Stop Sitting With Your AI in a Chat
I do not sit with AI in a chat anymore. I give it tasks in a tracker, the same way I give tasks to a team. It calls me only when it cannot go on without me. It has been about a month. I have ten AI assistants, each with its own role. I set a task and answer inside the task, often with one emoji. I use AI as much as before. Only the place where we meet has changed. My tracker is Todoist. The tool…
I no longer converse with AI in a chat. Instead, I delegate tasks to ten AI assistants, each with a specific role. I assign tasks through a tracker, such as Todoist, using voice or text commands. The AI only reaches out when it cannot proceed without my input, which has been consistently true for about a month now. I employ ten AI assistants, each handling a distinct task, and communicate via task comments, often using emojis.
The tool I utilize is simply a task tracker, and the hosting method is unimportant. My server is an old laptop that runs a small program every minute. Upon receiving a new task, the program adds a note to the queue, which wakes the dispatcher, or task assistant. The assistant assigns a short title, maintains a personal note, and assigns the task to its owner.
The owner then replies with "Taken" for acknowledgment. Each assistant maintains its own memory for its area, requiring only two lines of communication. AI comments start with a unique tag, indicating whether my intervention is required. Only three tags prompt my response: "For review" (result ready), "Question" (await further instruction), and "To send" (message awaiting approval).
All other comments remain within the task. Each task includes a default prompt, "I propose this," and AI responses are numbered lists with default options. When an assistant seeks my input, the task receives the "Waiting on me" label, and a Slack notification alerts me. If no response is received within an hour, a single Slack message is sent.
There is no second notification. The label "Waiting on me" includes two axes: importance and urgency. Importance refers to financial matters, clients, personnel decisions, safety, and personal goals. Urgency denotes deadlines, today's or tomorrow's meetings, or the presence of a live person. Four filters are preset in the tracker: important and urgent, only important, only urgent, and neither.
I open the "Important and Urgent" filter, which brings me to a list of tasks. I respond with reactions: a checkmark for approval, a cross for rejection, a number one to four for selection, a thinking face for a simplified version, pause to postpone, and an eye symbol to mark as seen without notification. The AI never initiates conversation; all assistants operate under my account, making their reactions appear as mine.
Occasionally, a checkmark is insufficient, necessitating a written response. Sending messages follows a similar process; the assistant composes the recipient and text, which I confirm with a checkmark. The small program then sends the exact text. Messages are signed "Sent by the Spartan, Pavel's silicon personal assistant." If the content changes after my review or is sent to an incorrect channel, it is not delivered.
Upon receiving a task, I get a message summarizing open tasks, pending requests, and ongoing AI activities. Every hour, the program identifies stalled tasks and returns them to their owners. If an assistant remains inactive for over 20 minutes, I receive a notification. I no longer manage individual AI sessions; instead, my AI assistants form a staff of specialists, each with a defined role.
A senior assistant, called the Spartan, oversees new requests and delegates tasks to the appropriate specialists. The Spartan also proposes merging, removing, or adding assistants when repetitive tasks arise. The rest of my tasks are handled by the specialized assistants, who manage licenses, expenses, projects, and more. A senior assistant, known as the Spartan, receives all new and unusual requests and delegates the rest to their respective owners.
Weekly, the Spartan presents me with three options for review. If a task involves discussing with someone, it is added to the agenda for our upcoming meeting. If an email requires a response, it remains with the corresponding assistant. My task list stays clear. If there is no response from the person for a day, I receive an alert: "What do we do?"
After my calls, I do not take notes or make decisions. The assistants handle the work, and I provide only confirmation or rejection through a single tap.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.