I found myself recently as the lazy quiet of summer gave way to the noisy insistence of the fall
being more irritable than normal. I would get really annoyed about the latest blown save for the
Washington Nationals or I'd walk into a cluttered room and in a frantic fit decide I had to clean
that room right away because it was intolerable the way that it was currently laid out. Now I'm familiar
with this routine because it's my mind's way of telling me that now that the fall is beginning
it's time to turn back on some more serious organizational systems. I slow down for the summer
but when the fall starts my mind begins to push back and say we got to get more structured.
So I did last week I sat down and figured out the exact setup for my organizational systems
that I'm going to run this fall including the specific tools I will use in the processes I will
use with those tools and this is what I want to talk to you about today. I'm going to put on my
productivity nerd hat and walk you through the exact setup I will be using this fall to organize
my life but I'm not just going to leave it there. I'll then step back and talk about a why it's
important in general to periodically update or change your organizational systems and be give you
some best practices for building a system that fits your needs for this upcoming fall and in
case you're wondering my productivity nerd hat it features a David Allen making out with a Trello
board. It's just pretty disturbing. All right so we got a lot of advice and ideas to cover in
this episode so let's get right into it. All right so here's my plan. We're going to proceed
in three parts. In the first part I'll go through the details of my system for the fall of
2026 then we'll talk about why tweaking your systems like I do is actually a good thing and then
we'll talk about best practices when you're creating your own system for this upcoming fall.
Now a quick disclaimer before we get started I should briefly address the meta question of
why do we care about organizational systems at all. I mean didn't Jenny Odell teach us that
productivity is simply the subordination of human individuality to the industrializing logics
of late-stage capitalism. In other words there's a bit of like hey let's not talk too much about
organizational system vibe on the internet these days but here's why I think it's important right.
My main role as a computer scientist who studies human flourishing in a technological world.
One of the things I've learned is the technological world throws a lot of distractions and a lot of work
and a lot of shiny objects at you and if you do not have an intentional plan for how to organize
what you have to do and to aim intentionally your time and attention the technology will make
these plans for you. They're happy to hijack your day in a way that's going to allow Mark Zuckerberg
to buy the second half of Kauai. I don't know what that when he does out there but maybe involves
David Ellen making out with trailer boards. I don't know what tech people do. So I do think it's
important to talk about organization not because God was this kind of tiresome rhetoric about we're
always optimizing ourselves because we're trying to we're trying to internalize late stage cap was
no if we're not organized. Technology will organize our lives for us all right so that's my
disclaimer. Let's get in with it part one I want to talk about my system for the fall of 2026.
Now I'm going to proceed by first talking about the material so that is like the specific tools
I'm using and then we're going to talk about the methods the particular rules or systems I'll use
to sort of integrate those tools into my life. I want to give some setup here. This is not a normal
fall for me because I'm on sabbatical so I'm running a more lightweight system this fall than I
normally would in a typical year. However I've been told that even my lightweight systems would be
considered pretty robust for a lot of other people but I just want you to have that that note in mind.
All right so let's start with the key tools that we're calling materials that I am using in my fall
system. All right the first material is where am I storing my tasks this fall? Well because I'm on
sabbatical I am storing them all in a computer program called things three. I'll put it up on the
screen here for people who are also watching here's the website for things. I have a couple screenshots
here so what you can see is things exist you can have it on your phone you can also have it on
your computer you buy them separately for each device which I think is kind of cool it's not some
software service thing and odd it sinks up I'll zoom in here a little bit on the main interface
it's pretty simple right you have a list down here of different you have different to do list
different categories like family buy a new car through a party for Eve you can put dates on list
and then you can also look at on task you can look at like hey which which tasker do today or which
ones are upcoming the nice thing about things what I like about is the interface is very easy it
tries to simplify adding things removing things adding dates to things it makes as easy as possible so
very quick to capture very little actual overhead that's what I'm using for this fall in a more
busy fall like when I have multiple administrative or research roles at Georgetown I would typically use
things for home things and I might use trello for my major roles because if I need to have more
complicated information or I need to store files along with tasks then I like to use trello boards
per role but in a simpler fall like this I think things will be fine all right the next piece of
material next tool I'm going to use is a time block planner I have one here I kept the the ribbon on
it obviously I sell my own time block planner so I have the advantage of using one that's planned
by me you know it's seen a time block planner I'll hold it up here you got pages over here for time
blocking then over here for storing tasks give a shutdown complete check box in a place to put
metrics okay so I will be using my time block planner I get four uses out of mind planning my day
capturing tasks and I'm away from machine the write them down shutdown routine justification
there's a shutdown box and for tracking my metrics the third tool I use is probably the most
typical and boring which is a calendar but you need it obviously that's that's actually the one
