dev blogbehind the scenes

Building AxiomWeaver: One Developer, One AI, and a Lot of Rust

By Rob Chipman

I should be upfront about something: AxiomWeaver is built by one person.

No team. No funding. No investors. Just me, a day job I work alongside this, and Claude — Anthropic's AI — as my development partner.

I'm telling you this because I think it matters. If you're going to trust a tool with your creative work, you deserve to know who's behind it and how it gets made.

What "built with AI" actually means here

When people hear "built with AI," they picture someone typing "make me an app" and shipping whatever comes out. That's not this.

I'm a software engineer. I design every system, architect the data model, and make every product decision. Claude is a collaborator — fast, knowledgeable, and available at 2am when I've got an idea I can't let go of.

A typical day looks like:

  • I design a feature's architecture
  • Claude helps implement the Rust backend while I sketch the UI
  • I review every line, rewrite what needs rewriting
  • I decide what to test, Claude helps me write tests faster
  • I debug the edge cases that only surface with real data

The narrative engine — temporal state projection, the event ledger, the entity scanner — that's architecture I designed. The 1,200+ Rust tests? I defined what correctness means. Claude helped me get there faster.

The most fun I've had in years

I need to say this because it's the truth that makes everything else make sense: building AxiomWeaver has been the most fun I've had in my career in a long time.

My day job is meetings. Strategy decks. Cross-functional alignment sessions. I'm good at it, but I rarely get to actually create anything anymore. When I started this project, the goal was modest — learn how to work effectively with AI agents. See what was possible.

Three months later, I have a company. A real product. A community of authors testing alpha builds. I went from "I wonder if AI can help me code faster" to "I accidentally built a business" in less than a quarter.

The first few weeks felt like black magic. I'd describe a feature — the entity scanner, the temporal state projection, the composable template system — and Claude would help me build it. Not by writing everything for me, but by being the collaborator I'd never had. The one who knows the algorithm you need (Aho-Corasick, in the scanner's case), explains why it works, and helps you implement it at 2am because neither of you needs sleep.

I'll be honest: in those early weeks, I'd wake up in the morning half-expecting it to be gone. Like I was cheating somehow, and someone would take it away. The productivity felt too good to be real. A solo developer doesn't ship a full prose editor, entity system, event ledger, character sheets with equipment grids and skill trees, a five-format import pipeline, and a real-time entity scanner in six weeks. That doesn't happen.

But it did. 1,200 Rust tests. 1,200 frontend tests. And the thing that surprised me most: I understand every line. This isn't generated code I don't comprehend. It's code I designed, reviewed, tested, and can explain at a whiteboard. Claude made me faster. It didn't make me a passenger.

Why solo + AI fits this product

There's something fitting about a tool for solo creators being built by a solo creator.

The advantage of working alone is focus. Every feature exists because an author mentioned a pain point on Royal Road, or because I hit a wall while thinking through how LitRPG writers actually work. No feature committee. No product manager. Direct line from the community to the code.

The speed is real. In February 2026, the project was a blank directory. By mid-March, it was a real product. That timeline isn't possible for one developer without AI. It's barely possible with it.

What I won't pretend

I won't pretend this is a big company. It's Tiny Forest Labs — named for exactly what it is.

I won't pretend the alpha is perfect. There are rough edges. Features I'm still building. Things I'm figuring out in real time. But every week it gets better, and every feature ships because a real author needs it.

I won't pretend I'm not using AI to build this. I am, and I think that's a good thing. The alternative isn't "a team of 10 engineers building it properly." The alternative is this tool doesn't exist.

The tradeoff

When you choose AxiomWeaver, you're choosing:

  • Someone who listens. I'm in the Discord. I read every message. Feature requests become code within weeks, not quarters.
  • Speed. New features ship weekly during alpha. The narrative engine, character sheets, skill trees, import pipeline — all shipped in the first month.
  • One person. If I get hit by a bus, that's a real risk. I take it seriously. The codebase is clean, tested, and documented specifically so it could survive without me.

I think that's a fair tradeoff for a tool that no big company would build. The market for "LitRPG writing software with temporal state projection" is too niche for venture capital. It's exactly the right size for one developer who cares about the problem.

Try it

The alpha is free. I ship new features weekly. The community is small but growing — authors from Royal Road who are done with the spreadsheet and ready for something built for the kind of fiction they write.

If that's you, request early access. Or just come say hi on Discord — I'm always there.