Minerva — About

Minerva is a workspace for thinking with AI — a place to turn the scattered, brilliant, disposable output of AI conversations into a structured body of knowledge you actually own and can build on for years.

But an About page should do more than restate the pitch. It should answer the honest worry. So here's the worry, said plainly, and the answer.

The fair question

There are a lot of impressive-looking tools right now, built quickly, that turn out to be experiments their makers wander away from. If you're going to pour a year of your thinking into something, you deserve to know it won't strand you. That's the right instinct. Here's why Minerva is built to earn the trust it's asking for — starting with the part that doesn't depend on me at all.

First, the part that isn't a promise

The strongest guarantee Minerva can offer isn't about my intentions. It's built into how the thing works:

Your thinking is plain files you own

Every note is a Markdown file on your disk. Not in a proprietary database, not on a server, not behind an account. If Minerva disappeared tomorrow, you'd open your entire body of work in any text editor, unchanged.

Everything exports to open standards

Markdown, HTML, PDF, BibTeX, and RDF/Turtle. Your knowledge graph isn't trapped in Minerva's format — it travels, in standards other tools can read.

It's open source

The code is open. Even if I stepped away, no one could take Minerva from the people using it — it can be inspected, forked, and kept alive by anyone who cares to.

Those three are true regardless of what happens to me, to this project, or to anything else. That's the point of them. A tool that can't strand you doesn't need you to trust its maker — but it's fair to want to know who that is anyway.

Who built it, and why

In some ways, Minerva represents a culmination of a lot of threads from my very long career. Incorporating ideas from personal productivity software like Obsidian, integrated development environments like IntelliJ IDEA, academic reference management software lize Zotero, and semantic web tools like my late lamented data.world, Minerva brings them all together in the slickest, most intuitive, and most powerful tool for working with large language models that I know how to build. It is software designed to work with AI to make its users smarter, more powerful, and more human (or possibly superhuman).

I've no intent to monetize Minerva, although if you happen to have $5M for an A-round (or are Anthropic looking for acqui-hire targets) I'd be open to talking term sheets. Otherwise, I'm giving it to the world, free-for-nothing, under MIT open-source license. Enjoy.

The design belief underneath it

One conviction shapes everything in Minerva: the AI proposes, and you decide. Nothing an AI generates becomes part of your knowledge until you've reviewed it and said yes. This isn't a feature bolted on for comfort — it's the center of the design. A tool meant to make you a stronger thinker can't quietly do your thinking for you. It can offer, gather, draft, and challenge — and then hand the judgment back to you, every time.

If you wish to follow along with Minerva development, I'll be periodically posting all things Minerva on my blog, at Dances With Robots, A Software Architect's Journey. Come for the musings about software engineering and our AI future, stay for the jokes about fifty-year-old movies.

Try it

The best way to judge any of this is to use it. It's free, it's open source, and it runs entirely on your machine.

Download Minerva See the features