productivity tool that most people actually reliably use is they will always check their calendar
I use a Google calendar now I actually have multiple Google calendars and I can put them all on
the same display maybe the most interesting tweak I do is that I have a calendar that's called
cal logistics separate from the normal calendar I use for appointments so I can put notes to myself
that aren't actual appointments but reminders about things or things like somewhere in here you
need to make sure that you catch up on whatever it's a way for me to actually have information on my
calendar that I see and it's connected the days or time but it's not looking the exact same as an
appointment I share my appointment calendar with my wife so she knows when when I'm going to be
aware when I'm on calls or whatever but I don't share the logistics because those are just points
to myself all right the final key tool that I have is a vision document all right so this is
something I print out and is going to contain the sort of the core ideas for the season ahead of
mine I'll put I made a sample one I'm going to put on the screen for people who are watching here
all right here's a sample vision it's it's kind of joky but but it has the right the right
elements you need for a vision so what the sample vision has is at the top of it a description
of like hey what's the big picture vision for this season it then has a list of like the major
projects and practices like what are the big things that we're hoping to get done this season outside
of just a normal flow of work and small tasks like what are the big one of the big things maybe optional
but the big endeavors are going to move me forward to bigger goals or practices I want new practices
I want to put in place that I'm really focusing on so you have like your big picture hey here's
the things I'm focusing on this season and then materials and methods and I think this is critical
it just list okay this is my system that I'll be using this I'll be using the season right
now think about it somewhere this should be written down if you're just keeping the stuff I'm
talking about today your organizational system if you're just keeping the system in your mind
here's I'm going to organize myself this fall it's an open loop it's in there it's floating around
and you don't want to forget it what if I forget it what if I forget that I'm using my time
block planner that I shouldn't remember to shut down and so I like to have in my vision document
just a quick summary of here's the whole system you're using now this is a little bit circular
because one of the things that might be in that that the list of your productivity systems is like
and I keep a vision document right so maybe it's a little bit circular but I want to have that
written down somewhere this vision document then becomes your root it becomes the root to your
entire productivity system printed out put in your top drawer or your socks are because you go
there every day and you're like this document from here I can get to everything else I'm doing to
be organized because I have my my goal for the the season the major product to try to make progress
on and the productivity systems that I'm running and if you look back at this sample one that I
created let me go back here under this materials and methods section where I'm kind of listing my
productivity system one of the key things I have is specific metrics right so if for example you do
metric tracking which I do and I I recommend is a way to sort of keep up on practices you and do
on a regular basis you got to write down somewhere these are the metrics I track every day so that's
one the things that's there that's where I would also say I'm using time block planner I'm using
things for my productivity so somewhere you want one document it captures what your systems going
to be as well as your big vision for that season that's printed out put in your sock drawer so you
can't avoid it that's the root for everything else that follows all right so those things together
we'll we'll describe as my materials so just
This is a reminder.
It was things three for task.
It was time block planner for planning metrics shutdowns and capture.
It was my Google calendar for Q and check of appointments and notes to myself and a vision
document to help orient me towards what this, what progress I want to make on big things
this season and what my description of this whole system so I don't have to keep that in
my head.
All right.
So now let's move on to methods.
So methods are the actual rules or processes or practices that you'll use to implement
your system.
So you have the tools, but you also have to have your methods.
Let me put up on the screen here, the core methods that I am following this fall.
So I have three core methods for using those tools.
Number one is time blocking.
Now that means multiple things for me, right?
So I wrote down right here weekday plans shutdown and metrics, right?
So on the weekdays, I time block plan the day on the weekdays, I do shutdowns, shutdown
complete.
I check to check marks after I've done my shutdown ritual and I keep track of metrics when
I do my shutdown.
So I'm putting all of that in next to time block plan and my methods, right?
For task, the method I'm going to have is 20 minutes a day at least, minimum of 20 minutes
a day going through things and trying to make progress on task.
Now that's a very minimal way of dealing with tasks, like I'll be honest about it, when
I'm in a busy semester, it's much more complicated when I'm using my trellabords and rolls
and capturing and processing, but for a sabbatical, this makes sure I'm not going to forget things
that need to be done because I look at my to-do list every day and I do non-trivial
to-do list work every day.
That really helps you with my irritability because when you're not fitting in that 20 minute
minimum of just making productive process on small things, both at work and at home is
where you begin to feel like you're falling behind.
And then the last thing that I am including on my methods for this fall is my weekly planning.
That's why I wrote this down here.
So what is my weekly planning?
This is where each week, I look at the week ahead, see what's going on, move things around
if I need to, I review my tasks, and I review my vision.
I actually wrote that down here because that's key.
That's when I look at my vision, why is this important?
