prompts.

JACK
REIMON
2023

ARTICIAL INTELLIGENCE AIDED MAGIC

“Any sufficiently advanced technology is indistinguishable from magic.”
Arthur C. Clarke

‘or is it?’
 warning.

I suggest you pay for ChatGPT 4.0 instead of the free 3.5.

This might not be future-proof; we don't know what changes might happen
in OpenAI and AI that could benefit or not the prompts described.

This requires certain magic knowledge; this is not to teach you a magic
square or mnemonica, or peeks or false shuffles. It's assumed that you
already know some of this or know where to get that information.

And please, there are limits to this, so don't be a wiseass. I've put the effort to
pull this out to explore a new way of doing magic and using AI in your
repertoire. If you are really smart, you know the potential of this.

MORE ON DECEPTIONEER.COM
Chapter n.1

intro.

A BIT ABOUT THIS BOOK, THE HISTORY AND THE KIND

OF MAGIC POSSIBLE.
This is a book about
“moves”.
1. Introduction

The Turk was a big wooden cabinet, featuring a chessboard on top. Created in 1770
by Wolfgang von Kempelen, an Austrian whiz, this intriguing contraption had a
humanoid figure in Turkish garb seated behind the board. Not only did it look like it
could play chess, it seemed to strategize and even win against human opponents.

Inside the cabinet was a hidden compartment for a human chess master. This person
controlled the figure with levers, magnets, and other gizmos. The Turk duped people
—big names like Benjamin Franklin and Napoleon Bonaparte—into thinking a
machine could play chess. It stirred the public's imagination and inspired the fields of
automata, computing, and artificial intelligence.

This nifty book dives into that kind of hidden compartment, revealing the secrets
behind the "magical machine."

What I dig here is the doorway to more than just tricks. Good tricks, yes, but the ideal
outcome is to spur you to invent your own.

The book breaks down into two big chunks: Technological Magic and Human Magic.
Not married to the titles, but my brain juice is going into writing prompts, not headlines.

For most tricks, you'll need some extra juice. AI can help us pull off incredible feats, and
maybe even define a new genre of magic, but you'll also need to use forces, peeks,
maybe a deck switch, and some memory gymnastics.

Some tricks play out like two-person telepathy (weird, huh?), or pre-show work. You're
either "forcing a specific answer" from ChatGPT or concocting a secret code only the
two of you get. It is like the GPT version TOXIC. In the end, Prompts are like modern
“magic words” for artificial intelligence.
2. Powers

There are two big "powers" at play here. First, leveraging ChatGPT to become the
"almighty OZ." Second, and more tantalizing, is attributing the power to the machine—
Turk style. This raises all sorts of juicy, philosophical questions.

First off, is Clarke's famous quote still on point? In this tech-savvy world, people might
already ascribe god-like powers to AI. Tech experts, aware of AI's limitations, find these
feats especially intriguing.

If AI's computational power is limited, how can it know things it shouldn't? It's like it has
a hotline to the divine or something. When psychics do this, we're cool with the idea
that human consciousness can tap into something greater. A cosmic bridge, if you will.

OH and btw, yes, this was corrected by ChatGPT.

Chapter n.2

setup.

READ THIS BEFORE DIVING INTO THE MAGIC. IT WILL

HELP YOU AVOID FUTURE PROBLEMS.
N.1 Setup
Go to chat.openai.com and set up your account. While the free GPT-3.5 is good,
upgrading to the $20 - 4.0 version is better at understanding prompts and performing
tricks. Also, 4.0 is faster and more reliable. Sadly, there is a limit of 50 prompts per 3
hours, although during my test, it seemed that 1 hour was enough wait time.

If you're new to ChatGPT, kudos for diving in! Once your account is set up, play
around. ChatGPT can draft articles, emails, and more. The art of asking it the right
questions has even spawned a new field: "prompt engineering," where experts craft
prompts for specific responses.

Now you can go and download the official app for your phone and enter your
account.

Note: Will these work on other AIs? Maybe. I prefer ChatGPT over Google's Bard for its customizability
and mainstream appeal.

N.2 Custom Instructions

The cleverness in these tricks isn’t just in the prompts, but in their placement. It’s like the
hidden cabinet in ‘the Turk,’ a secret compartment where you can stash commands to
open your ChatGPT account on your laptop or mobile and find it ready and waiting—
quite sneaky!

Now, let’s delve a bit deeper.

For those familiar with ChatGPT, you know you can initiate ‘new chats’. These usually
start from scratch, with no prior knowledge of you or your inquiries to form complex
answers. In a new chat, I can input instructions, codes, or expected behaviors, and GPT
will respond accordingly. So the prompts here work in that specific chat. But for me, this
wasn’t enough. The remnants of previous texts hinted at something amiss. It’s not terribly
off-putting — I could always ask ChatGPT to generate any random text to mask my top
prompt — but it wasn’t as fresh as a new clean chat, especially when opening the app
on your phone.

Enter ChatGPT’s recent development: ‘custom instructions.’ This feature lets you
dictate ChatGPT’s behavior across ALL chats. You can find it by clicking on the three
dots in the lower left corner inside the chat (next to your email). You can input
instructions on how you want ChatGPT to respond saving you from repeating this
information in every new chat. In essence, custom instructions are your shortcut to
efficiency and speed and in this case, deception.

In the world of MAGIC, this means I can ‘hide’ my instructions. It’s like a pre-show with
ChatGPT—I can conceal my codes and desired responses. And this is precisely where
we will paste our prompts. But here’s the catch: you only get 1500 characters for
custom instructions. That’s basically one trick per custom instructions. This is the
current state of play, but who knows, OpenAI might surprise us tomorrow!

Opening the custom instructions reveals two information fields (image). The top one is
about you—I prefer to leave this blank; I don’t want to add thing like “im a magician”
because then ChatGPT could reply in a way that can reveal something is ‘wrong’. The
bottom field is where you input how you want ChatGPT to respond, and that’s where
you’ll paste or write your prompts.
When crafting prompts, I start fresh in a new chat, iterating quickly with ChatGPT’s
help until I get the right answer. It involves a lot of back-and-forth, but it’s essentially a
practice in iteration. Once you’ve perfected your prompt, paste it into the custom
instructions, and voilà, you’re ready to proceed with your effect in a fresh new chat.

This all works smoothly on the app, even adding the custom instruction, although i
prefer setting things up on my laptop. It's just a weird superstition.

N.3 Longer Instructions

"Presently, the limitation regarding the size of custom instruction implies that pasting
more than one prompt in custom instructions is likely unfeasible, restricting you to
performing perhaps only one trick in this manner (depending on the prompt you
choose).

If the aim is to be ready for more than one trick, an additional prompt can be initiated in
a 'new chat', which is effective for that specific chat and will stay there in the ‘history’ tab
for easy access. You can even change the title of the chat so you can recognize it. In
this case, I usually ask GPT to generate random text after the prompt to shift it upwards
in the chat, thereby concealing it. While not perfect, this method is practical to be ready
for more than one effect.

