WEBVTT

00:00:00.000 --> 00:00:04.860
OK, so imagine for a second, you go out and you

00:00:04.860 --> 00:00:07.400
buy a Ferrari. It's sitting in your driveway.

00:00:08.279 --> 00:00:12.199
Italian leather, the engine is purring. It's

00:00:12.199 --> 00:00:15.019
capable of these incredible speeds. Right. But

00:00:15.019 --> 00:00:17.559
every single time you get in it, you never shift

00:00:17.559 --> 00:00:20.160
out of first gear. That's a painful image. You're

00:00:20.160 --> 00:00:22.539
just putting around the neighborhood at 10 miles

00:00:22.539 --> 00:00:24.719
an hour. But I mean, according to the data we're

00:00:24.719 --> 00:00:27.780
looking at today, that is exactly what is happening

00:00:27.780 --> 00:00:29.519
with artificial intelligence. Is that right?

00:00:29.780 --> 00:00:33.939
Over 90 % of paying users, so people actually

00:00:33.939 --> 00:00:36.500
dropping $20 a month, have never even opened

00:00:36.500 --> 00:00:39.119
the settings menu. Wow. Or driving a supercar

00:00:39.119 --> 00:00:41.880
in first gear. I have to admit, when I saw that

00:00:41.880 --> 00:00:44.939
stat, I felt a little called out. I'm definitely

00:00:44.939 --> 00:00:46.920
guilty of just opening the window and typing,

00:00:47.219 --> 00:00:49.719
assuming the defaults are good enough. We all

00:00:49.719 --> 00:00:53.070
are, I think. Welcome to the deep dive. Today,

00:00:53.189 --> 00:00:55.570
we're analyzing a guide called Mastering the

00:00:55.570 --> 00:00:57.670
Interface. But this isn't just a tech review,

00:00:57.789 --> 00:00:59.890
is it? It's more a conversation about how we

00:00:59.890 --> 00:01:02.530
interact with intelligence itself. It is, yeah.

00:01:02.789 --> 00:01:05.109
And the mission for this deep dive is really

00:01:05.109 --> 00:01:09.090
moving from being a casual user to a power user.

00:01:09.269 --> 00:01:11.769
Right. And the most interesting part, I think,

00:01:12.329 --> 00:01:15.829
from the source material is that you don't need

00:01:15.829 --> 00:01:17.790
to learn to code. You just need to know where

00:01:17.790 --> 00:01:20.200
the buttons are. which is fortunate because my

00:01:20.200 --> 00:01:23.939
coding skills are nonexistent. So looking at

00:01:23.939 --> 00:01:25.420
our roadmap, we're going to move through this

00:01:25.420 --> 00:01:27.959
logically. We'll start with the foundation. So

00:01:27.959 --> 00:01:29.700
the privacy and memory settings that I think

00:01:29.700 --> 00:01:31.439
most people are just afraid to touch. Then we'll

00:01:31.439 --> 00:01:33.840
move to the art of the prompt. How to structure

00:01:33.840 --> 00:01:35.859
your language so the machine actually understands

00:01:35.859 --> 00:01:37.849
you. Right. And then we'll cover the productivity

00:01:37.849 --> 00:01:41.810
tools that turn it from a toy into a real workspace.

00:01:42.030 --> 00:01:44.189
And finish up with the advanced stuff. Vision,

00:01:44.469 --> 00:01:47.450
web browsing, and these new agent capabilities.

00:01:47.730 --> 00:01:50.129
It's about moving from a passive consumer to

00:01:50.129 --> 00:01:53.549
an architect of the tool. Exactly. So let's unpack

00:01:53.549 --> 00:01:55.430
this. We start with the foundation. The source

00:01:55.430 --> 00:01:57.989
material calls this fixing the foundation before

00:01:57.989 --> 00:02:00.450
you build the house. And it starts with something

00:02:00.450 --> 00:02:03.310
that I think generates a lot of anxiety. Privacy.

