{"id":51076957,"url":"https://github.com/zozo123/wolfram-fb0","last_synced_at":"2026-06-23T15:02:04.511Z","repository":{"id":359994167,"uuid":"1248281984","full_name":"zozo123/wolfram-fb0","owner":"zozo123","description":"AI-written x86_64 asm + eBPF for Wolfram fractals direct to /dev/fb0. Entire build runs in a real-VM islo sandbox in your browser. No install. Fork-and-rerun in one click.","archived":false,"fork":false,"pushed_at":"2026-05-24T13:59:02.000Z","size":129,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-24T15:07:42.246Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/zozo123.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-24T12:42:25.000Z","updated_at":"2026-05-24T13:59:05.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/zozo123/wolfram-fb0","commit_stats":null,"previous_names":["zozo123/wolfram-fb0"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/zozo123/wolfram-fb0","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fwolfram-fb0","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fwolfram-fb0/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fwolfram-fb0/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fwolfram-fb0/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/zozo123","download_url":"https://codeload.github.com/zozo123/wolfram-fb0/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fwolfram-fb0/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34694786,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-23T02:00:07.161Z","response_time":65,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-23T15:01:58.602Z","updated_at":"2026-06-23T15:02:04.504Z","avatar_url":"https://github.com/zozo123.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# wolfram-fb0\n\n\u003e **AI wrote x86_64 assembly + eBPF for Wolfram fractals direct to `/dev/fb0` — every step (build, kernel trace, qemu boot, framebuffer) runs in a real Linux VM in your browser via [islo](https://islo.dev). No install. No drift. Fork-and-rerun in one click.**\n\n[![Live demo (islo sandbox)](https://img.shields.io/badge/live%20demo-islo%20sandbox-7c3aed)](https://zozo123.github.io/wolfram-fb0)\n[![Fork the sandbox](https://img.shields.io/badge/fork%20the%20sandbox-one%20click-22c55e)](https://zozo123.github.io/wolfram-fb0#fork)\n\n---\n\n## The pitch in one paragraph\n\nMost AI coding sandboxes are containers. You can't write eBPF, touch `/dev/fb0`, or run nested qemu in a container — no kernel modules, no real kprobes, no framebuffer. **islo gives you a real virtualized VM in your browser**, which is what makes this demo legal: an AI (smallest open-weight model we could fit) writes pure x86_64 assembly and eBPF programs that render Stephen Wolfram's cellular automata (Rule 30, 90, 110), the Mandelbrot set, and a Julia animation directly to the Linux framebuffer. Inside the same sandbox: `qemu-system-x86_64` boots a 4 MB Linux straight into the agent-built ELF. eBPF traces every syscall and every byte mmapped to fb0. Both stream out to your browser tab as the fractal blooms.\n\n## Why this exists\n\n* Fractint (1988) and [FractalAsm](https://github.com/mrmcsoftware/FractalAsm) showed that hand-tuned assembly + framebuffer is one of the most visually rewarding low-level art forms in computing.\n* Almost every AI-coding demo today is a webapp. The opposite of a webapp — pure ELF, no libc, kernel ABI only — is exactly the kind of work that's *hard* for LLMs (no Stack Overflow corpus to copy) and *easy* to verify (pixel-diff is unforgiving, byte-count is a scalar).\n* And every \"reproducible\" AI demo dies on install friction. The README rots. The toolchain drifts. The model gets deprecated. **The cloud-sandbox-as-replay-link is the reproducibility primitive these demos have been missing.**\n\n## What's in the box\n\n```\nsrc/        — x86_64 assembly (rule30.s, mandel.s, julia.s)\nbpf/        — bpftrace programs that X-ray the asm at runtime\noracle/     — Python reference renderer + pixel-diff + binary-size scorer\nharness/    — agent loop: islo skills orchestrate, opencode codes, oracle judges\ndist/       — final agent-built ELFs\niterations/ — every loop iteration, committed (the convergence record)\nqemu/       — boot.sh + tiny initramfs for the framebuffer demo\nsite/       — GitHub Pages: plot, live viewer, charts, fork link\ndocs/       — design notes\nislo.yaml   — declares the real-VM sandbox shape\n```\n\n## Run it (the islo way)\n\nYou don't need anything installed. Click the **Fork the sandbox** button at the top, or:\n\n```bash\nislo use wolfram-fb0 --source github://zozo123/wolfram-fb0\n# → boots a real VM with nasm, qemu, bpftrace, opencode, model, all warmed.\n# Inside: `make demo` boots the fractal, opens the eBPF trace, streams to your browser.\n```\n\nTo watch the agent rebuild from scratch:\n\n```bash\nislo use wolfram-fb0 -- make agent-loop\n```\n\n## Run it (the hard way, on your own Linux box)\n\nYou'll need `nasm`, `ld`, `qemu-system-x86_64`, `bpftrace`, Python 3 with `pillow`/`numpy`, an Ollama with a small Gemma pulled, and `opencode`. We support this for masochists. The full sandbox image is one `make sandbox-snapshot-import` away if you really want it.\n\n## Targets\n\n| Target | Asm budget | Math | Why it's interesting |\n| --- | --- | --- | --- |\n| Rule 30 | ~256 B | integer XOR | Wolfram's signature 1-D automaton. Trivial math, ruthless on byte budget. |\n| Mandelbrot | ~1 KB | SSE float, no libc | Forces the model to use raw SSE; no `sin`/`cos`, no libm. |\n| Julia animation | ~1.5 KB | SSE + frame loop | Adds time as a dimension; tests state-machine discipline. |\n\n## Convergence\n\nAfter the agent loop finishes, this section will embed the charts: binary size descent over iterations and pixel-diff convergence, by target. *(Will appear after first full run.)*\n\n## Provenance \u0026 lineage\n\n* **[Fractint](https://www.fractint.org/)** — the 1988 ancestor; integer-math fractals on a 386.\n* **[FractalAsm](https://github.com/mrmcsoftware/FractalAsm)** — modern asm fractal renderer.\n* **Stephen Wolfram, [A New Kind of Science](https://www.wolframscience.com/nks/)** — the cellular automata.\n* **[islo](https://islo.dev)** — the real-VM sandbox that makes the whole thing one click.\n* **[opencode](https://github.com/sst/opencode)** — the inner-loop coding agent.\n\n## Try islo\n\nEvery new account on **[islo.dev](https://www.islo.dev)** ships with **$50 of free credit, no card required** — enough to spin a real-VM sandbox like the one this demo runs in and reproduce the full convergence loop yourself.\n\n## About\n\nBuilt by **[Yossi Eliaz](https://www.linkedin.com/in/yossi-eliaz)** ([@zozo123](https://github.com/zozo123)) for [islo.dev](https://www.islo.dev).\n\n## License\n\nMIT. Fork freely.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzozo123%2Fwolfram-fb0","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fzozo123%2Fwolfram-fb0","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzozo123%2Fwolfram-fb0/lists"}