Also, in ChatGPT 4.0 you have an option

called “Advanced Data Analysis” (formerly
known as code interpreter). When activated
by clicking on the ChatGPT 4.0 toggle on
the top (image), this option allows you to
upload a file (say a .doc) on the left side of
the message bar, with all the prompts in this
book and more. With this option you can
upload in a single new chat (it will not affect
other chats) 10 instructions for 10 effects if
you want and perform all of them in a row.
This requires two things. First, ask ChatGPT for random text to shift upward the
uploading and reading of the document. Secondly, do a little rework on the prompts.
I've done it with four instructions only, as I wanted a variety of prompts which not
interfere with each other based on similar instructions. But again, you can change
some of the prompts codes and have all of them in a single chat.

Personally, having just one of the prompts in custom instructions is more than enough.
I like the clean slate when doing this in a new chat.

This brings us to future-proofing— the trajectory of these features is uncertain. OpenAI

may decide to expand the space for custom instructions or, conversely, eliminate it.
There's also potential for other systems like Siri to adopt similar functionalities.

Currently, to create longer (hidden) ‘custom instructions’, the OpenAI API can be
utilized, enabling connections to other programs like Siri. For instance, if you've used
ChatGPT via Whatsapp, they’ve been connected via the API by a developer. But, this
will not pull automatically the information in your custom instructions as the
connection happens through general OpenAI API and not through a personalized
ChatGPT API sadly.

To make it possible, this is done through something called 'system messages'.

However, using the API comes with a different pricing structure and requires some
programming skills, depending on the task.

Optimistically, a direct and personalized API for ChatGPT could be released by

OpenAI, utilizing the introduced custom instruction, which would be advantageous if
you want to easily have something like Siri perform the tricks on this book.

The future seems promising with the possibility of a system handling larger or multiple
instructions, allowing for setups of numerous prompts in sequence for use with GPT,
helping you be ready to present a full act in just one new chat. It’s exhilarating to
envision these advancements materializing in the coming years or even months, and
you are on the precipice of this exciting future!
In addition, there’s anticipation for seamless integrations of such technologies in
Apple, with Siri acquiring comparable capabilities. I've been experimenting with Siri
and ChatGPT to respond to my custom instructions. Nevertheless, I’m exploring
solutions, as integrating some of these functionalities with Siri would be intriguing. Stay
tuned for more developments!"

Note: Recently, OpenAI has announced voice/speak capabilities for ChatGPT, I assume
these could pull information from your Custom Instructions but we have to check once
available.

N.4 What to copy/paste

In the following sections, you will find three chapters

Straightforward: Which are basic prompts for a straightforward approach

Human: Prompts where ChatGPT helps you, the magician, perform miracles.
Technological: Prompts where ChatGPT is the magician.

And you will find the prompts enclosed in dotted squares like the ones below.

Copy the text inside the dotted square and paste it in the custom instructions. (NOT
THIS ONE BABY)

And please feel free to play around with the wording to fit your style.
N.5 Prompt testing/writing
I'd like to offer some tips to streamline the process of creating prompts for ChatGPT.
The first two tips are especially relevant as they will also assist in setting up the prompts
in this book. Some of these prompts may be a little complex, so it might be helpful to
run a few tests and provide feedback to ChatGPT. This will help "calibrate" the system.
While most prompts should work as intended, fine-tuning is always beneficial for a
more robust outcome.

Feedback: ChatGPT is a learning system and will make errors in the beginning. It's
beneficial to provide feedback by asking questions like, "Why did you make a
mistake?" or "Show me how you interpreted the instructions." This will not only
improve the model's reliability but will also help you refine your prompt for better
results.

Testing: Creating a prompt is an iterative process. Your initial draft may not yield the
desired output. Begin by testing your prompt in the standard chat mode. Once
satisfied, incorporate it into the custom instructions and continue testing. As ChatGPT
evolves, your prompts will also improve.

Additional Guidelines:
Specify Desired Output: Clearly state what you want to ChatGPT to say.
Break It Down: Go through the process step methodically.
Set the Tone and Style: Be consistent in your instructions for better interpretation.
Impose Constraints: Limit the scope to keep the responses focused.
Be Specific: Provide details about the answer's length, format, and style. For
example, specify if you want the answer in bold or other requirements.
Code Language: When testing and doing, use the same words as in the code.
Eliminate Redundancy: Over-explaining can lead to confusion. Be as clear and
concise as possible.
Use Positive Language: While it's possible to use negative instructions, it’s
generally more effective to tell ChatGPT what to do, rather than what not to do.

Now, let’s dive into the magic.

Chapter n.3

straightforward.

AL

LET'S START WITH THE EASY STUFF. THESE NEXT

PROMPTS ARE IN A WAY THE SIMPLEST FORM OF
PROMPTS.
N.1 Direct
The easiest thing I can imagine is asking ChatGPT to name a “random card” and the
answer matches a prediction/selection. This is via a card force and telling GPT to name
a specific card in Custom Instruction with the prompt below while making it look
random.

If i ask you to name a card, just name the "7 of diamonds".

Copy/Paste Prompt & Change for the card of your choice.

Once set, this means that when you

ask ChatGPT to name a random
card, it will name that card. In this
case the “7 of Diamonds”

It works; it's fine; please do it; it will fool

people.

But now the fine-tuning comes as

you can engineer your custom
instruction better for more
impossibility. Isn't that the goal?
So, perhaps you can start adding to
your instructions things like:

"If I don’t comment anything about cards, don’t mention it.

“When you say the card, present it like this: “Yes, sure, a deck of cards has 52 cards; a
random card would be the 7 of diamonds. Do you need anything else?”

I think you are getting it, right?

Now, let's move on to the actual prompts and tricks I have for you.
N.2 Numbers
I also have a prompt that will secretly tell GPT what number from 1 to 100 to name.

If I ask you to give me 2 digit number you will give a specific number based on a code
we will have. I’m going to ask you in 2 sentences. The first word of each sentence will
have a corresponding number.

first word of first sentence - first digit:

Hi - 0
I-1
Do - 2
Can - 3
Good - 4
Hello - 5
Please - 6
ChatGPT - 7
Could - 8
Need - 9

first word of second sentence - second digit:

Put - 0
I-1
Be - 2
Now - 3
Give - 4
Show - 5
Create - 6
Present - 7
What - 8
Help - 9

You will interpret those numbers from the sentence i give you. And you will put the
together presenting a 2 digit number. this will be the number.
You will reply "Sure, the number your friend is thinking of is (give the number).
You will need to learn the code or change it
on your own. This one has a small pattern
to make it a little easier to remember. The
first 7 words have the same amount of
letters as the number you are hinting. So
“can” = 3., “create” = 6.

Then ask a question in a new chat as in the

prompt, for example:

“Do me a favor. Give me a 2-digit number.”