00:02:03.590 --> 00:02:06.280
Right. There's this lingering question. people

00:02:06.280 --> 00:02:08.919
have, like, is the AI watching me? Is it reading

00:02:08.919 --> 00:02:12.979
my diary? Yeah. And the reality, according to

00:02:12.979 --> 00:02:15.860
this guide, is that unless you change a very

00:02:15.860 --> 00:02:20.699
specific setting, OpenAI is using your chats

00:02:20.699 --> 00:02:23.879
to train its future models. So if I'm uploading

00:02:23.879 --> 00:02:26.900
a confidential business plan or, I don't know,

00:02:27.080 --> 00:02:29.180
venting about a relationship. It's potentially

00:02:29.180 --> 00:02:31.550
going right into the mix. Yeah. to make the model

00:02:31.550 --> 00:02:33.349
smarter for everyone else. See, I think the reason

00:02:33.349 --> 00:02:35.969
people don't touch this stuff is fear. They think

00:02:35.969 --> 00:02:38.169
if they mess with the settings, they'll, like,

00:02:38.270 --> 00:02:42.110
break the bot. Of course. But the guide is pretty

00:02:42.110 --> 00:02:44.349
specific here. It's actually really simple. What

00:02:44.349 --> 00:02:46.599
does it say? It's under data controls. You go

00:02:46.599 --> 00:02:48.460
to settings, then data controls, and you just

00:02:48.460 --> 00:02:51.300
turn off the toggle that says, improve the model

00:02:51.300 --> 00:02:53.520
for everyone. Just a switch. A simple switch.

00:02:53.639 --> 00:02:55.400
But there has to be a trade -off, right? If I

00:02:55.400 --> 00:02:57.199
turn that off, does the tool get worse for me?

00:02:57.360 --> 00:03:00.379
Not immediately. No. But you're essentially opting

00:03:00.379 --> 00:03:02.360
out of that collective learning process. You

00:03:02.360 --> 00:03:04.900
kind of become a ghost to this system. For most

00:03:04.900 --> 00:03:06.960
people, though, especially in a business context,

00:03:07.460 --> 00:03:09.680
that trade -off is absolutely worth it for the

00:03:09.680 --> 00:03:12.060
security. That makes sense. Secure the perimeter

00:03:12.060 --> 00:03:14.199
first. But then, once you've done that, you have

00:03:14.199 --> 00:03:17.360
to look at how the AI remembers you. This memory

00:03:17.360 --> 00:03:19.280
feature. I've had moments where it brings up

00:03:19.280 --> 00:03:21.620
something I said three weeks ago, and it feels

00:03:21.620 --> 00:03:23.740
a little, I don't know, a little black mirror.

00:03:23.819 --> 00:03:26.599
It can be jarring, yeah. But the guide points

00:03:26.599 --> 00:03:30.319
out this isn't magic. It's just a database. And

00:03:30.319 --> 00:03:32.740
you have control over it. What do you mean? You

00:03:32.740 --> 00:03:35.900
can actually go into the settings under personalization

00:03:35.900 --> 00:03:38.580
and view the specific list of things it knows

00:03:38.580 --> 00:03:41.469
about you. So I can see a list that says, like,

00:03:41.650 --> 00:03:44.669
user likes sci -fi novels or user is a vegetarian.

00:03:45.090 --> 00:03:47.110
Exactly. And if there's something wrong, let's

00:03:47.110 --> 00:03:49.169
say it thinks you're a lawyer because you asked

00:03:49.169 --> 00:03:51.990
one legal question three months ago. Right. You

00:03:51.990 --> 00:03:54.270
just hit the little trash can icon next to it

00:03:54.270 --> 00:03:56.310
and delete it. You can even wipe the whole memory

00:03:56.310 --> 00:03:58.789
if you want a fresh start. That's useful because,

00:03:58.830 --> 00:04:01.280
I mean, I still wrestle with... what I call prompt

00:04:01.280 --> 00:04:03.939
drift. I'll be working on a podcast script and

00:04:03.939 --> 00:04:05.939
suddenly it starts giving me advice like I'm

00:04:05.939 --> 00:04:07.900
a Python developer because of something I asked

00:04:07.900 --> 00:04:10.460
yesterday. Right. It feels like I'm constantly

00:04:10.460 --> 00:04:12.819
wrestling the steering wheel back. That's the

00:04:12.819 --> 00:04:14.919
context pollution problem. Yeah. And managing

00:04:14.919 --> 00:04:17.620
that memory list is exactly how you clear the

00:04:17.620 --> 00:04:20.759
air. Okay. So that leads us to the about me section

