Why I built Madless
A short note on what Madless is for, what it isn’t, and why I built it on my own.
I built Madless because I was tired of AI that thinks for me.
The web is full of tools that summarise, draft, decide, suggest. They’re fast. They’re fluent. And they do the one thing I don’t actually want done for me — the thinking.
What I wanted was simpler. A place where my own ideas live. A place that helps me see them. Not a chatbot pretending to be smart. Not a wiki I have to maintain. A canvas I can drop a thought into, and a quiet system in the background that ties it to everything else I’ve ever written.
That’s the substrate. The thinking is still mine.
What Madless does, in one paragraph
You capture — articles, voice notes, screenshots, files, anything that catches you. The canvas connects them. You draw the connections that matter; the system suggests the ones you missed. Once a week, on Sunday afternoon, Mirror reads everything you wrote that week and writes back. Patterns. Contradictions. A short note in your own voice on what your thinking did this week. Not a notification. Not a feed. A letter you open when you’re ready.
That’s the whole product line. Capture, Canvas, Minds, Mirror, Memory. Each one earns its place.
What Madless doesn’t do
It doesn’t chat with you. The chat you want — interactive, fast, frontier-grade — lives in whichever AI you already pay for. Madless makes itself readable to that AI through MCP, and through a Memory layer your AI inherits. The conversation belongs to the model. The substrate belongs to you.
It doesn’t notify. It doesn’t nudge. It doesn’t ping. The product never asks for your attention; it waits for you to bring it.
It doesn’t train on your data. It doesn’t sell ads. It doesn’t sell your data. None of that ever happens. The product is paid because it has to be — running the substrate costs money — and being paid is what lets it stay that way.
Why I’m not raising
I’ve thought about taking VC money for this. I’d be a fool not to. But the moment I take it, the product has to behave differently. It has to grow at a rate that requires me to do things I don’t want to do — ads, retention loops, attention extraction, things that make the product less yours and more theirs.
So Madless is independent and self-funded. It will grow slowly. That’s the price of building something I actually want to use, and it’s a price I can pay.
The closest thing I’ve ever loved in this space is are.na. I want Madless to feel like that — quiet, weird, durable, in your hands. Different shape, same spirit.
The Mirror cadence
The thing I’m most proud of, today, is the cadence.
Most AI products try to be in front of you constantly. Madless does the opposite. The synthesis happens once a week, on a premium model, on a quiet Sunday. The rest of the week, the substrate is silent — embedding what you capture, connecting it to what you already wrote, getting itself ready for the next reflection. You don’t see any of that. You see the Sunday letter.
Rare and good beats often and cheap. That’s the bet.
What’s next
Today, Mirror writes one report a week and one Portrait — a standing read of what your thinking is about, organised by theme, named from your own words. That’s v1.5. The roadmap from here is depth, not breadth: better synthesis, more languages, deeper Minds, the released-scope Portrait that makes your /minds/handle page feel like you.
There’s a one-page vision under /welcome if you want the longer version. The pricing is two plans, $15 and $29 per month, both with a 30-day free trial — no free tier on purpose. Free tools have to find some other way to pay for themselves; this one doesn’t.
If any of this resonates, you’re the person I built it for.
— Slawek