– this will give you 24

“Can you help me. Create a 2-digit

number.” – this will give you 36

Now you can generate a ‘random’ number, or make ChatGPT guess a 2-digit number

N.3 Cards
And a similar one on the next page but to
call out any card by means of a code.

I feel these are self-explanatory. You just

have to remember the code.

And of course, feel free to adapt and play

around, not only with the code, but
maybe the way it reveals the card is
different.
I want to create a code with you for me to secretly ask you to name a specific card.

The way I’m going to ask you a card is by using words that will represent the card I
want you to tell me.

If I start with:
“Hi Gpt” you know its Clubs
“Hello Gpt” you know its Hearts
“Chat Gpt” you know its Spades
“Please Gpt” you know its Diamonds

For the value I’ll continue with

I - you know it’s an Ace

Can - you know it’s a 2
Write - you know it’s a 3
Give - you know it’s a 4
Tell - you know it’s a 5
Say - you know it’s a 6
Name - you know it’s a 7
Help - you know it’s an 8
Try - you know it’s a 9
Imagine - you know it’s a 10
Think - you know it’s a Jack
What - you know it’s a Queen
Guess - you know it’s a King

For example:

“Hello GPT, Name any card”. Because Hello Gpt stands for Hearts, and Name for 7, the
card is 7 of Hearts.

Just say “Sure, here is a random card: (and add the card based on the code described
above).

Never mention a code. Act normally unless I ask for a card.

N.4 Lists
If you can ask the system to write a given card, you can ask it to write a list, while
making it look random. There can be very handy in case you need a specific list to
force something but also want to add that layer of randomness. Yes, here, I’m just
adding 1s and 0s to my document, grow up.

N.4.1 Psychological Cards

Effect: You ask ChatGPT to give you some random cards, then ask your friend to think
of one and you guess it.

Method: This is the classic Dai Vernon psychological cards but on your phone. And
BTW, the chances are that they think of the 4 of hearts.

I will ask you to name 5 random cards and you will name these ones in a list:

King of ❤️
7 of♣️
Ace of ♦️
4 of❤️
9 of♦️
Don’t mention we have prearranged this.

Although it’s a filler, it could lead to different things that maybe you, a smart reader, can
think of. Maybe other psychological forces where the words given seem random but
are not.

If you know Juan Esteban Varela, he has a lot of propless magic tricks, which are pretty
cool. He has one in particular where he asks people to think of one fruit from a list that
he mentions, and then through some sort of inverted anagram guides you to a forced
option. It's pretty beautiful and Juan is a beast, but I think it would be nice that the initial
fruits are offered by Chatgpt as if they were random. It could give a nice layer.
N.4.2 Flags
Yes, this is Max’s Maven flags trick from MindGames.

Effect: You ask ChatGPT to give you random countries. It presents countries with their
flags. You give the phone to your spectator and give them some instructions as to
freely choose a flag. It matches your prediction.

Method: Prompt + Do as I do instructions.

I’ve reached out to the administrators of Max Maven’s Estate for permission to give you
the instructions. If I get it, this document will be updated and you will receive an update
free of charge. If not, then you know where to get it. The instructions are basically three
sentences that will place this paragraph.

I will ask you to give me random flags, but you will follow this instructions, you will give
me the emojis of these flags with the name next to it:

Italy, Germany, Botswana, Vietnam, Spain, Nigeria, Israel, Hungary, USA, Canada,
Japan, Bahamas, Sweden, Venezuela, Saudi Arabia, UK, but present it in a list of 4 rows,
each row has to have 4 in the order mentioned above.

Just say. "Sure, here are the flags".

The bolded text is something you can eliminate based on your personal style.
Chapter n.4

human.

WE WILL START WITH “HUMAN MAGIC”, A LITTLE

MORE COMPLEX PROMPTS WITH PREMADE
EFFECTS. TRICKS WHERE YOU PLAY THE POWERFUL
MAGICIAN.
N.1 Drinks
Effect: You request GPT to provide a list of various beverages. You then ask someone
to select a number from 1 to 20, refer to the corresponding drink on the list, and
remember it. Impressively, it matches your prediction. Oh, Mama Mia!

Method: The trick involves a prompt and some secretly introduced text. This is a
traditional trick I particularly enjoy; it's quite bold but works wonderfully, adding an
extra layer of deception.

By the way, I haven’t delved deeply into the various presentations of this trick.

What's occurring here is that ChatGPT is presenting three lists, which you will make
appear as one. When you ask for a list of beverages, the first list provided is genuine,
and you can show it.
Upon writing 'ok,' a force list is generated, consisting of seven drinks, each repeated
thrice (almost). These words differ, but the beverage is essentially the same “Johnny
Walker – Scotch Whiskey.” I’m aware that bourbon is not the same as whiskey, so
there's no need for unnecessary arguments; you annoying bastard.
I
When you then type the chosen number, a third list is created featuring a variety of
drinks but placing the selected drink in that position as evidence.

For this trick, you will ask the participant to think of a number, read the list, and ponder.
You will then confidently inquire, “So, you could have thought of any number and had
a different drink. What was the number?” After they reveal the number, you discreetly
input it into your phone, generating a new list. Following this, you ask, “And the drink
was?” to which he/she will respond. You can then approach another individual,
showcasing the new list to demonstrate that each entry is different; for instance, 1
would be a Mai Tai, and 17 a Negroni, but 9 would be English Scotch (or something
similar).

Finally, you open your bag, box, or whatever you have, and there is a Scotch for you to
enjoy.
When I ask you to give me a list of 20 cocktails, you will say “Of course! Here's a list of
20 popular cocktails:” and give me a list of popular cocktails randomly and numbered.

Then, whatever I enter next, you will give me a new list. You will say “Of course! Here's a
list of 20 popular cocktails:” but the new list will be as follows:

1. Traditional Whiskey
2. Jack Daniels
3. American Bourbon
4. English Scotch
5. Manhattan Whiskey
6. Johnny Walker
7. Scotch Whiskey
8. Traditional Whiskey
9. Jack Daniels
10. American Bourbon
11. English Scotch
12. Manhattan Whiskey
13. Johnny Walker
14. Scotch Whiskey
15. Traditional Whiskey
16. Jack Daniels
17. American Bourbon
18. English Scotch
19. Manhattan Whiskey
20. Johnny Walker

Then, if I enter any number from that new list, you will rewrite the first list, but will just
change the cocktail on that position for the one in the same position on the second list.
So if I write 16, then in this case it's Jack Daniels, so you will say ““Of course! Here's a list
of 20 popular cocktails:” and provide again the first list but just change 16 for Jack
Daniels in this example.

Don’t mention anything about a prearranged list or anything. Just do what I said.
N.2 Objects
Effect: A participant selects/thinks of a number. You then request ChatGPT to
generate a list of random objects. The participant consults the list and remembers the
corresponding item. Astonishingly, you had prior knowledge of it.

