I built my daughter a tool in half an hour

My daughter and I are doing our first aid certificate at the moment. She needs it for Scouts, and I figured it was a good thing to have anyway, and a nice excuse to do something alongside her. The course is with Business Wise First Aid Training in Hamilton. They're a client of ours, which is honestly part of why I picked them, though the schedule and the location suited too.

There's a fair bit of reading in the course material, and the other day she asked me a fair question: was there a way she could just listen to it instead? She likes podcasts.

Fair enough. So I built her one.

It's called Speaky. You give it some text, a link, or a file, and it reads it back to you as a podcast. Word for word like an audiobook, or with one host presenting it, or two hosts chatting about it like a real episode. It handles about a dozen languages, and the voices are genuinely good now, not the robotic text-to-speech you might remember. It's live here if you want a go: https://speaky.jezweb.workers.dev

That's the whole thing. But the tool isn't really the point.

The point is that building it was a completely reasonable thing to do. Not a project, not a product, not a "let's see if there's a market in this." Just a small tool that does one job, made for one person who wanted it.

And yes, there are a thousand sites that already turn text into audio, and she could probably get close with one of the big AI apps. But the free ones are wall-to-wall ads, and none of them do exactly her thing. That's the shift worth noticing. I didn't have to settle for the closest available option. I could make the exact one, no ads, that does just this and nothing else.

Here's what "reasonable" means now, in actual numbers.

It cost me about $13 of AI time to build, in roughly half an hour. The whole app, the dozen languages, file uploads, the lot, is around 1,450 lines across one backend file and one web page. I went the developer route, because that's my world. But you really don't have to. In Claude's Cowork app (the one I wrote about last week) there's a feature called Artifacts: you describe the tool you want in plain English and it builds a working version right there, no code. I actually tried this while writing this email, I asked Cowork for a simpler version of Speaky, and a few minutes later it had built one I could use and share by a link, no account or key needed at the other end. Lovable, Bolt and Base44 do a similar job, and OpenAI's Codex Sites is the same idea (it even runs on Cloudflare underneath, same as Speaky). This isn't one company's trick. It's the direction everything is heading.

Then there's running it. Speaky costs about one cent per podcast, and nothing at all when nobody's using it. No server humming away, no monthly subscription ticking over in the background. It runs on Cloudflare's edge with the AI built right in, so it sits there costing zero until someone actually presses go.

Built for the price of a couple of coffees. Runs for a cent. For one small job.

That combination is the new bit, and it quietly changes the maths on a whole category of "I wish something just did this" annoyances. Building a tool used to need a real justification: a market, a budget, a business case. Now the bar is closer to "would this make my week a little easier." A reminder that chases unpaid invoices. A thing that renames and files your photos. A page that turns your scribbled notes into a tidy summary. Whatever your version of "can I just listen to it instead" happens to be.

So that's the real lesson from a half-hour job for my daughter. Not that the AI is clever. That the cost of making a small, specific, genuinely useful thing has fallen through the floor, and "I'll just build one" is now a sentence a normal person can say and mean.

Have a play with Speaky if it's useful to you. And if there's a one-job tool you've always wished existed, that wish is a lot more answerable than it was a year ago.

Cheers,
Jez