14 months in. The technical problems are mostly solved. The explanation problem is not.

This is the thing nobody warned me about when I started.

Building the product is hard. Getting it to actually work -- the edge cases, the integrations, the latency -- that's all solvable. Hard, but solvable. You know what you're trying to achieve and you can measure whether you got there.

The explanation problem is different.

Genie 007 is a voice AI assistant that works on any website. You talk to it, it responds, it learns your tone over time. Runs locally. Three seconds to say that. Five minutes to understand why it matters.

That gap between three seconds and five minutes is where all the distribution effort goes. Every bit of content, every landing page update, every cold outreach attempt is trying to collapse those five minutes into something faster.

I used to think it was a landing page problem. Then a copy problem. Then a targeting problem. It turns out it's something more fundamental about how people learn that a tool is for them.

The way I eventually understood it: people don't buy tools, they buy a clearer version of something they already want. My job isn't to explain what Genie 007 does. It's to find the people who already feel the pain I built it to fix and make the connection obvious.

Three types of people feel that pain:

1. People who type all day and their hands or wrists are taking damage (RSI, carpal tunnel). Voice is faster and hurts less.
2. People who think faster than they type and lose ideas mid-sentence. Voice is closer to how the thinking actually happens.
3. People who need to produce a lot of words (support, ops, writers, founders) and want to get the output up without burning more hours.

When I talk to any of those three people, the five minutes drops to about 90 seconds. The product didn't change. The framing did.

Still figuring out how to scale that framing past conversations.

14 months in and that's the real work. Not the code.

Building Genie 007 at genie007.co.uk -- voice AI for Windows, Mac, and Chrome.