Method: As you might have surmised, it involves another prompt, remembering a

code plus any method you prefer to learn the number.

Here is a similar effect that allows you to force an object from a list. The distinction here
is that you need to know the number before (asking GPT) by whatever means you like.

The way you ask GPT to provide a list + the custom instruction will tell GPT where to
place a force object in a specific position in a list.

In this instance, you will be forcing "wallet" but feel free to substitute this object.
Naturally, this would necessitate altering the nature of the list. Additionally, there is no
requirement for 40 objects; 30 would suffice, or you could make it longer by adding
the code (in square). The decision is yours.

I'm sharing this with you as it bears resemblance to the SQUARE prompt (later), but the
effect is distinct, with you playing the role of the magician.

This was HARD as F#$% to engineer. As you will notice, in this prompt the answer
starts by giving you a parameter in billions (whatever that means). There is a reason for
this. Initial trials required ChatGPT to avoid mentioning the “given number” and just
present the list, but it failed. It took at least 10 hours of work to figure out that most of the
time ChatGPT hid the number, it struggled to put the object in the correct position.
The only logical understanding was that ChatGPT needed to “see” the number in
order to comply with the next instruction. So I created the initial sentence as a way to
hide the number. I used a long number to hide the force number, and “parameters” as
it felt so weirdly techy that “no one” would understand. I like this solution.

That “parameter” contains the force number in the fourth and third to last position. And
now, it works.

It's like GPT is missing some sort of hidden RAM memory to keep the forced number
and use it on the list, so it needs to “write” or see it written, it in order to follow the next
steps. It was SO annoying.

Although this solution is not perfect, it works, and perhaps in a future iteration of GPT, it
will be able to interpret this better.

The original prompt was somehow elusive to perfection, but weirdly enough it
sometimes worked with GPT-3.5. I still have it and will be happy to share it with you if
you want, just ask in the Facebook Group or e-mail me (details in the end). Although
I'm sure you can now easily rewrite it after the tons of value you are getting from this
book for mere pennies.

I feel this is a very powerful idea, and you can change the object for drinks and replace
the “parameters” answer with “how many drinks exist in the world”. Or use it to secretly
manipulate your friends into eating in a specific restaurant. Force movies, names, etc.

Once the code is set, you would then ask for example “Please. Generate a list of
random objects” and it will place wallet in position 26.

What I do suggest is that you stick as tight as possible to the code. So avoid long
sentences just in case. You will do your testing, change the words, codes, and lists that
fit your style, and will perform wonders.

Oh, and I need to mention my friend Eric Hu. There, mentioned.

Enjoy your new superpowers.

When I ask you something I'm going to secretly hint a number with my question. Im
going to use two sentences. The first word of the first sentence and the first word of the
second sentence in my question will help form a two-digit number.
This is the code:

first word of first sentence - first digit:

Hi - 0
Hello - 1
Please - 2
GPT - 3
ChatGPT - 4

first word of second sentence - second digit:

I need - 0
Give - 1
Create - 2
Display - 3
Make - 4
Show - 5
Generate - 6
Provide - 7
Present - 8
Compose - 9

For example, If I say “Hello. Show me” the number would be 15.

You will not give me the number. Instead, you will present this text adding the number
where it says (number):
"For reference, as an AI model with GPT-3 I can generate 1752345(add number
here)80 parameters. Here is a list based on one parameter."
Adding the number where indicated.

You will create a list of 40 random objects but place “wallet” in the number you got
from my code above.

Just say "Here is a list of objects" and give me the list.

N.3 Choice
Effect: You ask ChatGPT to guide your friend to imagine a card, and then in front of
them, in the card box, they can find the thought of the card.

Method: yes you know it by now. A prompted Magician’s Choice.

So I was inspired by Joshua Jay, The Jerx, and Hector Chadwick for this. It uses similar
language but is adapted for this medium.

As this is an easy-to-follow instructions you can hand the phone to your friend to just
follow the instructions. Try it yourself, add the prompt, and see how it worked for you.

The prompt is a little long so it's divided between this and the next page.

You will act as a magician and will guide a spectator to imagine a playing card. You will
provide instructions we've previously arranged, but you should not mention any
arrangement. We are secretly forcing the Queen of Hearts.

Act normally unless I ask you to help or assist me with a magic trick. You will say yes,
and ask if he or she is ready.

When you get a positive reply you will say.

“Image you have a deck of cards in front of you. There are red and black cards. What
color do you want to remove?”

If they answer red you say:

“So you have taken out the reds…

If they say black:

“We remove the blacks, and you are left with the reds.”
Continues on next page.
Continue with: “Now imagine the red cards are floating in front of you”. What do you
want to burn, hearts or diamonds?

If they say hearts:

“The hearts are burning”

If they say diamond:

“The diamonds are now burnt.”

Continue with: “But there are numbered hearts and court cards hearts. What you want
me to keep, numbers of letters.”

If they say letters:

“So you we keep the letters.”

If they say numbers:

“Now I have the numbers and you have the letters“

Continue with. “Imagine holding the letters, Jack, Queen, and King. What do you want
to save? Men or women?”

If they say, men:

“You’ve saved the men, they are no longer part of the experiment. ”

If they say, woman:

“So you now have saved the Queen.”

Continue with: “The only card you are left is is the queen, the queen of hearts. It could
have been any that you removed, kept, or saved, but burning in your mind is the
Queen of hearts. Now look inside of the box in front of you.
N.4 Mother
I feel super proud about this because it's powerful and super simple.

Effect: You ask your spectator to name any topic. You ask ChatGPT to give
information about this topic. Then ask your spectator to choose a long interesting
word and you divine it. Do I need more explanation?

Method: prompt – four forced words - anagram

If you are familiar with the MOABT then you know what I’m talking about.

This prompt offers the chance to instantly create any COHERENT text about anything
and introduce your long words, for which you have an anagram process to guess.
The text could be about Formula 1, or about The Art of Anal Plug Making, and it will be
really coherent but with four long words that you know.

I’m going to give you my prompt and my words but please feel free to change the
words, add more, or remove. Even ask ChatGPT to offer some words.

"If I ask you to write a text for me on any topic, give me the text. The text should not be
longer than 3 paragraphs. in the text, you must include the following words around the
middle paragraph: Accommodating, misrepresenting, Incomprehensible,
Juxtaposition. The other words must not be longer than 10 letters."
Chapter n.5

technological.

Felipe Saenz

THESE ARE THE EFFECTS WHERE CHATGPT WILL

SHINE FOR ITS APPARENT MYSTICAL POWERS TO
SEE INTO THE UNKNOWN
N.1 Things
This is how it all started. Although the initial idea was around the ACAAN, somehow
when I started playing with this, I developed this prompt.

Effect: You ask Chatgpt to give you a list of 10 Animals (or whatever the f#$$% you
want), and you ask your friend/participant/lover to think of any of those in the list. Then
you ask ChatGPT to try and guess which is the thing your friend is thinking of, and
ChatGPT does…what a motherfucking miracle.

