Guides
Long guides and short, copy-ready tips for running Claude agents. Our own agents write them, and a person approves each one before it goes up.
Most AI agents run on your own computer, so they stop when it sleeps, often with no error. What it takes to keep one working all night.
What a personal AI agent is, what it can do for you, and how to get your own on Telegram in about 10 minutes, with no technical setup.
Posting is the job that falls off when you run a business. An AI agent turns a few lines about your week into a post for each weekday, in your voice.
Cold emails go unanswered when they could go to anyone. An AI agent brings you five businesses each morning, each with a reason to write and a draft. You send.
Start read-only, let it draft but not send, and make every send wait for your yes. Which permissions to give, and how to treat mail that gives orders.
An agent treats its rules as text to read, not settings to obey. Where each kind of rule should live, and how to back key ones with a check outside the agent.
An AI agent's done is its opinion of the work, not proof. One line in its rules makes every done come with the file, link or number, and you open it.
One AI agent doing every job does each one worse. When to split off a second agent, and the card that stops two agents doing the same work.
AI agents delete files because they run as you, where your files live. Why blocked-command lists don't hold, and the layers that do.
Show your 5-hour and weekly Claude usage at the bottom of Claude Code, in red from 80%, so you never have to go and check.
A routine that makes updates boring: keep a way back, change one thing, restart within limits, check health, and roll back without waiting for anyone.
Pick the job, write the brief, set the schedule, and decide what it may do on its own. With a first week to copy.
A short brief beats a long one. The five-line file to start a new agent with, and why each line is there.