Disposable app instances
Blink
Five and a half seconds. Three tenths of it is the app.
A site that hands you a running application. You open it on a phone, tap one of five open source apps, and about five seconds later that app is running in your browser: your own private copy, already filled with sample data, not a demo and not a shared instance somebody else has already broken. Ten minutes later it deletes itself and tells you what it cost, which is under a cent.
What you get
Five open source applications: Gitea, Jaeger, Excalidraw, Uptime Kuma and Metabase. Tap one and you get a private instance of it, seeded with enough sample data that there is something to look at rather than an empty install asking you to configure it. Nobody else is using it and nobody else has used it before you.
It runs for ten minutes and then deletes itself. The last thing it shows you is the bill.
Why it is quick
Nothing is installed at launch. Each application is pre-built as a snapshot of a machine that has already booted, already started its services and already loaded its data, and launching is waking that frozen machine rather than assembling one. The work that would normally sit between the tap and the first paint has already been done and paid for.
Swarm
There is a button called Swarm. Press it and twelve real browsers load your instance at once, and you watch all twelve live in a grid with their load times running. It is one instance under twelve simultaneous cold clients, which is a more honest question than a single tab on a fast connection.
The number, and the honesty about it
Every launch is measured and every measurement is public: how long each one took, how often they fail, and how much credit is left in the account paying for the whole thing. A site that shows its own failure rate is making a claim it cannot quietly walk back.
The number is 5.5 seconds from tap to a usable application, and only 0.3 of those seconds is the application. The rest is a real computer waking up and the network carrying the page to you. Quoting the 0.3 alone would be the more flattering figure and the less true one.