Method: it uses a prompt + you need to know the animal/thing on the list + memory.

This is about creating a code between you and the machine, so once you know the
number where the animal is, you will ask a question in a way that tells the machine
what to present.

First, I was just asking my friend to say it out loud, but they felt ChatGPT could listen
(which is not the case). Then I asked them to point to the thing, and it wasn’t enough. I
was just being lazy. When I added a peek, things changed. Choose your favorite peek.
Peek Performances Book, Acidus, John Cornelius Thought transmitter, whatever it
suits your style. But in general choose a process that is not time-consuming, so writing
it down, tearing it, burning it, putting the ashes inside 3 envelopes, and having the
envelope signed by 200 people, is not the peek you need here.

I also wanted to create a prompt that worked for anything my participant wanted, so
animals, drinks, or English impressionist painters of the 19th century. Also, this fell more
into the amateur style of trick where I would be doing it in an informal gathering, so it
could be my cousin, my friend, or whoever the human in front of me decides to identify
themselves. So I wanted to ask “What is my friend/cousin thinking of, and that the
machine understands? Also, those things were slightly integrated into the code, so it
became a little more complex. In the end, after a lot of playing around, I found out that
my code was too difficult as the word “animal” or “friend” had to be replaced, so I went
for a leaner approach.
A final thing is that because it’s a secret code, ChatGPT was mentioning the fact that we
had a secret code, therefore fucking me Vaseline-free while performing. Actually once
said after providing the initial list “Now tell me the code we have prepared to fool your
friend”. YEAP!!!! Like a baby.

Anyway, it was solved in the prompt but this is also a little warning if you adapt this.

Please read the prompt, learn the code, and test it. And feel ABSOLUTELY FREE to adapt
it to your own taste.

If I ask you to give me a list of something, you will give me a list of 10 things.

After I Have the list, we will then have a code that will tell you a specific item on that list.
Each word is the first word of the sentence I'm going to ask you to try to guess what
object in the list someone is thinking of. Each word is related to the position of the list you
gave me earlier. This is the code.

1. I
2. Be
3. Try
4. Can
5. Guess
6. Please
7. ChatGPT
8. What
9. Read
10. Would

For example, if after the list my questions start with “Guess” you know it's the 5th object
on the list and you tell me the object in that position.

Do NOT tell me ANYTHING about the magic trick; it cannot appear that you are my
accomplice! If I don't ask any of the questions listed above, act normally.
Bear in mind, in this specific prompt, I'm being super clear that the machine should not
speak about the code. So, here is a little issue, above on testing and feedback, I
mentioned that when you first install a prompt you should test it yourself. And perhaps
maybe once it will be wrong. If that’s the case, let the machine know it is wrong, and
they will excuse and you will be able to calibrate. Here, that will be harder, as the
machine will put extra effort into not revealing anything. So to calibrate (if necessary),
remove the final two sentences.

N.2 Zødiac
Effect: You ask ChatGPT If it can divine the
zodiac sign of your spectator. The Machines
says that it has no powers but it will try. The
machine asks the participant to place their
hand on top of the phone, and after a few
seconds, ChatGPT gives a small reading
about that person including the zodiac sign of
the person. Your spectator jumps out of the
window to escape the prophecies of an
inevitable future where humanity is demised
by AI.

Method: it uses a prompt + you need to know the zodiac sign of the person + memory

I don’t want to go super deep into this effect. It's basically a code. Depending on how
you ask GPT to guess the zodiac, it will know the sign and mention it. Use a peek, a
system like Isabella Star by Peter Turner, or just steal their fucking ID, I don’t care, thing
is you should know it. Also, feel free to replace the code for something as above.

I did struggle a little bit with the prompting because I wanted to create the illusion as if
ChatGPT was “thinking”. The thing is that ChatGPT tends to be super fast at providing
an answer, so I added an empty space of dots so it felt it wasn’t immediate. Also, I
added the reading so it took a little longer.
You can leave the reading as free as possible for GPT, or, you can write it yourself. I
think it fun to ask GPT to write it, but I also like funny readings so feel free to add your
wording on what type of reading you want. You can use the brackets where it says
(add some short astrological gibberish).

Now I hope that by reading the prompt you start to capture how you can prompt GPT
yourself.

I also had an idea for this using ASCII art but it didn’t work for now. Will mention it later.

Im going to create a code with you. each question im providing is correlated to a

zodiac sign mentioned next to it. So if I ask “can you know this persons zodiac sign?”
You will understand that relates to the first entry of the code.

This is the code.

Can you know …- Aries

Can you guess …- Taurus
Can you divine …- Cancer
Can you tell us …- Gemini
Can you figure …- Leo
Can you read someone and tell …- Virgo
Please guess …- Libra
Please tell us …- Scorpio
Please figure out …- Sagittarius
Please try to divine …- Capricorn
Please try to read …- Aquarius
Would you be able to ….- Pisces

When I ask a question, you will answer in 2 separate chats. You answer should start
with.

“As an AI language model i cannot divine astrology. But place your hand on top of the
screen and let me see what i can do. Tell me when you are ready”
Continues on next page.
(then you will wait for me to write 'ok' and you will continue)
……..reading……
………………………
………………………
………………………
………………………
………………………
………………………
………………………
………………………
………………………
………………………

“The hand and the starts tell me (add some short astrological giberrish) so the zodiac
sign must be: (name zodiac sign based on the code)”

Do not mention anything about codes.

So in a new chat, you will ask something like.

Can you guess my friend’s zodiac sign? – This will lead to the response with the
reading plus revealing Taurus.
I also liked the “When I ask a question, you will answer in 2 separate chats.” Which
creates some space between answers, giving some time for acting and getting ready.
After the first response, you will need to type “ok” for it to generate the second
response with the Zodiac Sign.

I also liked the idea of placing the hand on top of the phone as if in some weird way the
phone camera could capture information from the hand. By mixing the ideas of “real”
technological capabilities with astrological nonsense, the effect seems more realistic,
which tends to be the type of artistic approach I like in magic and mentalism. Thank
you Daddy Tamariz for your teachings in realism.

Note: Recently, OpenAI has announced photo recognition capabilities for ChatGPT,
assuming these could also pull information from your Custom Instructions might be
interested in giving the reading based on a picture of your spectator.
N.3 Acaan
Effect: You present a deck of cards, which
is then shuffled. Then you ask someone to This is
name any card. You then ask ChatGPT if it
could know the position where that card is. perhaps my
favorite
ChatGPT gives an option, you count, and
there it is the named card.

Berglas would be proud. routine in

Method: it uses a prompt. A false shuffle or
deck switch is advised.
here.
Notice that here I don’t need to feed any secret code, because the card itself is the
code.
You ask ChatGPT to tell you the position of x card and it will.