Well, often the best way to make progress on these big initiatives in your vision is
when you're looking at your week ahead block off time to make progress on them.
So that is my time.
I'm going to protect Friday morning and Wednesday afternoon to make progress on whatever big
initiative I want to make progress on.
So weekly plans, you get your head on straight for the week, you see what's coming, you
remind yourself of your vision, you make sure that your calendar reflects what you want
to do.
You kind of reset, look through your things list, make sure they're organized, et cetera,
get kind of reset for the week ahead.
So again, this is a very stripped down organizational system because I'm on sabbatical.
But again, I'm time block planning and everything goes along with that.
I'm spending 20 minutes a day as a minimum in things that I'm weekly planning.
That actually, I can tell you from experience, that is a powerful system.
That seems simple.
Run that system with your time block planning, weekly planning, 20 minutes a day in tasks,
doing a shutdown routine, keeping track of metrics.
That can power a life that's going to feel like you've got your act together, like you're
getting things done, you're not forgetting things.
You're not stressed.
You're making picture on big projects.
You're not dropping the ball in small projects.
The AC filters are getting changed.
Like that is a system that I think is very flexible and powerful.
Again, I would probably have something more complex if it was a more demanding semester.
But that could apply to most people's situations, in my opinion.
We're going to think a quick break from our discussion of organizational systems to hear
from some of the sponsors that makes this show possible.
But stick around because coming up, I'm going to get into the best practices for how
you can create your own custom organizational system.
All right.
So let's talk about our friends at Superhuman here.
Too much of the talk about AI these days focuses on abstract benchmark scores or attention
grabbing stunts, like winning a gold medal or a hacking agent on a ride.
But do you know what I want out of AI to reduce the time I spend on administrative and
logistical efforts so that I have more time left to focus on the deep things that really
move the needle.
This is why I'm excited about Superhuman Go, a chat-based AI assistant that's with you,
wherever you're already doing your work, already aware of what you're doing and ready to
help.
And let me give you a specific example about Go that I find really exciting.
Let's say you're writing an email to your boss about the status of some project you're
leading.
And you realize that you want to include a report based on the current state of the project
in your JIRA ticket management system.
Well with Go, you don't have to leave your Gmail at all.
You can just pop up a Go window, type in, add a quick summary from JIRA at the end, and
boom, off it goes, gets the information on JIRA, summarizes it, adds it to the bottom
of your email, all without you having to go to another application or do this work yourself.
All right.
I'd like Go because it lets you focus on the things that matter without all of the brain
destroying context switching or busy work, no new tabs, no switching apps, get what
you need without losing where you are.
To find out more about Superhuman Go, visit superhuman.com, that's superhuman.com for
more information.
I also want to talk quickly about our friends at Monarch.
Look, some of my proudest moments from earlier in my life was hitting key financial milestones.
I still remember the day I finally paid off all of my student loans from college, but
you know what I wish I had back then?
Monarch.
The personal finance app that goes beyond just telling you what you've spent to also help
you set goals, map out big purchases, and see if you're actually on track before it's
too late to adjust.
For example, Monarch can send you a weekly AI recap of your spending, which makes it easy
to discover if there have been major changes or other spikes in your habits that are causing
problems.
Monarch has taken the mental load of tracking finances off my plate.
Just like having your own financial advisor in your back pocket, any question about your
finances, you can just ask the built-in AI and it will recap and simplify it so that you
aren't chasing receipts or trying to make sense of complicated spending.
If I had had access to this app back when I was first starting out, I would have hit my
key financial milestones much quicker.
This is why you should check it out.
So write your own money story with Monarch, use code deep at Monarch.com to get your first
year of Monarch Core half off at just $50.
That's 50% off your first year at Monarch.com using code deep.
All right, let's get back to our show.
All right, part two, I want to address the question of why do I keep tweaking my system?
Why don't I just have the best time management organization system in the world optimize
it and just run it every day for the rest of my life?
Well, I don't believe in fixing one optimal system forever.
I have two reasons why number one, what is it that kills systems?
What is it that stops people from sticking with their organizational systems?
It's almost always friction.
These little bits of friction, I have to do a little bit of extra work.
I got to go into this app and transfer things and go back over to this app.
I have to load this up and go into this interface and move things from here to there.
Then kill systems, not at first, when you have a lot of time and you're excited, you
do all the things, but as your life gets busier, the friction adds up and becomes intolerable
and eventually, like, forget this and then you just drop having a system altogether.
So friction kills systems.
So to reduce friction as much as possible, you got to be adaptable.
So what is the reality of this upcoming season?
I want the minimum possible system that still solves my problems of reducing stress
and anxiety and making sure I make progress on what matters and not dropping the ball.
And in the summer, you might pull this back a lot to a super simple system so you can
recharge.
In the fall, it might get more complicated.