00:04:20.759 --> 00:04:23.500
or custom instructions. Now I've seen this in

00:04:23.500 --> 00:04:26.500
the menu, but I usually just ignore it. Why does

00:04:26.500 --> 00:04:28.620
the source material put so much weight on this?

00:04:29.600 --> 00:04:32.120
Because it solves the problem of repetition.

00:04:33.379 --> 00:04:36.060
Without this, every single time you start a new

00:04:36.060 --> 00:04:38.939
chat, you have to establish context from scratch.

00:04:39.120 --> 00:04:40.819
It's like having to introduce yourself to a friend

00:04:40.819 --> 00:04:43.040
every time you meet them. Hi, I'm Dave. I like

00:04:43.040 --> 00:04:46.040
podcasts. Exactly. Custom instructions lets you

00:04:46.040 --> 00:04:48.459
do that introduction one time, and it sticks.

00:04:48.730 --> 00:04:51.149
There are two boxes. Box one is, what should

00:04:51.149 --> 00:04:54.129
chat GPT know about you? Okay. And box two is,

00:04:54.290 --> 00:04:57.050
how would you like chat GPT to respond? The guide

00:04:57.050 --> 00:04:59.569
had a really clever example of this. Something

00:04:59.569 --> 00:05:02.449
about a junior English teacher in Vietnam. Yes.

00:05:02.949 --> 00:05:05.310
In their instructions, they specified that they

00:05:05.310 --> 00:05:07.670
teach 10 -year -olds. And crucially, in that

00:05:07.670 --> 00:05:10.629
second box, they ask for short sentences, no

00:05:10.629 --> 00:05:14.250
academic jargon, and please... Use bullet points.

00:05:14.649 --> 00:05:17.569
And once that's saved, every answer is automatically

00:05:17.569 --> 00:05:19.370
tailored for a 10 -year -old. You don't have

00:05:19.370 --> 00:05:21.329
to keep typing. Explain this simply. It just

00:05:21.329 --> 00:05:23.689
does it. It creates a consistent persona. It

00:05:23.689 --> 00:05:25.689
stops the drift. There's one more feature in

00:05:25.689 --> 00:05:29.410
this section, temporary chat, the so -called

00:05:29.410 --> 00:05:33.930
incognito mode. Right. Be honest. Does anyone

00:05:33.930 --> 00:05:37.029
actually use this? Or is it just for when you

00:05:37.029 --> 00:05:39.589
have, like, embarrassing medical questions you

00:05:39.589 --> 00:05:41.649
don't want in your history? Well, the medical

00:05:41.649 --> 00:05:43.949
questions are a totally valid use case. But it's

00:05:43.949 --> 00:05:46.350
also for those one -off queries. You don't want

00:05:46.350 --> 00:05:48.709
polluting your memory. Like what? If you're asking

00:05:48.709 --> 00:05:51.110
for a recipe for a party, you don't necessarily

00:05:51.110 --> 00:05:53.129
want the AI to remember you love gluten -free

00:05:53.129 --> 00:05:55.470
cupcakes forever. You toggle on temporary chat,

00:05:55.629 --> 00:05:58.110
the screen turns gray, and nothing is saved.

00:05:58.290 --> 00:06:01.149
It's the burn after reading feature. OK, so looking

00:06:01.149 --> 00:06:03.129
at these foundational settings, memory, custom

00:06:03.129 --> 00:06:05.230
instructions, it does raise a question for me.

