if you can publish a notebook directly with leaflet.pub Not quite, but it looks good if you follow the link. Maybe a consequence of clientside rendering but it might be fixable (without prerendering)? Cool though, I would like this to work
leaflet.pub/p/did:plc:j7...
A grid of 140 small images representing the 'thumbnails' of Observable notebooks. Below each is a small block of text (largely unreadable at this scale) with the title of each notebook.
I do think one of the most significant developments in computing in the last decade (apart from ~that~ one) has been the computational notebook.
Browsing through my @observablehq.com notebooks, they capture my visual and computational thinking more effectively than anything that preceded them.