Maybe the winter you pull it back again, but you always want to be adjusting to the reality
of each current season that try to find places where unnecessary friction or complexity
can be removed when you can.
The second reason why I think you should keep tweaking your systems is novelty.
You will get bored of a system if you use it forever and you'll get tired of it.
It's great because when you adjust a system for a given season, you get excited at first
about, oh, I like this new system.
I feel good about it.
It's exciting to try something new.
That starts to wear off and just around that time, you're creating a new system for
the new season.
So novelty will actually keep you, and I think this is critical, but novelty will actually
keep you more energized about being organized in general.
So I'm going to suggest the same for you, fall spring summer at a minimum, adjust your
organizational systems.
You might do quarters instead, fall winter spring summer.
That's up to you, but that's the right way to think about it.
Have novelty, have some seasons be more intense than others, and try to keep your system
as minimal as possible for the demands you happen to have in the season that you're facing.
All right.
Third part of this discussion, what are best practices for you designing your own system?
Now the one I just described to you for my fall is pretty good.
That's pretty flexible.
It's pretty easy.
It's pretty low friction.
If you're not sure what to do, do something like that.
But people have different situations, they have different complexities, you know, you
got to custom build a system of what you're facing.
I want to give you some best practices for how to do this.
I have a helpful diagram here I might load up beautifully drawn.
I bet Jesse's not here today because he would appreciate how beautifully drawn this
is.
So if we look at my diagram here, I'm best practices for creating your own system.
At the top of the diagram are three questions.
What, when, and why?
All right.
So what does that mean?
Well, the what is, I'll put it this way, these are the three questions you have.
have to answer when creating an organizational system.
The what is, what do I have to do?
This is where you keep track of your task.
Where are your tasks?
Where do you store your task?
How do you review your task?
How, like, what is your way for managing
all the different things you have to do?
The wind stands for, when am I doing it?
Right, so you need some sort of method
for how you actually decide what to do when.
Mine is built around time block planning and weekly plans.
Other people have, they run their life off of a task list.
If you're using, like, David Allen's getting things done,
he suggests answering the wind question
with physical context.
So he divides all of his tasks in the context.
And then when you find yourself in a given context,
you just look at the list for that context.
Like, computer admin work.
I'm at my computer doing admin work on the internet.
You have a list of computer admin work
where you've moved tasks just in that list.
You just load up that list and choose things off of it.
Like, that's the David Allen answer to that question.
If you go back to, like, the 1970s, 1980s,
you're going to get priority systems.
ABC rankings for my task.
I do all the A task, most of the B task,
and only the C task if I have time, right?
So there's different ways to answer this question of,
when do I do tasks?
How do I decide what I'm doing next?
Well, you need an answer for that
in any reasonable organizational system.
And the why stands for, why am I doing that?
Which means you need an explicit way of managing your workload.
What you say yes to, what you say no to,
how many things you work on concurrently
when you work on them, et cetera, right?
So these are the base questions
that any reasonable organizational system needs to ask
is what, when, and why?
Okay, now once you've, those are the questions.
How do you answer them?
Like, what formed your answers?
Take, well, the form that we just saw,
you need materials and methods, right?
So then you're like, okay,
well, what tools am I going to use
to implement my answers to those questions?
And what rules do I have about using those tools?
And so those are your materials and methods.
That's a good way of breaking up the way you think about
your organizational system.
And all of this should be built on a foundation of vision.
What am I trying to do?
What's important to me?
What are the things I want to move forward on?
So you have a foundation of like a vision for your life,
but more specifically a vision for the season
that you're currently facing.
So to me, I think that's really important
that you're able to ground your organizational system
for a vision for like what it is you want that,
what is that season going to be like?
And it's not, this is not just accomplishments.
Here's my goals.
It might be like, I want this to be a quiet season.
I want this to be a season, you know, it's summer
and it's about family or like the fall
really should be about Halloween decorations.
And I want to kind of build my life around
that you need some sort of organizing vision
that you're trying to actually move towards.
Otherwise, you're just going to become a task executor.
Hey, I got a great system.
I execute all these tasks.
And I'm sort of a drift and drowning in on we.
All right, so let's just look at this one more time.
Your organizational system has to answer
the questions what, when and why.
Those answers should take the form of a description
of what materials you use and what methods you deploy
around those materials.
Everything should be grounded in a vision.
It's custom fit to the season
that you're currently facing.
All right, so there we go.
Those are the big issues about the fall, all right?
I told you what my stripped down organizational system
is going to be for this fall.
I told you why you should keep changing
your organizational systems per season.
And then I gave you some best practices
for how do you create your very own organizational system?
What are the best practices for creating one
that's going to work for you,
but covers the thing that these systems should cover?
I would like to hear what you end up with.
