I did not set out to build another chatbot. I had used enough of them.
You open a tab, ask a question, get an answer, copy it somewhere else, and then do the actual work yourself. The help is real, up to a point. But when the tab closes, nothing remains. The tool was never really part of your computer. It was a guest in a box, answering through a small window.
I kept coming back to a simple question: what if the AI was not an app inside the operating system? What if the operating system was the AI?
That question became LazySiren. I built it alone, from Oliver, a small town in British Columbia, Canada. There is no team behind a curtain and no large company behind me. It is one person, one computer, and a lot of long evenings deciding what a computer should feel like when the system itself understands what you ask.
The difference is where she lives
LazySiren is an AI operating system, and its AI is called Siren. She is not a chat panel added on top of a desktop. She is the desktop. She sees the same workspace you see, works with the same apps you use, and acts inside that shared space instead of describing, from the outside, what you could do yourself.
That changes the relationship in a practical way. With a normal chatbot, you bring the work to the AI and explain the context again every time. With an agentic OS, the context is already there, because the agent lives where the work happens. You give her a goal in plain language, in English or in French, and she works toward it inside the system.
I will not pretend this is magic, and I will not claim more than what the product does today. Siren can misunderstand. Tasks can stall. That is why she reports where she is in a job, and why you stay in control. The point was never to remove the human. It was to stop making the human do all the carrying between tools.
What a morning in it actually looks like
I open my computer and ask Siren to prepare the day. She starts from the calendar, my notes, and the files I already have open. She opens the five or six usual apps, arranges them, starts with the important tasks, and gives me one view of the day instead of five small windows scattered on the computer.
If I need something bigger, I talk to her. I explain the goal. She works toward it, step by step, using the same tools I would use. I can see every step. I can direct it whenever I want.
If she needs to interrupt, she shows a clear message. She does not sit in a corner and wait for me to notice. She tells me what she is about to do, asks when that is necessary, and continues when I approve. The computer can then keep working while I think about something else, instead of making me decide every time it makes a small choice.
Why she talks at all
Every action I make talks. I type a direction, like: open it, put it on the screen, or turn it into a simple cover. I can change my direction whenever I want, and I do not have to keep my hand on the mouse the whole time. Siren does the actual work quickly and carefully.
She uses a single speech voice, but I can change that voice at any time when my hearing or intelligence prefers a different tone. I can disable spoken output completely, or turn it off for this session only. Everything she says is also shown on screen, first in a single-message view, and then in the running transcript that follows, marked in one place.
The AI who remembers what she did
Siren is the same AI across every app in the system. That means the work can move across the boundary between apps. For me, it started with the moment I understood there is no need to explain everything in order. I can stop in the middle of a task, leave, come back, and find the state intact. When I come back in an hour, or the next day, she picks up where she left off. I no longer have to remind her what she is doing. Time in the system is not lost.
If I ask her to start a task and go away, I can leave the computer and come back calmly. Nothing is gone, and nothing is lost in the conversation. I just ask her what happened, and she tells me.
When I take a flight, she keeps working on the plane. That is what works for me.
Here is what she can actually do
- She uses the system directly, without needing code or hand-crafted workarounds.
- She can design, plan, and create a presentation with images and video in PowerPoint, and then hand it to me for review.
- If a solution does not work, she can try another one, without me having to give a second prompt.
- She can work with me through speech, and so can take control of the system.
- I use my voice to give instructions, but I can also write them or click them.
- She remembers more than one context and can switch between projects in the middle of a task for the right reasons.
What I learned building it alone
I have lived with LazySiren for a long time now. I use it for real work, not as a demo. When something gets in the way, I change it. When I find a dead end, I spend an evening on it.
I rebuilt a lot of the system with Siren herself. She helps me write scripts, clean up the system, and fix the places where the product still breaks. When I am tired in the morning, I ask her to be the one to clean up the system. When I change something in my notes, she knows about it the next time it matters.
That is how I understand it now: it is not a product I finished and shipped. It is a place we use together, and I keep changing it as I keep living in it.
A computer that works like a person
I built something small, but I think it is important. It might be that I am slower than some other products. Some are stronger in some ways, and many are older in terms of age. But in LazySiren, the AI is no longer sitting in a box. She is in the computer, where the work is done. She is the desktop. She sees everything, and then she does it.
I want a computer that works like a person. It starts with what you mean, not with what you type. It handles the small details, and leaves you the choices that actually matter. Siren is my attempt to make that happen.
I am going to keep working on it. If you have spent time with AI apps, you know what I mean: the conversation is somewhere else, the work is somewhere else, and the tool that could really help is nowhere near either one. I wanted to close that gap. That is what LazySiren is for.
Top comments (1)
The morning routine example makes the idea concrete. Starting from the calendar, notes and tasks, rather than from a blank chat box, shows why having Siren live in the desktop matters. The detail about Siren trying another solution without a second prompt is the part that stands out, since that is where most assistants hand the work back. It would be interesting to hear how the offline work on a plane is kept in sync once the connection returns.