Hey HN,

About 10 years ago, before I had any real experience in the industry, I kept finding myself with ideas and things I wanted to teach. I'd write blog posts. Unpaid, unstructured, no audience. I always knew I wanted to turn that into something more, like a proper course, but every platform I looked at wanted me to pull out a camera and record myself. I never did. The ideas just sat there.

Fast forward to now. I've got 10+ years building products, and I finally built the thing I wanted back then. Lesso (https://lesso.app) is a course platform designed specifically for people who teach through writing. No video, no camera. You write lessons, organise them into a course, put a price on it, done.

There's also a Substack import. Paste your Substack URL, it pulls your posts in, you drag them into a lesson order and publish. Most newsletter writers don't realise they're already sitting on a course, it's just scattered across their archive in chronological order instead of structured as a learning path. I added a PDF import feature, completely client-side, and a markdown formatting feature too.

Whenever I decide to make a change, I ask cursor, and then push it straight to prod, no staging environment, just raw speed. I want to get features shipping as fast as humanly possible, if something breaks, I fix-forward.

Creators keep 85% of revenue. No monthly fees. No feature tiers.

The build:

I built the whole thing in 2 days using Cursor and Claude. The stack is React, Supabase, and Vercel. I barely touched the code. I put my product hat on, described what I wanted, and let the AI build it. The loop was: describe a feature, AI builds it, test it, tell it what's broken, it fixes it, repeat. I was making product decisions, not engineering ones. That's the bit that surprised me most. 10 years of building products and the skill that mattered here was knowing what to ask for, not how to build it. Claude handled the non-code side too. Business model validation, legal docs (privacy policy, terms, cookie policy), compliance research. I registered as a sole trader in the UK, ICO registration was £52, and that was my entire legal spend. The AI-generated docs aren't lawyer-reviewed, but they're good enough to launch with. I'll get a proper review once there's revenue worth protecting.

Stripe for payments. The whole thing was live in 2 days.

What I got wrong:

I assumed the hard part would be building. It wasn't. The hard part is everything after.

I designed an affiliate programme before I had a single user. Creators keep 85%, affiliates earn 50% of the platform's net cut, 90-day cookie, $10 minimum payout. I thought this was such a good offer that affiliates would be lining up. Nobody was. I hadn't thought through the basic problem: why would anyone promote a platform with zero courses on it?

I spent days doing cold outreach on X, searching for people who might want to create courses, DMing them, posting in founder threads. Barely any traction. An empty platform is almost impossible to pitch, no matter how good the economics are. So I wrote the first course myself. It's about this whole process, building and launching a SaaS with AI, the real version including the parts that didn't work. At least now the platform isn't empty. Whether that actually moves the needle, I genuinely don't know yet.

Where I am now: Sole founder, pre-revenue, very early days. The product works. The first course is live. I'm figuring out distribution one conversation at a time. The honest reality is that AI compressed the building part to days, but it can't compress the part where you earn people's trust. That still moves at human speed.

Happy to answer questions about the build, the AI workflow, the business model, or the cold start problem. Especially the cold start problem, because I'm still in the middle of it.