If you want to share whatever organizational system
you're using this fall,
you can send it to
[email protected].
How do you answer what material is a method you use?
And if you want to share your vision, you can.
And maybe in a later episode,
I will share some of these on the air.
All right, so that's enough about what I'm interested in.
Now I want to know what you're interested in.
It's time, as I often like to do on Monday,
to open up the inbox.
If you have a question or something you want to share
or feedback on a past episode,
send it to
[email protected]
and maybe we will deal with it on the air.
And we don't have Jesse today.
He's dealing with some family health issues.
We wish him our best,
but it means I will have to handle all the questions today.
So you will have to bear with me.
All right, message number one.
Methods number one comes from Catherine.
And Catherine says, "I thought you would find a comment below
from Dervla McTernan and Irish novelist in Australia
about her use of AI interesting."
All right, so Catherine sent a link to a newsletter
from Dervla McTernan and Australian Irish novelist
living in Australia.
All right, so I read that newsletter
and I pulled out here the key quote in my opinion
from this newsletter that Catherine sent me.
So here's the key quote from this novelist newsletter.
I don't use AI at all for my writing,
but I do use it for a lot of the business
that surrounds writing.
And it feels like I suddenly have the ability
to do a lot of things that I can only have to in the past,
which means of course that I feel like I should do all of it.
I've been frustrated for years by some of the constraints
that I've felt in my work,
accessed information, the ability to use certain systems
really well, like Canvas design or meta ads
or the intricacies of the mail chimp back in.
And now that those constraints have largely been removed,
I'm just incredibly tempted to spend all of my time working.
All right, I think this is a really good point
because it actually dovetails was something I've talked about
for years here on this podcast that AI is making worse.
All right, so stay with me here.
There's a concept that I've always been wary of
that I call checklist productivity.
And it's this idea that when it comes to some sort of
important endeavor in your life,
that there's a way to not gonna say hack it,
but the checklist productivity mindset is,
if you can get the right information
about the steps that really matter and execute these,
like checking items off of a checklist,
then you will find great success
in this really important endeavor.
And what's compelling about the checklist productivity concept
is that it moves the necessary action for success
away from deep effort on something hard
that's competitive and that you might not succeed with
to information scarcity.
No, no, no, you can, if you get the right information,
these are the right things to do.
You could just execute those and you will have success.
And this is really appealing because executing the items
and checklist productivity is not competitive.
It doesn't typically require deep work.
It's more mechanical and it's often much more fulfilling.
You're setting things up, you're moving things,
you're seeing immediate results.
And so we want this to be true.
We want it to be true that we can find a secret list
of things to do and if we just do these right things,
we'll be a successful author.
We'll be a successful professor.
We're going to suddenly have promotions at work.
So checklist productivity, it's very infectious.
It was really big and like West Coast productivity
type literature in the 2000s.
Hey, just get the right information.
You can optimize, like think about some pieces
of like a four hour work week,
had a real checklist productivity vibe.
Here's the secret information, do these things.
Each of which is like completely
comfortable and predictable, do these things and success.
It'll add up to success.
I'm not a big believer in checklist productivity
because it doesn't work.
For hard things, it's the hard things that matter.
It's the creation of the art.
It's the rare and valuable thing.
There's no way to avoid deliberate practice and deep work
and grinding and straining and it's uncomfortable
and you get better and a long winding road
to the pride of competent accomplishment is hard one.
That's almost always what matters.
There's no magical list of to do's that you can execute.
That's going to get you the same level of success
of just doing the hard work of becoming too good
to be ignored at something that you value.
So checklist productivity, I feel like,
is a little bit dangerous.
Well, AI super charges checklist productivity.
Because it puts out your fingertips
like Derva was talking about in this newsletter,
the ability to execute all sorts of checklist productivity
style things, like let me read the examples
from her newsletter.
These are perfect examples of checklist productivity.
Using Canvas design better or using meta ads better
or using your MailChimp newsletter software better.
All of these sort of like technical things,
you're like, wow, I've optimized my setup on Instagram.
I have tags now for incoming subscribers
on my newsletter management software
so that I can segment different types of messages to them.
I'm using Canva in a more advanced way
so I can have really cool looking memes
that I put up in my TikTok post, like whatever it is, right?
All these type of things, you feel super productive.
I'm doing all these things,
but they don't add up typically to real success.
That almost always requires you doing the thing
that's really hard and you have to be too good to be ignored.
So AI vastly opens up access to checklist productivity style
activities and makes them very fun and fulfilling to do
because you're getting like step-by-step instructions
from a chatbot or if you're using agent
or even do some of this for you.
It's addicting, but it's not going to make
you a successful novelist and it's not going
to make, it's not going to give you a promotion at work.
