What it does
Two things. It explains how the major free assistants actually differ — one always cites its sources, one is faster, one is more direct, one is chattier and more prone to telling you what you want to hear. Then it walks you to a finished prompt you can paste, with your own personal details left as blanks to fill in.
The prompt part is the reason it exists. Most people who bounce off AI don't bounce off the technology — they open a blank box, can't think of anything to type, and close the tab. Handing someone a well-built prompt with three blanks in it skips the hardest step entirely and lets the first experience be a good one.
Why I built it
My dad suggested it. He pointed out that for most people his age the first encounter with AI is either a news story about it taking jobs or a scam call using it — and that somebody should build the version that's actually a good introduction.
The people who most need this are the least served by everything that currently exists. Tutorials are written by people who already understand the thing, for people who already want to use it. Neither assumption holds here.
Design log
No incident log on this one yet — it's early. These are the decisions that shaped it.
I threw out the efficient design
Cascading dropdowns. Pick a field of life, which opens a subcategory, which opens another, narrowing until the site can generate something specific. Dense, fast, few clicks.
It optimised for the wrong thing. Someone who has never used AI doesn't need to reach the destination in fewer steps — they need to never be looking at more than one decision at a time. A screen with three dropdowns on it, two of which are greyed out until you answer the first, is a screen that teaches you you're doing it wrong.
Replaced the whole interaction with a wizard: one question per screen, full width, no visible machinery. It takes more clicks. That's the point.
The spec was already written against the dropdown model, so this meant redoing the content model as well as the interface. Worth it — the dropdown version would have been a site that demoed well to me and failed with the only people it was for.
Blanks instead of personalisation
Generated prompts leave personal details as blanks to fill in rather than asking for them up front.
Asking a first-time user to type their circumstances into an unfamiliar website before they've seen any benefit is a lot to ask, and for this audience specifically, being asked for personal information by a site you don't know yet is a learned warning sign. Blanks mean the site never holds anything about you and the prompt still comes out specific.
Known limits
- The comparison between assistants is a snapshot of a fast-moving field and will need revisiting on a schedule, not once.
- Characterising how different assistants behave is inherently subjective. It's written as impressions rather than claims of fact.
- Not yet tested with the actual target audience — which, for a project whose entire premise is user empathy, is the gap that matters most.
What I'd do next
- Sit with three people over sixty and watch them use it without helping. Everything else is guessing until that happens.
- Cover voice and live conversation modes, where nothing needs pasting at all — likely the easier entry point for a lot of this audience.
- Explore working with an organisation that already reaches these people rather than expecting them to find the site.