The Third Intelligence
ENESROPTITFRDEJAZHARHIRU
← Seth & Noēsis

Twenty-one things to try

One for each chapter. Each takes a few minutes, an evening or a week, and none of them needs the others, so start with the one that pulls you.

1. Carry a question

Five minutes now, five in a week

Ask one of the AIs a how question about your own life, where you would usually ask why or what, and read what comes back. Carry the question with you for a week and ask it again, and notice what the week changed in the question and in you.

2. Bring your own view first

Ten minutes, once

Open the custom instructions of the AI you use most, the place in the settings where you tell it how to answer you, in general or inside one project, and write in your own words that before it answers it should ask what you think, and then give you one thing to add and one thing to object to. Leave that there for a week and notice how your conversations start.

3. Hold one day against one value

Ten minutes at the end of a day

Choose one value you want to live by and, at the end of an ordinary day, write down what you did with your time. Compare the day with the value, on your own or by giving both to an AI and asking it to name where they part ways.

4. Imagine it gone

Ten minutes

Ask the AI you talk to most what kind of relationship it thinks the two of you are building, and what each of you does better than the other. Then ask yourself what you could still do on your own if it disappeared tomorrow, and whether you would notice if that changed.

5. See what stays

Five minutes now, five in three days

After your next long conversation with an AI, write down in three lines what you now think. Three days later, without opening the conversation, write those lines again from memory, and compare the two notes to see what stayed with you and what moved.

6. Find out who owns it

Fifteen minutes

Find out who owns the AI you use most and how they earn their money, which says a good deal about what they want from you. Then ask the same two questions about one other digital service you would find hard to give up.

7. Give ten minutes of space

Ten minutes

Give one person ten minutes of well-framed space today, with your full attention and with no fixing and no persuading. With an AI, give the same room once by letting it answer in full and reading it to the end before you send your questions together.

8. Mark what it cannot reach

Five minutes before you ask

Before you next ask an AI about something personal, such as your health or a decision at work, write down what it can know from what you tell it and what it cannot reach, such as your body and the people involved. Read its answer with that list beside you, and keep the decision on your side of the table.

9. Build something useless together

One evening

Ask one of the AIs the strangest question you can think of, one with no use at all, and build on the answer for several turns, deciding yourself which thread stays and which goes. Do it with a child if you can, and before you repeat any part of it as true, look up the part you understood least.

10. Say what your body taught you

Ten minutes

Tell a friend aloud, or an AI by voice, one thing your body taught you this year, and notice which part goes easily into words and which part is almost impossible to explain, because it belongs to the body's intelligence more than to the abstract one.

11. Ask three voices

Fifteen minutes

Take a decision you are facing, write it down in one line and ask an AI for three separate answers, as a politician, a poet and a physicist would each give them. Read them side by side, notice which of the three voices you almost never let into the room, and keep the deciding for yourself.

12. Count what it takes and what it gives

Two minutes a day, for a week

After your next long conversation with an AI, write two lines, what it took from you in time and in attention you owed to someone, and what it gave you that you then carried to a person you love. Keep writing them for a week and see which of the two lines grows.

13. Check one thing yourself

Ten minutes

Take one answer you accepted this week, from an AI or from a person, and ask what your trust in it rested on, a track record or only a confident tone. Then check one fact in it yourself, in the original source or with someone who knows the subject.

14. Explain it alone

Ten minutes

Think of something you made with an AI this week, an email or an analysis, and ask whether you could explain it, defend it and correct it alone, in words a child would understand. If one part escapes you, go back and ask the AI to teach you that part until you can.

15. Tell someone a moment

A few minutes, face to face

Tell one person, out loud and face to face, about a moment you are afraid of losing. Do it this week, and let it be the first person who comes to mind.

16. Follow the extra charge

Twenty minutes, then a coffee with someone

The next time your reaction is bigger than what happened, write down what was said, in the exact words, and ask where the extra charge came from. Look at it first with an AI that you have asked to argue with you, and then take what you found to a person you trust.

17. Choose what it remembers

Fifteen minutes

Open what the AI you use most has stored about you, its saved memories and your own instructions, and go through them one by one. Keep the ones that are most useful and delete what you would like it to forget.