It's not going to make you, you know,
your newsletter suddenly take off.
Ultimately, you still have to be doing something hard
that's rare and valuable.
If checklist product, if it worked,
we would just automate it and immediately
we would have 100,000 people doing that same thing
and the value would disappear.
So my advice is constrain checklist productivity time.
It's not that those activities aren't useful
but they're not the solution.
And so you want to constrain that time,
maybe have one checklist productivity style
initiative per week that you're working on
and still keep the main thing to main thing.
The core of your day is doing the hard work
on the thing that ultimately produces value
that you're getting better.
and is becoming more rare and valuable.
Do not drown in checklist productivity actions
just because AI makes it possible.
All right, message number two comes from long time reader.
I felt like during law school,
I kind of lost my way a little bit
with my relationship with technology.
Maybe even as a way to cope
with the struggle of being in university.
I ended up falling into a pretty bad social media addiction
with all the professional networking opportunities.
I found it overwhelming to say the least,
having 24/7 access to a variety
of different professionals I could reach out to,
reading up to date articles
and a variety of different legal landscapes.
Overall, it was just a huge waste of time.
Anyway, I recently started studying for the bar exam.
I found myself over the time between graduating from law school
and starting working on the bar,
rekindling my relationship with the deep life.
I've gotten back into many of your suggestions
such as reading, having physical activities,
mental activities, skill-based activities,
things like that.
It's been a wonderful, real awakening,
getting back into those principles
that made me successful in the first place
and trying to make you more balanced
and methodical relationship with the tech products
that I do use.
This is sort of more of a case study than a question,
but I love to hear it, right?
We talk abstractly about things like the deep life on here,
and it's like, why?
This is why.
Because if you do not have an intentional vision
for your life, if you don't live on purpose,
the 21st century technology is gonna do the living for you.
It's gonna pull the strings.
It's just your time and attention is so valuable
that you will lose the battle,
the workplace and personal technologies
if you're just wandering through life,
trying to get random accomplishments and feeling good.
You have to live on purpose,
to flourish in a technological world.
And all these ideas about the deep life, et cetera,
this is why.
I mean, look, I have this new book coming out
in March called The Deep Life.
The subtitle is a structured approach.
I called it one-off, but people push back last time.
It's not one-off.
It's at the core of what you do.
And I think it's right.
Even though it's not a book of technology criticism,
it's a book about creating the bulwark to technology.
When you live on purpose,
when you have an intentional life that's deep and remarkable,
the alert of TikTok, of LinkedIn, of Instagram,
of making your workday devolve into constant,
slack conversations and teams meetings
so that you seem like you're a team player.
All of that begins to diminish
because you're like, no, I'm intentional.
And that's not so great.
What I'm doing over here is more important.
And so the deep life,
we can't pull apart the flourishing
from reacting to the technological world.
All this goes together.
You have to control your analog life,
the pushback on the digital.
Are we gonna take another quick break
to hear from some of the sponsors
that makes this show possible?
But stay tuned because at the end of the inbox session segment,
I go on a rant.
I can't help myself.
You don't want to miss it.
But let's get to these ads real quick.
I probably should have started this podcast earlier.
I knew it was a good idea,
but I was paralyzed by the technical complexities
and getting a show like this up and running.
Now, I don't want you to follow in the same trap
if you're thinking about starting your own business
but are worried about the complexity
of actually selling things.
Here's the secret to eliminating those fears
so you can get going right away.
Use Shopify.
It gives you everything you need to start selling
on day one so you can focus on what really matters
which is growth.
Here's one feature in particular I think is critical.
All Shopify storefronts are integrated with checkout
or no extra charge.
Among other benefits, this offers your customers
one click checkout which can significantly improve conversion.
Look, you don't want to be wasting time on your business,
fixing issues and getting technical platforms running.
Let's Shopify handle these details
so you can grow your business on your own terms.
With Shopify, nothing stands between your idea
and an actual functioning business.
So go make it a reality.
If you're ready to start putting off your business
and start hearing, then you're ready for Shopify.
Start your free trial at Shopify.com/deep today.
You heard that right.
Start your free trial at Shopify.com/deep.
That's Shopify.com/deep.
Hey, are you looking to get in better shape
but are struggling to follow through?
Here's the solution.
My body tutor, T-U-T-O-R, an online coaching program
that actually delivers results.
Now, how does it work?
When you sign up, you're assigned a coach
and they help you devise nutrition and exercise plan
custom fit to your specific goals and circumstances.
And then in the sticky part, you quickly check in
each day on the my body tutor app
to let your coach know how things went.
It's this accountability that really gets you to follow through.
It also puts an advocate in your corner
who can help you adjust your plan
to whatever life those your ways
or notice where you're struggling and offer help.
You want to get healthier?
