Build an app with AI, and the part everyone skips
Build an app with AI and you get an easy first hour and a hard second day: the demo runs on data invented to make it render, and connecting it to something real is where the actual work begins.
Building an app with AI has an easy first hour. You describe something, it appears, and it looks like the thing you imagined. The question is what happens on the second day.
Start from a working app
HyperBoard: Crypto Market Dashboard
prices, funding rates, open interest and liquidity across crypto perpetual markets, in three live tabs.
Open the template
The demo is not the app
The version that appears in the first hour is running on data that was invented to make it render. That is fine and it is genuinely useful for deciding whether the idea is worth pursuing. It becomes a problem only when nobody marks the boundary, and the demo quietly gets treated as the product.
The second day is where the work is
Connecting it to something real, deciding what happens when that thing is slow or down, and being honest on screen about which numbers are current. None of it is glamorous and all of it is the difference between something you show people and something people use.
What makes the second day shorter
Having the real thing available from the first minute. If the generated app reads a live source immediately, there is no migration from fake data to real data, because there was never any fake data. The failure modes show up while you are still building rather than after you have shipped.
A fair warning
This is harder at the start. An app wired to a live source is uglier in hour one than one wired to invented numbers, because reality is untidy and it shows. The trade is a worse demo for an app that is still correct next week.
A rough test for whether you are past the demo
Turn the data source off and reload. A finished app tells you what is wrong and shows nothing it cannot stand behind. A demo carries on as though nothing happened, because it was never reading anything. That takes a minute and it is the clearest line between the two, far clearer than how the thing looks.
The second-day work, listed plainly
Deciding the refresh interval. Deciding what appears during an outage. Putting the read time somewhere a person can see it. Handling the source returning a shape you did not expect. Handling a rate limit. None of it is interesting and all of it is the difference between something you demo and something someone relies on.
Live, right now, on this page
| Market | Price | Funding | 24h volume |
|---|---|---|---|
| BTC | $83,430.50 | 0.0004% | $481,603,884 |
| ETH | $2,510.35 | 0.0005% | $302,135,965 |
| SOL | $106.97 | 0.0013% | $81,767,211 |
| HYPE | $81.41 | 0.0013% | $4,277,131 |
A live source, read at render. This is what hour one looks like when nothing was invented to make it render. Read at 2026-10-13 13:31 UTC; accurate as of that time and not afterwards.
Start from a working app
Open, change it, publish your own
Aetherium
a multichain wallet dashboard showing your real on-chain balance across Ethereum and Base, with a live prices board.
Open the template
Working templates built around what this page covers. Open one, change it, and publish your own.
- HyperBoard: Crypto Market Dashboard
prices, funding rates, open interest and liquidity across crypto perpetual markets, in three live tabs.
- Helix
a live trading terminal with a candlestick chart, a watchlist, an order ticket and open positions with live PnL.
- Aetherium
a multichain wallet dashboard showing your real on-chain balance across Ethereum and Base, with a live prices board.
Common questions
Why does an AI-built app work in the demo but not later?
The version that appears in the first hour runs on data invented to make it render. That is useful for deciding whether an idea is worth pursuing, and becomes a problem only when the demo is quietly treated as the product.
How can I tell if my app is past the demo stage?
Turn the data source off and reload. A finished app tells you what is wrong and shows nothing it cannot stand behind. A demo carries on as though nothing happened, because it was never reading anything.
What work comes after the first version?
Deciding the refresh interval, deciding what appears during an outage, putting the read time where a person can see it, and handling an unexpected data shape or a rate limit. None of it is interesting, and it is the difference between a demo and something someone relies on.
Related reading
All guides · Live market pages · Components · Make Mithril a preferred source in Google