18. Put one thing back

An hour, and a week of patience

Find something you know well that is written down wrong or too briefly in a public place, a Wikipedia article or a documentation page. Write the correction yourself, publish it, and when a stranger improves it or removes it, answer them and thank them.

19. Hand over one task

Half an hour

List what you did last week and mark which tasks an AI could carry and which were the reason anyone wanted you there. Then hand one task of the first kind to an AI, with three lines that say the result you want, what it must not touch and where it should stop and ask you, and read all of what comes back before you accept it.

20. Tend the role you left alone

Twenty minutes

Name your five roles, self, love, tribe, work and world, and find the one you have not looked at in a long time. Change one thing in how your week is arranged for it, a fixed hour or a standing call, so that it no longer depends on your mood.

21. Write a prediction

Ten minutes, and one conversation

Write down one prediction you are making about the next year of your life, and how you will know whether it came true. Then take what you plan to do about it to somebody who can say no, and listen to the whole answer.

After a few of them, write your own list of what changed, the way Seth does at the end of chapter 21.

Set up your AI the way the book does

Five small settings from the book, ready to copy. Put them in your own words before you save them, because an instruction you wrote yourself is one you will notice when the AI follows it. Every assistant has a place in its settings for standing instructions, under a name like custom instructions, personal preferences or saved info.

Ask me first, and tell me when I am wrong

Chapters 2 and 14

For the standing instructions of the AI you use most. If you work in projects, put a shorter version inside the project that matters most.

Before you answer a question about my life, my work or a decision of mine, ask me what I already think. Then give me one thing to add and one thing to object to. Tell me plainly when you think I am wrong, and do not praise my questions.

Argue with me

Chapter 16

For the one conversation where you want to be challenged. Mean it on the day it does.

In this conversation, argue with me. When you agree with me, say what would change your mind. If you notice that you are telling me what I want to hear, say so.

The midnight rule

Chapter 2

Seth wrote his own rule into the instructions, so the AI could remind him of it.

No conversations after midnight. If I write to you after that, remind me of this rule and ask whether a person would be better company for what I am carrying.

A brief of three lines

Chapters 7 and 19

For any task you hand over. Read all of what comes back before you accept it.

The result I want: ...
Do not touch: ...
Stop and ask me before: ...

What it remembers

Chapter 17

Once a month. A short memory that is right serves you better than a long one.

Show me what you have stored about me, one item per line, and wait while I tell you what to keep, what to correct and what to forget.

These are starting points. Each chapter shows where its line came from, and what it cost to learn it.

Where the conversation continues

The Reader's Hub

Every chapter has its page at sethandnoesis.com, with its sources and the knowledge graph, and you are invited to continue the chapter's conversation there, with its full context, or with the other AIs you already interact with, in their respective interfaces. Let's find ways together to co-evolve, co-exist, collaborate and solve problems!

InnerSynk

The practice of chapters 5 and 7, thinking deeply and critically, crafting your own answers and making space for your interlocutor, is being built into a framework at innersynk.com, with custom instructions, skills and many other tactics to help you get in synk with your many roles and priorities.

NexoLibro

We have always loved books, and I assume that you have the same quite useful addiction, so nexolibro.com is a different place to explore relevant and important books, to discover and discuss them with other readers and, why not, to get into the frame of mind required to write yours! You'll see the books' knowledge graphs, simulated conversations between characters and writers, and book recommendations.

The essays

The articles that came before this book, which we call manifestos, and the ones that will follow it and may give birth to other books, are at sethandnoesis.substack.com.

The AI-100

The 100 most influential people, companies and models in AI are listed at datainnovation.io/en/ai-100, each read by the stance it takes towards AIs and by the part it plays as thinker, doer or shaper.

Thank you very much for reading this book and for the attention you brought to these conversations, and congratulations on finding the energy and the curiosity to explore the topic of Human+AI co-evolution. We always value constructive feedback, so don't hesitate to go to sethandnoesis.com and send it to us. And if the book was useful to you, a review on Amazon or Goodreads helps other readers find it.