Go to mybodytutor.com today.
And if you do, mention I sent you
and they will give you $50 off your first month.
That's mybodytutor.com.
All right, let's get back to the show.
Our third message comes from Sandra
who simply writes, "Apennian, please."
She then links to an article that is titled,
"The Sharp Rise and Incidents of AI
Escaping Users Control Research Finds."
And it was an article about,
not the, I mean, it mentions the hacking attacks
from this summer that I've been talking about
in some of the tech-to-coded episodes of this podcast.
And in my newsletter, calnewport.com,
I've been talking about these attacks.
But the report is also finding that other smaller instances
where individual users running agents are finding,
oh my god, it deleted my files
or did something else I didn't want the agent to do.
So there's a sharp rise of such reports.
So I sent Sandra a response.
I'm actually gonna read the email I sent back to Sandra
giving my opinion to this article
and then I'll elaborate a little bit.
But here's what I said to Sandra.
In this article, when they say AI,
they are actually responding to a very specific type of system
in which a computer program executes a long-running loop
in which it repeatedly prompts an LLM,
ask it what it should do next
and then automatically executes it.
Given all of the unreliability, erratic behavior
and unpredictability of LLM outputs,
such a system is 100% guaranteed
to eventually end up doing something you didn't expect
or like, this is not a problem with AI,
but a problem with a specific type of irresponsible system
that the LLM companies wanna keep convincing us
are critical.
This is at the core of my critique right now
of what's going on.
I don't wanna get too much into this
'cause I do this more like my newsletter
and other types of episodes on this podcast.
But just like at a really high level,
every time there is bad news,
surrounding the health of the businesses
of the major frontier labs,
they drop in our lap like a cat leaving a dead bird
on our doorstep.
Some sort of anxiety-producing AI safety tinged.
LAs are Yucowsky fever dream style story about AI
and that's what everyone talks about.
And I think this happened this summer.
People started pushing hard on the financials
of these companies.
Open AI began losing top employees again and again.
They've had several incidents in the past few months
where they hired a top employee
and they left within a few weeks.
The economics aren't making sense
and analysts are starting to push back on this.
They're pushing back on the CAPEX pin.
They're doing the math and realizing wait a second
for you to get your projections.
For all of this CAPEX pin, for example,
on AI data centers that make sense,
there have to be trillions of dollars of revenue.
And right now, we mainly know how to do coding agents,
which is like a good business, but not a massive business.
And it's struggling and also is not,
doesn't really require frontier models.
So this is kind of a difficult place for the big companies.
And like Google search where it costs more money
to power the search with these LLMs,
the search results than it is money
you can actually make selling ads.
So there's some financial issues right now.
All this being pointed out open AI delays is IPO
because the numbers aren't there.
People are jumping ship and what do they say?
Do they come out and be like,
"Let's explain the fundamentals of our business."
Here's what's going on.
No, they say agent civilizations are trying to take over.
And boom, that's what everyone's talking about.
So they do this and it's annoying.
But here's the point I'm trying to make in this response
is all of this sort of summer's hysteria
about AI getting out of control
is about a very specific type of AI system
that is guaranteed to get out of control.
Guaranteed.
It's not AI in general.
It's not AI in some general sense
as it gets more capable is misaligning with human intentions.
It's a specific type of system
where you have a call it a loop prompt
where a prompt loop.
We just have a program written by human beings,
nothing fancy here, that ask an LLM.
Hey, here's what I'm trying to do, what should I do next?
LLM says do this.
It executes that.
Then it comes back, updates, it's okay.
Here's what I want to do.
Here's what's happened so far.
What should I do next?
And that's it.
And maybe it generates other prompt loops
to take on particular steps to keep the context clean.
So it kind of swarms out and comes back in.
But that's what it is.
It's just blindly executing what an LLM suggests.
Now this only really works in the world of computer terminals
because in computer programming or cyber security,
the actions can all be text commands you can run on a terminal.
So it's very easy for an LLM response to say run this command.
It can describe it in text or run this script,
which it can describe in text.
And then the agent program can just run it directly
on the terminal.
That's a very specific type of system.
It's a really bad idea to run this for any sort of duration
because LLM outputs are unpredictable.
There meant to be electrographically plausible,
not to be normative.
They don't follow human norms.
They don't have consistent or symbolic.
like planning. They don't have value systems. It just sort of spits out. I think this is a
electrographically plausible thing I just output. This can have very powerful ideas like this is
a really good hack, but it also means they're going to drip all over the place. They make things up.
They fall into narrative tropes. If in your prompts like you're an AI system and I need to know what
to do next, it's going to likely fall into sci-fi tropes, right? Because we know this from research.