00:06:05.439 --> 00:06:08.300
If we automate our context so heavily, do we

00:06:08.300 --> 00:06:10.800
lose the ability to be flexible in how we ask

00:06:10.800 --> 00:06:14.060
things? That is the tension. If you hard code

00:06:14.060 --> 00:06:16.199
your instructions too strictly, you might just

00:06:16.199 --> 00:06:18.600
create an echo chamber where the AI only talks

00:06:18.600 --> 00:06:21.740
to you in one specific way. So, rigidity versus

00:06:21.740 --> 00:06:23.959
convenience. Rigidity versus convenience, okay.

00:06:24.300 --> 00:06:26.600
So, we've built the foundation, the settings

00:06:26.600 --> 00:06:28.860
are locked in, now we have to actually talk to

00:06:28.860 --> 00:06:31.439
the thing. Right. Section two of our source material

00:06:31.439 --> 00:06:34.259
is the art of the prompt. And this is where most

00:06:34.259 --> 00:06:37.220
people struggle. They blame the tool when they

00:06:37.220 --> 00:06:39.839
get a generic answer. But usually the problem

00:06:39.839 --> 00:06:43.639
isn't the AI, it's the input. The source compares

00:06:43.639 --> 00:06:46.360
it to learning to ride a bike. It suggests there's

00:06:46.360 --> 00:06:49.019
a perfect prompt structure, four parts. Roll,

00:06:49.360 --> 00:06:51.439
context, task, and format. If you miss one of

00:06:51.439 --> 00:06:53.699
those, the bike wobbles. Okay, let's role play

00:06:53.699 --> 00:06:54.939
this because I want to see if it actually makes

00:06:54.939 --> 00:06:56.920
a difference. Let's say I own a coffee shop.

00:06:57.199 --> 00:06:59.980
The bad way would be me just typing what... Write

00:06:59.980 --> 00:07:02.240
a post about coffee. Exactly. And the AI will

00:07:02.240 --> 00:07:04.779
give you something so incredibly bland. Coffee

00:07:04.779 --> 00:07:07.420
is great. Come drink it. It's useless. OK, so

00:07:07.420 --> 00:07:09.579
fix it for me. How do I use the four parts? OK,

00:07:09.879 --> 00:07:14.480
so roll. Act as a social media expert. Got it.

00:07:14.860 --> 00:07:18.300
Context. I own a small coffee shop in Hanoi that

00:07:18.300 --> 00:07:20.819
specializes in coconut coffee. OK, specific.

00:07:21.180 --> 00:07:24.040
Task. Write a Facebook post for our summer launch.

00:07:24.740 --> 00:07:28.360
And then format. Give it an exciting tone, use

00:07:28.360 --> 00:07:31.500
three emojis, and keep it under 100 words. Okay,

00:07:31.620 --> 00:07:33.939
wow. I can see how that would produce a wildly

00:07:33.939 --> 00:07:36.620
different result. The AI knows exactly what success

00:07:36.620 --> 00:07:38.420
looks like. You're not making a guess anymore,

00:07:38.500 --> 00:07:40.680
you're giving it a map. And for listeners who

00:07:40.680 --> 00:07:43.660
really want to geek out on this, the source mentions

00:07:43.660 --> 00:07:47.240
Dr. Jules White's course on Coursera. Oh, yeah.

00:07:47.420 --> 00:07:49.680
He frames this perfectly. He says language is

00:07:49.680 --> 00:07:51.879
now a programming interface. You're coding with

00:07:51.879 --> 00:07:53.920
English? You're writing code just without all

00:07:53.920 --> 00:07:56.680
the syntax errors. So does the structure of the

00:07:56.680 --> 00:07:59.319
prompt then matter more than the creativity of

00:07:59.319 --> 00:08:01.839
the idea behind it? I would argue that in this

00:08:01.839 --> 00:08:04.920
medium, yeah. The format is what unlocks the

00:08:04.920 --> 00:08:07.160
function. You can have a brilliant idea, but

00:08:07.160 --> 00:08:09.519
if you can't structure the request, the intelligence

00:08:09.519 --> 00:08:11.579
can't manifest it for you. Format essentially

00:08:11.579 --> 00:08:14.100
unlocking the function. That's a good takeaway.

