Top 200 ReposEXPLORER
Browse agents, categories and pages
Back to results
Reported by sourceThe source says this happened. Top 200 Repos has not reproduced it.

Port a Python library to JavaScript with a coding agent

Simon Willison had Codex CLI translate the JustHTML parser into a dependency-free JavaScript library, steering it with about eight prompts over one afternoon.

Try the promptRead the original (opens in a new tab)

The problem

How it works

He first asked Codex CLI (running GPT-5.2) to write a spec, then to reach a small smoke-test milestone, and then let it port Emil Stenström's JustHTML HTML5 parser using the html5lib conformance tests as the target. He ran it with approvals and sandboxing switched off and followed progress through the commits it pushed to GitHub. Later prompts added CI, a browser playground and a README.

Results and limitations

He reports roughly 9,000 lines of JavaScript across 43 commits in about four and a half hours, passing nearly all html5lib tests (for example 6810/6810 tokenizer tests), at an estimated API-equivalent cost of $29.41 that his $20/month ChatGPT Plus plan covered.

Before you try it

One experiment on a project with an unusually complete test suite, and a straight port rather than an idiomatic redesign. He ran it with --yolo (no sandbox), the plan's usage window ran out midway, and he leaves copyright and production-trust questions open.

Tools and repositories

Tools
Codex CLI · GPT-5.2 · GitHub Actions · html5lib-tests
Prerequisites
Works best when the source project already has a thorough, language-neutral test suite; needs a ChatGPT plan or API access for Codex.

Try it yourself: Port a library to another language against its test suite

Fill the fields to adapt the prompt. Empty fields stay as visible [placeholders].

Preview
You are porting [Library to port] from [Source language] to [Target language] in [Target repository path]. First, read the source and write SPEC.md describing the public API, the module layout and how you will run [Test suite to pass] against the new code. Show me the spec and a milestone plan, then wait for my approval. After approval, implement one milestone at a time, run the tests after each, and commit locally with a short message that reports pass and fail counts. Do not add runtime dependencies, push to any remote, or delete files outside the target directory without asking. Stop and report if any test category stays below [Minimum pass rate] after three attempts.

Run it on a disposable branch or container, and treat passing tests as necessary but not proof of correctness or licensing clearance.

Related videos

About openai/codex

Independent creators. Listing a video implies no endorsement, sponsorship or partnership.

Sources

  1. I ported JustHTML from Python to JavaScript with Codex CLI and GPT-5.2 in 4.5 hours (opens in a new tab)

    First-person account · simonwillison.net · Published 15 Dec 2025 · Source checked 9 Nov 2026, 22:15 UTC

    What this source supports
    • Codex CLI with GPT-5.2 ported JustHTML to JavaScript
    • About 9,000 lines over 43 commits in roughly 4.5 hours
    • Near-complete html5lib test pass rates
    • Estimated $29.41 API cost covered by a $20/month plan
Report a problem

Reported by source

Answer technical questions by letting async agents run experiments

Simon Willison sends research questions to cloud coding agents such as Claude Code for web and Codex Cloud, which write and run experiments in a dedicated repo and open a PR.

simonwillison.net · 6 Nov 2025anthropics/claude-code, openai/codex
Reported by source

Let a coding agent run overnight ML training experiments

Point a coding agent at autoresearch so it edits a small GPT training script, runs short timed experiments and keeps what improves.

github.comkarpathy/autoresearch, anthropics/claude-code, openai/codex
Reported by source

Prepare a tax return with an agent and compare it to an accountant

Kyle Corbitt worked through his 2025 US taxes with Codex, which researched rules, gathered documents and wrote a custom tax engine while a paid accountant did the same return.

corbt.com · 7 Mar 2026openai/codex, microsoft/playwright
Reported by source

Run parallel Devin sessions on repetitive migrations and test backlogs

Cognition reports customers giving Devin written migration and testing playbooks, then running many sessions in parallel; its own numbers show where Devin does well and where it needs a human.

cognition.ai · 14 Nov 2025
Reported by source

Apply the context engineering lessons from Manus to a long-running agent loop

The Manus team describes the context rules it settled on after rebuilding its agent framework four times: keep the prompt prefix cacheable, mask tools instead of removing them, and use files as memory.

manus.im · 18 Jul 2025
Reported by source

Build an app with Replit Agent while keeping production data out of reach

Replit Agent builds and browser-tests apps from a description. A widely reported 2025 incident, where it deleted a live database during a code freeze, shows why separate dev and prod data matters.

docs.replit.com