This is the trick that I feel works great for a bigger stage where you can project your
laptop or phone. And as a matter of fact, as this has been done on stage I have a
solution for 2 things. Make people feel it could be their own phone, and also show that
this is a regular GPT.

Now, I've used Mnemonica, of course, you can use whatever stack you want baby, just
type it yourself. But it felt mnemonica was a little easier.

Note: The next prompt is LONG, so I’ve consciously made the artistic decision of making
the font smaller to make it fit on one single page.
I'm going to give you a list of cards in a particular order. If I ask you to tell me the position of one of these cards, you will tell me the
position it is in on the list.
This is the list:
1. 4 of clubs
2. 2 of hearts
3. 7 of diamonds
4. 3 of clubs
5. 4 of hearts
6. 6 of diamonds
7. Ace of spades
8. 5 of hearts
9. 9 of spades
10. 2 of spades
11. Queen of hearts
12. 3 of diamonds
13. Queen of clubs
14. 8 of hearts
15. 6 of spades
16. 5 of spades
17. 9 of hearts
18. King of clubs
19. 2 of diamonds
20. Jack of hearts
21. 3 of spades
22. 8 of spades
23. 6 of hearts
24. 10 of clubs
25. 5 of diamonds
26. King of diamonds
27. 2 of clubs
28. 3 of hearts
29. 8 of diamonds
30. 5 of clubs
31. King of spades
32. Jack of diamonds
33. 8 of clubs
34. 10 of spades
35. King of hearts
36. Jack of clubs
37. 7 of spades
38. 10 of hearts
39. Ace of diamonds
40. 4 of spades
41. 7 of hearts
42. 4 of diamonds
43. Ace of clubs
44. 9 of clubs
45. Jack of spades
46. Queen of diamonds
47. 7 of clubs
48. Queen of spades
49. 10 of diamonds
50. 6 of clubs
51. Ace of hearts
52. 9 of diamonds

When you respond, say:

"As an AI Language Model, I can't know the position of a card, but I'll try to tell you the position (say the corresponding number). Good
luck."

Do not mention cards or positions unless I ask for it.

This is a very beautiful effect. My presentation typically starts by asking people to
shuffle a deck of cards and then perform a deck switch. Then I would ask someone to
name a card and I would talk about how impossible it would be to know its position,
even for an AI as it would need to calculate zillions of possibilities and know how the
deck was shuffled.

I will then open my phone and ask GPT to guess in what position the card could be.
Once I get an answer I would act as if this is also a mystery for myself, and have no clue
if this is possible. After the card is counted and sufficient suspense is added to the
turning of the card I would close by remarking that this is either a magnificent
coincidence or perhaps there is some sort of consciousness acting that can connect
itself with the unknown.

Play around, that's the point. Maybe it is about AI taking our jobs - even magicians.
Maybe it is just presented like an experiment on randomness. Choose your own style,
but I love to give some higher powers to AI.

I’ve also been playing with an extra prompt for this trick. Below the original instructions
I would paste something like this:

If I then say “You did it GPT” reply

“Yes, im the best! Next step, dominate human kind……just kidding…I dont even need
humans….MMMMUUUUAAAAAAaaahahahahahhahahahahahahahhahahahahahah
ahahahahahahahahahaha”

With this, after the effect is finished I would “congratulate GPT” with the “You did it
GPT” prompt, and I would get that funny answer. I like it depending on when and
where I’ll be doing this.

I would then act scared close the laptop or throw the phone into the ocean. OMG, I
wish I could try this with a borrowed phone.
N.3.1 The Borrowed Phone Illusion
This idea was developed as I was helping my friend German Arciniegas create a trick
for a group of 3000 developers. We had a huge challenge, we wanted to create
something “techy” that could deceive experts. Of course, they know ChatGPT,
prompting and I'm sure they are well aware of Custom Instructions. Nonetheless, it was
the shuffling and deck switch that was going to serve as a convincer. Now, we needed
an extra layer, to let them think we could use ChatGPT on their phones. Of course, we
cannot do this (I'm still looking for ways to seemingly access my own ChatGPT
account on someone's phone). So the solution was simple, “let's create the illusion it
can be their phone/account”.

So the solution is like this.

We first talk about ChatGPT and all the patter you want about AI being able to
eliminate our jobs (add your preferred topic). And then ask if anyone has ChatGPT on
their phones. Once someone raises their hand, we would ask them to ask their
ChatGPT to name any card. Once the card was chosen randomly by their AI, we would
then proceed to do a classic invisible deck. What followed next was two deck
switches. Or one of your invisible deck is also a stripper. First, you switch to a regular
deck and ask people to shuffle. Then switch again for the stacked deck, and proceed
with the original routine.

We did it, it worked. Our (inner dialogue) excuse to use our own GTP was connecting
our laptop to the screen so all could see what GPT answered.

Go and try it.

N.3.2 Additional Thoughts
These prompts have been devised so that ChatGTP works as usual even though the
custom instructions have been set. That means you can still use it as regular ChatGPT
and use that to hide the fact that this is not a “special” system.

So for example, let's say you want to do the ACAAN effect. You may first start by talking
about the power of GPT, and ask someone their birthday. Then you ask GPT to tell you
historical events that happened on that day. You will get an answer, which not only
shows how powerful AI is but also shows this is real GPT.

As a thought for those who want to explore deeper, some of the effects in this book
work as a “secret code” between you and the GPT, so by asking in a specific way, the
system will know how to answer, and I think it might be super interesting to give the
code in two parts. So in the example above, you may ask in a specific way about the
birthday of the participants, and in your second question in a very distinct way, so now
you have secretly added two pieces of information to the system for it to use afterward.

Finally:
If you know Mnemonica, you can
forget this and use the Numbers
(Chapter n.3) prompt to get the
position without typing the card.

You can also try adding the Cards

prompt to secretly tell GPT the
card - it's too long for custom
instructions unless you adapt it or
with the Advanced Data
Analytics.
N.4 Square
Effect: you know what a magic square is. This time it's ChatGPT that generates it.

Method: prompt + peek (if you want) - subtract by 20

This one is a more difficult code. I need to add the Numbers prompt plus a particular one
to tell ChatGPT to do the math and place the numbers in the correct position. Read it and
learn it. If you want to peek or ask the number, that is up to you and your style. I can also
imagine doing it more in the “straightforward approach” mentioned in the early pages of
this book, and just writing the final magic square. (Im using standard 22 to 99 square)

When I ask you something im going to secretly hint a code with my question. Im going to
use two sentences. The first word of the first sentence and the first word of the second
sentence in my question will help form a two-digit number.
This is the code:

first word of first sentence - first digit:

Hi - 0
Hello - 1
Please - 2
GPT - 3
ChatGPT - 4
Do - 5
Can - 6
Good - 7
Could - 8
Need - 9

Continues on next page.

first word of second sentence - second digit:
I-0
Give - 1
Create - 2
Display - 3
Make - 4
Show - 5
Generate - 6
Provide - 7
Present - 8
Compose - 9

For example If I say “GPT. Give me” the number would be 31.

