While I sleep: the system catches every incoming
Sergei Pak · designs AI operating systems · русская версия
An email lands at 3:12. A document from a counterparty drops into a work chat on Sunday. A court case gets a new filing on a holiday, when the court is supposedly closed. I'd have missed every one of these – I'm not at the screen. The system misses none.
What "always online" means
My system isn't a browser tab I have to remember to open. It's a process that runs on a server around the clock, watching every channel where business data arrives: mail, chats, document feeds, the court registry. A new signal shows up, and it's caught that same minute, not whenever I next get to my laptop.
For a person, incoming passes through one throat: one pair of eyes, working hours, weekends, holidays, sleep. An important email gets stuck between two unimportant ones and sinks. The system isn't bound by that throat. At two in the morning it reads as carefully as at noon, and it doesn't get tired at the two-hundredth message.
What happens to an incoming
Catching is half the job. Then the system reads the signal and works out what it is: a client email, an invoice, a court update, or just noise. Then it decides what to do. Routine it closes itself: reconciled the invoice against the contract, put the court filing in the calendar, filed the boilerplate email away. What it can't close without me, it sets aside in a short list for the morning.
By six I wake up not to a hundred unread, but to three lines: this needs your call. Everything else got handled overnight.
Where this saves you
The most expensive failures in business aren't loud mistakes, they're quiet misses. An email that sat unread for three days, and the deal went cold. A document that arrived over the weekend and surfaced only on Monday, when the deadline was already burning. A court move spotted a week later than it should have been.
They all grow from one place: incoming depends on whether a person managed to look at it in time. The system removes that "in time". The signal is caught and read always, whether I'm at the screen or asleep.
What it means for your business
When someone shows you an AI system, ask one plain thing: what happens to an email that arrives at three in the morning on a Saturday. If the answer is "it waits until someone opens the inbox", that's not an operating system, it's another tab. The working answer is different: it gets caught, read, and either closed or placed on my desk by morning.
What such a system is built from, and why it starts with context rather than a model, is in the post on what AIOS is. And how many hours a day it carries without me, I measure in hours of autonomy.
FAQ
- Does the system really run around the clock?
- Yes. It isn't an app I open, it's a process on a server that never shuts off. An email, a chat, a document, a new filing on a court case – each gets caught the moment it appears, not when I next happen to check my inbox.
- What if something urgent comes in overnight?
- The system reads the incoming right away, figures out what it is, and decides: handle it, file it, or wake me. Anything genuinely urgent reaches my phone immediately. The rest waits in the morning digest, so I'm not pinged over nothing.
- Is it safe to give the system access to everything incoming?
- Catching and reading is not the same as acting. Anything irreversible (sending money, replying to a client, deleting) waits for my tap. Most of the reading runs on a local model, and only a trimmed slice goes out. More on that in the post on local models.