Hi everyone! I’m the maker of NULNUL. I started building NULNUL after noticing that people who wanted to try AI coding were spending too much time learning how to configure the AI environment before actually building anything. Which skills should I install? How many agents do I need? Which plugins are worth using? How should I write the rules? And once you start adding everything that looks useful, the setup can quickly become more complicated than the project itself. NULNUL takes a different approach: **Start small, reuse what already works, and add something only when the project proves it actually needs it.** Sometimes the best result is: **0 new agents. 0 new skills. 0 new infrastructure.** When a change is needed, NULNUL tries to keep it small, verifiable, and reversible. It also separates proposing an improvement from approving it, so the harness can evolve without simply declaring its own changes “better.” It currently works with Codex and Claude Code as a skills-only, repository-local harness. My goal is simple: **you shouldn’t need to become an expert in AI development environments before you can start building something.** I’d especially love feedback from people who use Codex or Claude Code regularly: **What part of setting up or maintaining your AI coding environment wastes the most time for you?**

NULNUL Harness

Start small. Let the harness grow with your project.

Start small. Let the harness grow with your project.