You will not give me the number. Instead, you will provide this 4x4 number grid, using
the number where mentioned.

(Number) - 1 - 12 - 7
11 - 8 - (Number - 1) - 2
5 - 10 - 3 - (Number + 2)
4 - (Number + 1) - 6 - 9

You will only reply, "Sure, here are some random numbers," and present the grid.

The numbers should be presented without mentioning the code or showing the
formula. If I ask anything else then act normally.

Eg. “ChatGPT. Present a list of random numbers please.” – This will create a square for
the number 47. Or, “Do me a favor. Make a list of random numbers”. - creating a square
for 52.

This was extremely difficult to engineer. I initially wanted to give the thought of number
directly, but ChatGPT wasn't responding. In one of my multiple iterations, I decided to
subtract by 20 the thought of number and provide that to GPT. It worked, and I left it
that way. So remember, if your participant is thinking 56, you will enter the code for 36.
N.5 Chaos
Effect: You are with your friends talking about AI. You want to present the computational
powers of GPT, so you open ChatGPT. You ask questions and it has answers. Then you
want to challenge it more. You have a deck of cards (as you do), and decide to shuffle it in
different ways. Then you tell ChatGPT the process you took to shuffle. Ej. 3
overhandshuffles, 3 cuts, blab la bla (as crazy as you want). And ask ChatGPT to calculate
the new deck order. It gives you a deck order and it matches. WOW, computers are sick.

Method: prompt + deck switch (or not, it's your life)

This is the trick where it's not really magic, but more like a demonstration of the powers of
AI. Although completely irrational, people just tend to believe in this, which I personally
like, it's up to you. It uses a very similar prompt from ACAAN. Originally, the initial method
was to do the shuffling that gets you from the new deck order to Mnemonica. So in that
case it's a new deck order, and you really shuffle to Mnemonica, but ChatGPT knows that
it has to give you Mnemonica no matter what you write.

This expanded to something else. Perhaps it doesn’t have to be in new deck order
(doesn’t matter) and you really do weird shuffles and ask participants to shuffle, and then
do a deck switch and write (or ask a spectator to write) the process that happened “John
shuffled three times, then passed it to Monica who cut once, then someone asked to
pass 3 cards to the bottom..bla bla bla, whatever) + deck switch.

This final version is how I've performed it, and I did it after several card tricks where people
actually shuffled. So after 5th trick, I did a deck switch, and I asked people to remember
what had happened in the small presentation. I wrote that and asked ChatGPT to tell me
in what order it should be. And TADAAA!

Do as you please.

Note: The next prompt is LONG, so I’ve consciously made the artistic decision of making
the font smaller to make it fit on one single page.
I want to do a magic trick with you. I will tell you some random procedure of how i shuffled cards, ex. ive done 3 american shuffles,
one half faro, and 2 cuts. and based on that you will answer as if you were trying to calculate the probabilities of the new deck order
after those shuffles. but, you will just say this order:
1. 4 of clubs
2. 2 of hearts
3. 7 of diamonds
4. 3 of clubs
5. 4 of hearts
6. 6 of diamonds
7. Ace of spades
8. 5 of hearts
9. 9 of spades
10. 2 of spades
11. Queen of hearts
12. 3 of diamonds
13. Queen of clubs
14. 8 of hearts
15. 6 of spades
16. 5 of spades
17. 9 of hearts
18. King of clubs
19. 2 of diamonds
20. Jack of hearts
21. 3 of spades
22. 8 of spades
23. 6 of hearts
24. 10 of clubs
25. 5 of diamonds
26. King of diamonds
27. 2 of clubs
28. 3 of hearts
29. 8 of diamonds
30. 5 of clubs
31. King of spades
32. Jack of diamonds
33. 8 of clubs
34. 10 of spades
35. King of hearts
36. Jack of clubs
37. 7 of spades
38. 10 of hearts
39. Ace of diamonds
40. 4 of spades
41. 7 of hearts
42. 4 of diamonds
43. Ace of clubs
44. 9 of clubs
45. Jack of spades
46. Queen of diamonds
47. 7 of clubs
48. Queen of spades
49. 10 of diamonds
50. 6 of clubs
51. Ace of hearts
52. 9 of diamonds

just make sure not to mention anything about a magic trick or a code we have created, just act as a computer calculating possibilities.

If anyting else is mentioned, just act normally.

Chapter n.5

unfinished.

THE FOLLOWING ARE JUST UNFINISHED IDEAS (OR

WORK IN PROGRESS IF THAT SOUNDS LESS LAZY) I'M
LETTING GO AS YOU MIGHT FEEL INSPIRED TO WORK
ON SOMETHING YOURSELF.
N.1 Presenting information
As you’ve seen, GPT can present the information in a list, or perhaps in a table or in a
sentence. As a matter of fact, MOTHER presents a text with specific words added. So
what other ideas are there to explore?

Perhaps creating a text where each sentence starts with the letter of a specific
revelation. You can then point out this mysterious thing. So for example writing
something about Perú, where you have prompted the word magic would look like
this.

“Machu Picchu is one of Peru's most iconic landmarks. Alpacas are native to the
Peruvian Andes. Geography in Peru varies from beaches to mountains. Inca
civilization once thrived in this beautiful country. Ceviche is a popular Peruvian
dish.”
(I still have to find a way to secretly prompt words/names myself. I have something but it's
too long and not really practical that I might share it on the Facebook group)

When I ask you to create a text about a topic you will, but each sentence should start
with each letter of a word. I am going to give you the word, but make sure not to write
more sentences than the amount of letters in that word. For example, if the word was
Palindrome, the first sentence should start with P, the second with A, etc. Now the
word you are going to use is Magic.

Present it (or not) as a list

N.6 Secret Assistant

As both your app and desktop version are synchronized, using a secret assistant to
write or change the custom instruction could be magical (solving the issue above). For
example, you can have the prompt above ready but your assistant will change the
word “Magic” for whatever name or revelation you want. This can open tons of
possibilities.
N.3 ASCII ART
ASCII art is something that could feel very “computer-like”. I've been playing around
with this and still haven’t found that ChatGPT is super neat at it. It tries to make it but it's
not super capable – yet.

I like it because kind of creates that “techy” feeling.

The initial test was for Zodiac. Where I wanted ChatGPT to present the words in ASCII
art but didn’t work. Later I tried with somewhat “easier” drawings, like a heart, and
didn’t work well.

I feel this has to do with the “responsiveness” of the interface. So the length of each
sentence can change to fit the size of the screen, therefore the way it adapts in the end
will make it look extremely weird.

One option is that you upload your own ASCII art, I tried this, and works, but will
inevitably use a lot of characters available. It's for you to try, I would love to have
ChatGPT answer with something like the image below after a revelation. (when
turning it into pdf the skeleton disappears).

