Finale: your HQ of three agents
Lesson 22 of 22 in the course “AI Agents for Life”. Free, with animations and a quiz.
From a chat to an HQ
Remember how it all began: Liza was always behind, and AI was just a chat that answered with text. Over six levels came the brief and stop rules, memory and your own skills, connectors, schedules, “if → then” chains, a Telegram bot, a browser agent and calls. And most importantly, an understanding of where an agent is dangerous and when you don’t need one at all. In this final lesson we’ll put it all into one system: your personal HQ of three agents, where you stay the lead.
Why three, not ten
Ten agents sounds cool, but within a month half of them will be sending reports nobody reads. Three is as many as you can realistically keep in your head and check. A good split is by area of life. The secretary handles the day: morning briefing, email, calendar. The treasurer handles money: subscriptions, receipts, price hunting. The mentor handles growth: study, job listings, the weekly review. Each has its own brief and its own minimal permissions. And you’re the lead: you set tasks, read reports and make decisions. Remember the team lesson: a lead without clear assignments means chaos.
7 rules that won’t go out of date
Product names changed before our eyes during this course: buying in ChatGPT moved into shops’ apps, the Atlas browser was shut down, ChatGPT Work and Claude in Chrome arrived. But the rules stayed the same. The five-part brief. A stop before money and any sending. Least privilege, with passwords only in your hands. Checking facts and prices. No lethal trifecta in a single agent. An audit once a month. And the important, heartfelt things: always yourself. If you remember just one slide from this course, make it this one.
The Agentdex: every brief in one place
Over the course you’ve collected 22 ready briefs, from the week audit to the minesweeper and the task sorter. So you don’t have to dig through lessons, the course page has the Agentdex. All the cards are gathered in one place, and each has a “Copy” button. Briefs from lessons you’ve completed are unlocked; the rest are waiting their turn. It’s your starter kit: take a card, adapt it to you and launch it. Good briefs, like good habits, only get sharper over time.
How to keep up
Agents change every few months, and half of what’s in the news today will have a different name tomorrow. To stay oriented, keep four habits. Look up new features on companies’ official pages, not in retellings. Try them on your own real task, not on a demo from a video. Check whether it works in your country and on your plan: a lot launches in the US first. And remember that principles beat buttons. Anyone who understands briefs, permissions and stop rules can figure out any new agent in an evening.
Final task: my HQ
Paste the brief below into a chat. The AI asks you six questions about your week, runs the ideas through the filter from lesson 21 and suggests three agents with cards in the course format. Launch one first, the simplest, and add a second a week later. After a month, do the audit from lesson 20. Done: you have your own HQ, and you’re in charge of it.
Check yourself
- Which agent should start the HQ?
Correct answer: The daily briefing: it repeats, mistakes are harmless, checking takes a minute. It passes all four filter questions. The first agent should be safe and useful every day. - What should change in the setup?
Correct answer: Take away the secretary’s permission to send email. Email plus untrusted messages plus sending is the lethal trifecta. Remove one leg and the trap has nowhere to send the data. - What should she do?
Correct answer: Find the official page and test it on one task of hers. A video shows the best case. The official page says where it works, and her own task shows whether it helps. - Which conclusion is right?
Correct answer: Two help; the mentor should be trimmed or turned off. That’s what an audit is for: keep what saves time and drop what nobody reads.