Implement an AArch64 CPU with JIT using an overnight agent herd
An author ran an overnight herd of agents to implement an AArch64 CPU with JIT for the kwemu emulator, since lots of software doesn't build for RISC-V. ·
Done with a custom-built or unnamed agent
- What they did
- Pooya Parsa had already built a RISC-V CPU for the kwemu emulator. He found that much software doesn't build for RISC-V, so he ran a herd of agents overnight to write an AArch64 CPU with a JIT for it. The post doesn't describe the setup, the tooling or where he stepped in.
- How it went
- The post says the AArch64 CPU with JIT was implemented. He expects Fx, OpenCode and Claude Code to run in the browser soon. It gives no test results, timings or numbers.
- Worth knowing
- The post is a short announcement. It gives no cost, agent count or prompts, so check his other posts or the kwemu repo for the actual method.
Try it yourself with Claude Code
In my repo [repo name or path], implement [a missing feature, e.g. support for a new CPU architecture or backend], working in small steps and running the existing tests after each one. Keep a progress log in NOTES.md of what works, what fails, and what you tried. Stop when [the target test suite or sample program] passes, or when you are blocked, and then summarize what is left.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Bluesky · Nov 18, 2026
Related use cases
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Debug frontend UI bugs with a browser agent
A frontend developer pairs Claude in Chrome with Claude Code to open the app, click through it, read the DOM, check console errors and network requests, and fix UI bugs.
Take ideas to implementation with a planning-command pipeline
A solo SaaS founder uses a custom five-command agent flow (intent, impact analysis, planning, execution, review) with lint guardrails to ship features in a production product.
Hand off coding sessions from local CLI to Devin
The poster starts in Devin CLI so it explores the repo and builds context, then hands the session off to continue independently.
Validate firmware in simulation with AI agents
Simantic gives AI agents virtual devboards and a firmware code reviewer so they can test firmware without physical hardware.
Custom agent
Write exhaustive test suites to clear a coverage backlog
A team had Claude write exhaustive test suites for their coverage backlog, but the tests overloaded the CI node and covered disabled features, so it only partly worked.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.