07:30, you switch on
You open the lid. There is no desktop waiting for you and no row of icons. There is a voice, a line of text and a question: what do you want to get done today?
You say, "I've got the investor call at ten. Pull together what changed in the numbers since last month, tidy the deck, and move my dentist to Thursday." Then you go and make tea.
When you come back, the deck is open with the changes marked, and the dentist has confirmed. You did not open a spreadsheet or a booking site. The computer did, on your behalf.
What an operating system is today, and why it looks like that
An operating system is the layer between you and the hardware. It runs the screen, the keyboard, the memory and the storage, and it gives you a place to put things. On a laptop, that place is a desk: windows you arrange, files you keep in folders, apps you open one at a time.
That picture was drawn for a different era. When the Macintosh arrived in 1984 and Windows 95 in 1995, computers had very little memory and no way to understand ordinary language. So the machine could not meet you halfway. You had to learn its way instead: where the file lives, and which menu hides the button.
The desktop was a clever translation, a way to make a rigid machine feel familiar. It worked brilliantly. But look at what it still asks of you. You are the one who knows the figure is in one app and the deck is in another, and you are the one carrying things between them.
10:45, the desktop becomes a conversation
The call went well. You say, "Send them the deck, and put a follow-up in for Friday."
You do not drag an attachment into an email or hunt for the right contact. The computer knows who was on the call, picks the version you just finished, writes the note in your voice and asks one thing: "Send it now, or after lunch?" Windows still appear when you want to see or edit something. You just stop managing them all day.
13:00, apps become things the computer can do
At lunch you say, "I need a table for four near the office, somewhere quiet, Thursday."
Nothing opens. Behind the scenes, your assistant talks to a booking service, checks your calendar, and comes back with two options. The restaurant never had to build a clever app for you. It had to make its bookings available in a way a machine can use.
This is the bigger change in what an app is. Today an app is a destination you visit. In an AI-first system an app is a capability: a set of things it can do, such as book, pay, look up, convert and send. You will still have brands you trust and prefer. You just will not spend your day tapping through their screens.
17:40, your files become memory
Late in the day you say, "What did we agree with the supplier last spring about delivery?" You do not search for a filename. The computer finds the email thread and the contract clause, and gives you the answer with the sources underneath so you can check it.
Your files stop being a heap of folders to keep tidy and become something closer to memory, ready to be recalled by meaning. Because that memory is so personal, it has to stay yours and private, and you have to be able to see what the computer did with it and undo it.
19:00, is this really where computers are heading?
This is where things are heading, and the finished version is not here yet. Assistants that can write and operate other software exist now. They are also still uneven: they get things wrong, they need careful permissions, and nobody has fully settled how much they should do before checking with you.
But the pieces are arriving together. Models understand ordinary language, and tools let them use other software. The companies that make phones and computers have all been adding assistants to their systems. The day above is where that road leads, and it will not arrive in one go. It will creep in, one task at a time, until the old way feels slow.
20:30, what this means for people who build software
If an assistant is going to use your product, then your product has a second kind of user, and it is a machine. It does not see your beautiful screen. It needs clear, well-documented ways to ask for something and to understand the answer.
So the unglamorous work gains weight: a clean interface for machines (an API, in the jargon), predictable names, sensible errors, and actions that are safe to repeat. A well-built service becomes something an assistant can use confidently. A messy one is likely to get skipped, and your customer never even sees that you lost.
None of this makes design less important for people. Screens still matter at the moments that count, such as checking a payment or approving a change. The skill is building both well: one door for people, one for machines, behind the same solid service.
22:00, what it means if you are founding something now
The good news is that you are not starting from a pile of legacy habits. If you are building a product today, ask two questions early. If someone's assistant wanted to do this job for them, could it? And what is the one thing we do so well that an assistant would choose us?
Build the core as a proper service first, with the screens on top. It is sound advice for any product, and it happens to be what an AI-first world rewards. It also keeps you flexible while the platforms settle.
Someone will switch on a computer, say what they want, and some product will be the one that does it. Build yours to be that one.