A small lab for weird, useful AI tools.

Last updated: 28 August 2026

The AI Ronin Lab is a one-person workshop. I build one weird or useful AI thing every week and put it here, in the browser, where you can just use it.

Built with AI — not necessarily powered by it. That distinction matters, so I want to be exact about it. AI helps me design and write these tools. Most of them do not call an AI model when you use them. The AI Tool Finder is a good example: it runs entirely in your browser against a list I maintain by hand. There is no model in the loop, no server, and nothing to sign up for.

The ronin part

A ronin is a swordsman without a master. That is roughly the working arrangement here: no employer, no client brief, no roadmap committee. I pick a problem that annoys me, build the smallest thing that solves it, and ship it. Some experiments are genuinely useful. Some are just interesting. Both are allowed.

Everything is deliberately small. No accounts, no dashboards, no subscription. Open the page, use the tool, leave.

Why AI Tool Finder exists

There are thousands of AI tools and roughly the same number of listicles ranking them. Most of those lists are the same twenty products in a different order, written by someone who has not used any of them, sorted by who pays the most.

The Finder tries a different approach. Instead of handing you a list, it asks what you are actually trying to do — the task, the exact outcome, your budget, your skill level, who you are building for, what you care about most, and whether it needs to connect to your existing tools. Then it ranks a curated database against those seven answers and shows the three that fit best, with the reasoning visible.

• 40 curated tools
• No signup
• Nothing leaves your browser
• Same answers, same result

How recommendations are decided

Every tool in the database carries scores for how well it handles each task and each specific outcome, plus how it fits different skill levels, budgets, audiences and priorities. Your seven answers are compared against those scores using fixed weights. The rules do not change between visitors, and identical answers always produce an identical result.

The scores come from reading current first-party documentation — the vendor's own product pages, help centres and pricing pages. They are reviewed in rounds, and when a score turns out to be wrong or stale, it gets corrected and the correction is recorded.

Some honest limits, because a tool that oversells itself is not worth trusting:

• I have not personally used all 40 tools. The scores are research-based, not a claim of hands-on testing of every product.
• The database is curated, which means it is a judgement call. A tool being absent does not mean it is bad — it may just not be in the list yet.
• The Finder is a starting point, not a verdict. It narrows thousands of options down to three worth looking at. You should still try them.
• Pricing and features change constantly. Always check the vendor's current pricing page before you pay for anything.

Independence, and money

You can read the full affiliate disclosure for the detail.

Get in touch

Corrections are genuinely welcome — if a tool is scored wrongly, priced wrongly, or has shut down, I would rather hear about it than not.

contact@theaironin.com — or see the contact page for what is most useful to send.