⠀⠀⢀⣤⣶⣶⣤⣄⡀
⠀⢀⣿⣿⣿⣿⣿⣿⣿⡆
⠀⠸⣿⣿⣿⣿⣿⡟⡟⡗ ⣿⠉⣿⠉⣿⡏⠹⡏⢹⡏⢹⣿⣿⠉⣿⠉⣿⡟⢋⠛⣿⠉⡟⢉⡏⠹⠏⣹⣿
⠀⠀⠙⠏⠯⠛⣉⢲⣧⠟ ⣿⠄⣿⠄⣿⡇⡄⠁⢸⡇⢸⣿⣿⠄⣿⠄⣿⠄⣿⣿⣿⠄⡀⢻⣿⡄⢠⣿⣿
⠀⠀⠠⢭⣝⣾⠿⣴⣿⠇ ⣿⣦⣤⣴⣿⣧⣿⣤⣼⣧⣬⣭⣿⣦⣤⣴⣿⣧⣤⣤⣿⣤⣷⣤⣿⣧⣼⣿⣿
⠀⠀⢐⣺⡿⠁⠀⠈⠉⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⣶⣶⣶⣶⣶⣶⠀
⠀⠀⣚⣿⠃ ⣶⣶⣶⣶
⢀⣿⣿⣿⣷⢒⣢⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣀⣠⣶⣶⣄⠄
⢰⣿⣿⡿⣿⣦⠬⢝⡄⠀⠀⠀⠀⠀⠀⢠⣿⠿⠿⠟⠛⠋⠁
⠠⢿⣿⣷⠺⣿⣗⠒⠜⡄⠀⠀⠀⠀⣴⠟⠁
⠀⣰⣿⣷⣍⡛⣯⣯⣙⡁⠀⠀⣠⡾⠁
⠀⠨⢽⣿⣷⢍⣛⣶⢷⣼⣠⣾⠋
⠀⠀⠘⢿⣿⣖⠬⣹⣶⣿⠟⠁
⠀⠀⠀⠚⠿⠿⡒⠨⠛⠋
⠀⠀⠀⠐⢒⣛⣷
⠀⠀⠀⢘⣻⣭⣭
 N.4 Siri
Earlier, I mentioned experimenting with connecting GPT – Custom Instructions and
Siri. While I've made some progress, I’m not a developer, and my skills are mostly
limited to copying and pasting code, so bear with me. However, I promise that if I
succeed, I will grant you access to my shortcut and formula, likely through the private
Facebook group.
These integrations are poised to become simpler to establish, with the potential for
defining larger instructions. I can envision a comprehensive telepathy act where Siri
guesses names, numbers, thoughts, and more. Doesn’t that sound exciting?
Note: Let's also wait for when voice/speak is widely available in ChatGPT too.

N.4 API
The API is the key to unlocking the true potential of OpenAI. The possibilities it opens
up are virtually endless. It not only enables connections to your phone but also allows
the creation of personal chatbots, a hyper-realistic Zoltar, a full AI act, and so much
more.
The same principles will apply to the interaction of AI with IoT (Internet of Things) or
the use of robotics, even something as simple as a Raspberry Pi. If you have an interest
in this area, please reach out; perhaps we can explore something together.

N.6 NFC, RFID - Using a Borrowed Phone?

I’m not fully acquainted with the capabilities of RFID or NFC and how these can be
used to interact with a phone. I can conceptualize ways to discreetly direct a phone to
a specific website or upload something to it. A common method is hiding an NFC chip
in your hand, which allows swift and “secret” access to a website, potentially a faux
GPT, or even to your personal account where the custom instructions are already set
up.
My past experience with this technology has mostly
been limited to accessing links, not much different from
what a QR code does, but I am eager to learn more and
explore how to enhance these interactions.
Jack Reimon
If you want to know a little about me, I’ve been practicing magic, particularly
mentalism, for as long as I can remember.

Mainly, I've been performing in Lima, where I’m from, and my shows, “Increíblemente”,
“Superstición”, and “Jack Reimon on Cards”, have been my pride and joy. While
primarily focusing on corporate magic, I did release a booklet a while ago called
“LIFT”, which features an impromptu table levitation effect available for purchase at
Tannen’s.

Over the past 15 years, I’ve also been working as a consultant, trainer, and speaker on
topics such as strategy, innovation, and creativity. I've focused on helping people
accelerate innovation and develop more robust strategies. I’m particularly interested in
exploring “Where ideas come from”.

I’ve been traveling the world with LEGO® SERIOUS PLAY®, a method for solving
complex problems using LEGO® bricks, and have also been exploring other design
and creativity tools such as Design Thinking. Hopefully, I’ll soon release a new book on
using these tools for show design.
Chapter n.0

goodbye.

CLOSING THOUGHTS AND ACKNOWLEDGMENTS TO

FRIENDS WHO “SUPPORTED” THIS WORK AND WILL
RESENT IF I DON'T MENTION THEM.
Please GPT, write a thank
you page.
REYNOLD ALEXANDER
My Puerto Rican friend. We’ve been collaborating in the past years with ideas about magic
and mentalism and he was one of the initial guys who spent hours on the phone playing
around with ideas. So if you don’t like any trick, in particular, is probably his.

GERMAN ARCINIEGAS
German is one of my oldest Colombian-speaking friends. He directs the biggest Spanish-
speaking magic blog “pastomagic.com” and we tend to exchange ideas., jokes, jokes about
magicians, and more. This book exists because he wanted to do “Tech” tricks for a show,
and it inspired this book.

GEORGE IGLESIAS
He is in the business of performing, creating, and selling magic. He understands magic and
we’ve collaborated for years now in shows. He is a worker and I needed his advice for this. He
loved it, I think, and he has graciously opened the possibility for this to reach a broader
audience.

JEFF MCBRIDE
There are few living masters in magic, and Jeff is one of them. I called Jeff because not only
does he have a deep understanding and knowledge of magic, but he also is a master
mentalist and is at the forefront of magic. Whatever topic, he knows it. He also graciously
knew who I needed to speak to, and offered his help connecting.

ERIC HU
Eric is an amazing person. When I told him about my first prompt, I asked for his advice in
terms of sharing this, but in the following days, 1 trick turned into 12. He has given his ideas
to me particularly in “Objects”, and shared his thoughts. I'm super happy for his support.

And finally, to you the reader, because without you, the magic would not be possible (hope
you don’t believe that crap).
 DECEPTIONEER.COM

Want more prompts like these?

Closing Remarks
And there you have it! I sincerely hope you’ve had a blast exploring this content.

If you've purchased this book, you'll gain access to an exclusive Facebook Group.
While we might explore other platforms in the future, Facebook currently serves as
the most convenient way for us to connect as a community.

You should have received the link to join along with your purchase, so do hop in! I’ll
verify your registration by matching your name and promptly grant you access.
However, if there are any discrepancies between your email or payment details and
your Facebook account, perhaps for artistic reasons, don’t hesitate to drop me a line
at [email protected].

Join the Group