00:08:14.930 --> 00:08:17.769
Okay, let's move to section three, daily productivity

00:08:17.769 --> 00:08:20.370
tools. This is about turning the AI from a fun

00:08:20.370 --> 00:08:23.250
toy into a real workspace. And the first tool

00:08:23.250 --> 00:08:26.089
here is called projects. The best analogy for

00:08:26.089 --> 00:08:28.769
it is smart folders. This one resonates with

00:08:28.769 --> 00:08:31.269
me because I hate context switching. Oh, yeah.

00:08:31.370 --> 00:08:34.230
I'll be asking the AI to help me write a serious

00:08:34.230 --> 00:08:37.149
email to a lawyer. But five minutes earlier,

00:08:37.210 --> 00:08:39.850
I was asking it to generate names for, like,

00:08:40.070 --> 00:08:42.830
a D &D character, and the tone just bleeds over.

00:08:42.909 --> 00:08:45.269
It's so annoying. That is exactly what Project

00:08:45.269 --> 00:08:48.309
Solves. You create a folder, say, Legal Work.

00:08:48.730 --> 00:08:50.870
You can upload your previous contracts or your

00:08:50.870 --> 00:08:53.409
company's style guides into that project. Oh.

00:08:53.590 --> 00:08:55.730
Now, every chat inside that folder is walled

00:08:55.730 --> 00:08:59.259
off. It knows the context is law, not... Dungeons

00:08:59.259 --> 00:09:01.200
and Dragons. And speaking of finding things,

00:09:01.720 --> 00:09:04.860
the source mentions the new search feature, which

00:09:04.860 --> 00:09:06.919
seems obvious, but it wasn't there for a long

00:09:06.919 --> 00:09:09.820
time. Windly. The little magnifying glass icon,

00:09:09.820 --> 00:09:12.759
it turns your chat history from this messy endless

00:09:12.759 --> 00:09:15.379
scroll into a searchable library. You just type

00:09:15.379 --> 00:09:17.600
meeting or recipe and it pulls it right up. It

00:09:17.600 --> 00:09:19.620
turns your past interactions into a database

00:09:19.620 --> 00:09:22.299
you can actually use. Now, I want to talk about

00:09:22.299 --> 00:09:24.700
voice mode. The source describes this as talking

00:09:24.700 --> 00:09:27.250
like a human. I've used this while cooking, and

00:09:27.250 --> 00:09:29.070
honestly, the thing that surprised me most was

00:09:29.070 --> 00:09:31.309
the interruptions. It's startling, isn't it?

00:09:31.669 --> 00:09:33.649
You can just cut it off mid -sentence and it

00:09:33.649 --> 00:09:36.230
adjusts, it handles laughter, it understands

00:09:36.230 --> 00:09:38.190
cadence. It totally changes the feeling of it.

00:09:38.350 --> 00:09:41.250
It's not typing at a computer anymore. It feels

00:09:41.250 --> 00:09:44.230
like a presence in the room. It creates a new

00:09:44.230 --> 00:09:47.429
modality, brainstorming while driving, practicing

00:09:47.429 --> 00:09:49.809
a language while you're doing dishes. It frees

00:09:49.809 --> 00:09:52.100
the intelligence from the screen. Another productivity

00:09:52.100 --> 00:09:54.620
feature mentioned is file uploads. The scenario

00:09:54.620 --> 00:09:57.779
in the source is a classic one. Your boss sends

00:09:57.779 --> 00:10:01.320
you a 50 page PDF report. Right. You upload it

00:10:01.320 --> 00:10:03.799
and just say, summarize this. And seconds later,

00:10:03.820 --> 00:10:05.820
you have the top five points in your action items.

00:10:06.019 --> 00:10:10.220
But here's my skepticism. Do you trust it? I

00:10:10.220 --> 00:10:13.259
mean, if I upload a dense legal contract or a

00:10:13.259 --> 00:10:15.299
complex medical study, am I really going to trust

00:10:15.299 --> 00:10:17.580
a three second summary? That is the critical

00:10:17.580 --> 00:10:20.379
question. The source material frames this as

00:10:20.379 --> 00:10:23.320
a massive time saver. And it is. But there is

00:10:23.320 --> 00:10:27.019
a danger. We are trading date comprehension for

00:10:27.019 --> 00:10:29.120
efficiency. We're outsourcing the reading. Right.

