An AI dashboard, and the number that is missing from most of them

Every figure has an age and most dashboards hide it

A price on a screen is a price from a moment. Without that moment shown, the reader supplies their own assumption, which is almost always that the number is current. Generated dashboards inherit this because the prompt says show the price and nobody says show when it was read, so the model does the thing it was asked.

The failure that looks like success

The source stops answering. The last value is still in memory, still on screen, still the right shape and colour. Nothing about the page changes. This is the worst failure a dashboard can have because it is indistinguishable from working, and it will stay indistinguishable for as long as nobody reloads and compares.

What to ask for instead

Say what should happen when the data is missing. A dashboard that renders an empty state and says the source did not answer is more useful than one that shows a stale figure confidently. Most models will do this correctly if asked, and almost none will do it if not, because a clean happy path is what the request described.

Refresh is a claim, not a setting

Polling every thirty seconds is a promise that the number is at most thirty seconds old, and it stops being true the moment a request fails silently. If you set an interval, also show the time of the last SUCCESSFUL read rather than the time of the last attempt. Those diverge exactly when it matters.

How to check one in ten seconds

Load it, note a figure, and cut the network. If the number stays and nothing on the page says anything, the dashboard cannot tell you the difference between a quiet market and a broken connection. The figures below this page carry the time they were read for the same reason.

Live, right now, on this page

Read from a live market as this page rendered, through a public endpoint that needs no key. Read at 2026-10-22 14:53 UTC; accurate as of that time and not afterwards.

Common questions

See the number itself

Pages that read a live market and show what they read, updated while you look at them.

• Live FX rates, and the three prices people mix up, a small live board, four pairs, showing what a dashboard has to keep current.
• Prediction market odds, live, the same idea on a different feed, questions priced as probabilities.
• Nvidia stock price, live, with the tape underneath it, the accelerator the whole AI build out runs on, with its realised volatility beside the price.
• Google stock price, live, and what GOOGL actually is, an ad business being repriced as a model business, which is what the volatility is measuring.
• Micron Technology stock price, live, with memory-cycle context, the memory that ships next to every accelerator, priced continuously.
• Meta stock price, live, with the week's tape, capital expenditure guidance moving a stock more than revenue does, visible in one week of candles.
• Marvell stock, live, through the tokenised share, the networking silicon between the accelerators, usually missing from the AI trade.
• AppLovin stock, live, through the tokenised share, a company valued on how well a machine learning engine targets, disclosed only through results.