How I got here

In 2020 I sold sunglasses.
By 2026 I was building the websites that sell them.

Five years. Four stores. One year of trying to build the AI. One good idea.
Scroll for the long version.

  1. 2020Sixteen

    I sold sunglasses on the internet.

    Bluelight glasses, technically. I called the store EyeShield.

    Built it on a Saturday from my bedroom.

    Made my first sale at 2 a.m. the next week.

    I called my mum. I thought I'd built a real business.

    EyeShield, 2020. I was sixteen. I thought "Stay Protected in Style" did the selling.
  2. 2020Still

    What I'd actually built was a pretty website that couldn't talk back.

    Customers came. Most left with questions, not glasses.

    The same five questions. Every customer. Every day.

    Do they fit small faces? Work over prescriptions? Real glass? Ship to Malta? What if they don't fit?

    The site couldn't answer one of them.

    So I did. By hand. At midnight. From bed.

  3. 2021Repeat

    So I built another one. Then another. Then another.

    Tea. Then headphones. Then watches.

    Each one prettier. Each one with better photos and better writing.

    Same problem every time.

    The site never answered the questions. The owner did.

    2020EyeShield2021Tea2021Headphones2022Watches
    Four stores, three years. Same idea, different product. Same problem every time.
  4. 2022Read

    I started reading instead of building.

    Books on why people buy. Why they leave. Why they come back.

    Everything I read pointed at the same thing.

    It wasn't the ad. It wasn't the product.

    It was the page itself.

    The page is a list. Customers don't want a list. They want an answer.

    Customers don't want a search bar that returns 47 cards. They want someone who knows the answer to talk to them.

    — Notebook, late 2022
  5. 2023ChatGPT

    Then ChatGPT happened.

    Every store rewrote their product descriptions in four months.

    They missed the point.

    Better writing on a page is still just a page.

    What customers wanted was a person who knew the answer to talk to them.

    Not a paragraph. A person.

  6. 2024Bet

    I bet a year on it.

    Quit running stores. Spent a year learning how to build the AI.

    First six months — everything I made was bad.

    Next six — less bad.

    I threw out three versions that looked great but lied to customers.

    I kept the one that didn't.

  7. 2025One thing

    Built one good thing.

    An AI concierge — like a really good shop assistant who never has a bad day.

    Greets every visitor.

    Knows the products, prices, policies, hours.

    Walks them to the answer in 90 seconds.

    Asks for help when it doesn't know.

    Ran it on a friend's tile shop in Malta.

    6.4s → 0.8s
    Page-load on mobile · friend's tile shop · day 11 of an 11-day build
  8. 2026Now

    Now I build it for other people.

    Two builds a month. Fourteen days each. Just me.

    Same idea every time — the website I wish I'd had at sixteen.

    A concierge at the front door. The visual site as the back-up.

    Customers walk in, get answered, leave happy.

    Owners stop answering the same five questions at midnight.

    2026. Same product page. Now it talks back.