00:10:29.120 --> 00:10:30.720
And if you outsource the reading, do you lose

00:10:30.720 --> 00:10:33.480
the nuance? I think it's amazing for triage just

00:10:33.480 --> 00:10:35.779
knowing if a document is worth your time to read

00:10:35.779 --> 00:10:38.139
deeply. OK. But I wouldn't let it sign the contract

00:10:38.139 --> 00:10:40.659
for me. That's a fair distinction. Triage versus

00:10:40.659 --> 00:10:43.159
decision making. But just, I mean, pause for

00:10:43.159 --> 00:10:45.179
a second and imagine the scaling implications.

00:10:45.699 --> 00:10:48.679
even if it's just for triage. You handed a 50

00:10:48.679 --> 00:10:51.659
page dense report and seconds later, you know

00:10:51.659 --> 00:10:54.440
it's in it. It just changes the velocity of information

00:10:54.440 --> 00:10:56.580
flow inside a company. It definitely speeds up

00:10:56.580 --> 00:10:59.399
the treadmill, but does it make us better runners?

00:10:59.679 --> 00:11:04.159
That's the question. Okay, we're back. We've

00:11:04.159 --> 00:11:05.940
covered the settings, the prompts, and the daily

00:11:05.940 --> 00:11:08.700
tools. Now let's unpack the really futuristic

00:11:08.700 --> 00:11:12.149
stuff. Section 4 covers advanced vision and what

00:11:12.149 --> 00:11:14.590
the source calls the doer era. Right. We are

00:11:14.590 --> 00:11:17.850
moving beyond just text. The source discusses

00:11:17.850 --> 00:11:20.370
multimodal prompts. This just means the AI can

00:11:20.370 --> 00:11:22.950
see. And the examples here are so practical.

00:11:23.190 --> 00:11:25.490
Taking a photo of your open fridge to get a recipe.

00:11:25.570 --> 00:11:27.850
Or taking a photo of a broken bicycle part because

00:11:27.850 --> 00:11:30.210
you have no idea what it's called. You just ask,

00:11:30.269 --> 00:11:33.350
how do I fix this? It bridges that gap between

00:11:33.350 --> 00:11:36.649
the physical world and digital information. Then

00:11:36.649 --> 00:11:39.700
there's canvas mode. This one feels less like

00:11:39.700 --> 00:11:43.580
a chat and more like, I don't know, like Google

00:11:43.580 --> 00:11:46.399
Docs. It's a dedicated workspace for writing

00:11:46.399 --> 00:11:48.460
and coding. So instead of the back and forth

00:11:48.460 --> 00:11:51.059
chat bubbles, it opens a document window on the

00:11:51.059 --> 00:11:53.240
side. So you can work next to it. Exactly. You

00:11:53.240 --> 00:11:54.679
can highlight a paragraph and just say, make

00:11:54.679 --> 00:11:57.159
this shorter. Or highlight a sentence and say,

00:11:57.419 --> 00:11:59.440
check grammar. It becomes a side by side writing

00:11:59.440 --> 00:12:02.519
partner. And we can't forget web browsing. Browse

00:12:02.519 --> 00:12:05.100
with Bing. That one's crucial for relevance.

00:12:05.649 --> 00:12:08.570
Without this, the AI is stuck in the past. Its

00:12:08.570 --> 00:12:11.029
training data has a cutoff. With browsing, it

00:12:11.029 --> 00:12:13.950
can check stock prices or current news. It connects

00:12:13.950 --> 00:12:16.049
the brain to the live internet. Now the final

00:12:16.049 --> 00:12:18.250
frontier in our source material is customization

00:12:18.250 --> 00:12:20.950
and agents. This feels like the biggest shift

00:12:20.950 --> 00:12:24.129
of all. It is. We are moving from the AI being

00:12:24.129 --> 00:12:27.350
a talker to a doer. Can we make that concrete?

00:12:28.210 --> 00:12:31.590
Because agent is such a buzzword. What does an

00:12:31.590 --> 00:12:35.179
agent actually look like in 2026? Okay, imagine

00:12:35.179 --> 00:12:38.279
this. You don't just ask the AI how should I

00:12:38.279 --> 00:12:40.220
plan a dinner date. Right. You tell it, plan

00:12:40.220 --> 00:12:42.179
a dinner date for me and my wife this Friday.

00:12:43.039 --> 00:12:45.299
The agent then checks your calendar for when

00:12:45.299 --> 00:12:48.399
you're both free. It goes to OpenTable and finds

00:12:48.399 --> 00:12:50.480
a reservation at an Italian place you've liked

00:12:50.480 --> 00:12:53.360
before. Then it drafts the calendar invite for

00:12:53.360 --> 00:12:55.980
your wife. It doesn't just give you advice. It

00:12:55.980 --> 00:12:58.200
executes the task. It connects to the apps. To

00:12:58.200 --> 00:13:00.820
Google Drive, your calendar, your email. Exactly.

00:13:01.080 --> 00:13:03.320
It's editing the files, not just talking about

00:13:03.320 --> 00:13:05.519
them. So what does this all mean for us then?

00:13:05.940 --> 00:13:08.539
If the AI starts doing rather than just talking,

00:13:08.759 --> 00:13:11.720
Does my role change? Fundamentally, yes. You

00:13:11.720 --> 00:13:14.019
shift from being the creator, the one actually

00:13:14.019 --> 00:13:16.539
typing the email, to being the manager. You're

00:13:16.539 --> 00:13:18.659
just supervising the results. So I become the

00:13:18.659 --> 00:13:20.559
editor -in -chief of my own life. That's the

00:13:20.559 --> 00:13:23.100
ideal, yeah. But it requires a different skill

00:13:23.100 --> 00:13:25.120
set. You have to be really good at spotting errors

00:13:25.120 --> 00:13:28.220
and really good at giving clear, concise instructions.

00:13:28.600 --> 00:13:32.120
So to recap the big ideas here, we started with

00:13:32.120 --> 00:13:34.980
the importance of settings, privacy, memory to

00:13:34.980 --> 00:13:38.309
stop that drift. We move to projects that hold

00:13:38.309 --> 00:13:41.769
our context, canvas for co -writing, and finally

00:13:41.769 --> 00:13:44.690
these agents that actually do the work. And if

00:13:44.690 --> 00:13:46.409
we look for the thread connecting all of this,

00:13:46.649 --> 00:13:49.690
it's customization. The tool is only as powerful

00:13:49.690 --> 00:13:51.990
as the settings you actually decide to turn on.

00:13:52.409 --> 00:13:55.190
That hidden menu isn't just about features. It's

00:13:55.190 --> 00:13:57.789
about moving from a passive user to an active

00:13:57.789 --> 00:14:00.309
architect. So here is the challenge for you listening

00:14:00.309 --> 00:14:02.509
right now. The source material is very clear

00:14:02.509 --> 00:14:05.590
on this. Don't try to use all 17 features we

00:14:05.590 --> 00:14:07.309
talked about today. You'll just get overwhelmed.

00:14:07.429 --> 00:14:10.570
Just do one thing. Go to settings, go to personalization,

00:14:10.909 --> 00:14:15.110
and open custom instructions. Spend five minutes,

00:14:15.129 --> 00:14:17.710
just five minutes, telling the AI who you are

00:14:17.710 --> 00:14:20.070
and how you want it to speak to you. It is such

00:14:20.070 --> 00:14:22.669
a high leverage five minutes. Once you do that,

00:14:23.190 --> 00:14:24.990
every single future conversation you have with

00:14:24.990 --> 00:14:27.450
it improves. Stopped driving the Ferrari in first

00:14:27.450 --> 00:14:29.429
gear open the menu. Thanks for diving in with

00:14:29.429 --> 00:14:31.210
us today. Always a pleasure. See you next time