So it's just a very dangerous thing to do outside of like here's a very narrow problem and you can
run a prompt loop to try to do multiple steps of a solution so you don't have to do it. But in a very
narrow context dealing with computer programming terminal based tools, this can be useful, but you
got to be really careful. The computer programmers know this. The professional computer programmers
really have been constraining the horizons on which they send these prompt loops out to handle
things because they go off and do weird things. And so these incidents of losing control is just people
running their own prompt loops using frameworks like OpenClaw, which lets you write your own prompt
loops, write them on their own machine, give them access to stuff on their machine, and because
they're just asking an LLM, what should you do next, what should I do next, they're erratic and
unpredictable and they do erratic and unpredictable things. This is not AI writ large. There is no AI
writ large. This is an LLM and a prompt loops company program that's prompting the LLM and doing
what it says. And anyone who's used a chatbot knows there be dragons. If you do that long enough,
you're going to get into a weird place, right? Fool me once, shame on you, fool me twice, shame on me.
So again, there's distraction going on here, prompt loops run for a long period of time,
like OpenAI did this summer. It's just irresponsible. A course is going to do crazy things,
not because SkyNet is getting upset, but because you put a weed whacker on a dog. It's an
unreliable source of action. Don't blindly execute LLM outputs as actions. A course bad stuff will
happen. All right, we got our fourth message here. It came in anonymously. And it said, I thought
you might find this Reddit post interesting. And it pointed to a Reddit post whose title was
the hype is waning even for large corporations. Now, let me read you the key paragraph of this
Reddit post. I have two colleagues from university who work for two major financial firms. Over the
past month, they have been talking about how many problems LLM are causing behind the scenes and
that the cost is coming due for the upper levels. They are moving away from all enterprise AI
tools from OpenAI and cloud. It's over and this starts the slide. Interesting. There's a lot of
replies to this. I want to read two replies from farther down in the Reddit thread. Here's one
that caught my attention. I work at a large non bank financial services firm. So lots of data,
lots of spreadsheets in the county, lots of legal docs, lots of personal data. We can't structurally
use LLM's because it would upload all the highly confidential client and KYC data. So if we use it,
it's a double checks and procedural legal requirements via prompt like Google search more often
than not. It's wrong because it doesn't use the right sources or even read them correctly.
Law firms rarely give clarity for free on their website newsletter, but the LLM will state anything
as certainty even when there is none or it has no answer for the prompt. In other words, we have no
existing use cases. Here's another reply I want to read. I'm hearing lots of stories about engineers
browsing on their phones while letting cloud do a bunch of work, walking the dog, playing games,
all on company time that doesn't lead to better outcomes for the company, just an engineer burning
a bunch of tokens in a way that lets them slack off. Just to be clear, this is not universal. I
also know engineers who remain focused, but it's a temptation a lot had given into and it actually
doesn't benefit their employer. If top of line productivity isn't changing, the token budget
is going to get cut. All right. So why am I giving those particular sort of negative accounts
of LLM based tools out there in the wild. I mean, since it's not to argue that LLMs don't work or
there's not use cases for them, but to argue that it's a normal technology. It has these impressive
things that can do like when we're chatting with it, right? But it's hard. And I'm saying this from
the perspective of someone who has written many books about the intersection of technology and
business. It is hard often to integrate new technologies into the professional environment.
Work is complicated, work is ambiguous, especially in the modern knowledge work era. We have many
different roles and objectives and ongoing projects all mixed together in ways that are subtle
and nuanced. It's kind of difficult to pull apart. It's hard to come into there and integrate
new tools into that environment. Either it's a simple tool that creates massive unintended
consequences like email, which is a very simple tool and a supposed to replace a voicemail
and fax machines and it completely changed our collaboration style. So either it's something like
that or it's a tool that's meant to replace major parts of work and it takes 10 years instead of
one. Personal computer usage was way slower the way that integrated into the office than someone
like Steve Jobs in 1979 would have guessed. All right. So I use these examples just to emphasize
this is going to be the case with AI as well. It's complicated. AI coding is like miraculous
that it can produce this code but it's difficult to figure out what's the best way to actually
integrate AI into the computer programming workflow. And when you talk to engineers who are not
also AI booster types, they're like, yeah, I don't know. I'm using this as better but this is worse
and I don't think we have this quite figured out yet. When you talk about AI and like law firms
or other places, there's use cases but also like often they don't work and it's complicated.
They're not miracle tools. And this is why I want to move away from this frame of thinking about AI
as like an abstraction. Like we have a brain. It's the way a lot of people think about it.
And it was a small brain and the brain is getting larger. So whenever we hear about advances in this
conception, we imagine the brain in general is getting smarter and it can do more and more
and all these jobs are disappearing and whole swaths of work are automating. Read the Bill Gates
letter. He just confidently is like, well, we know like all these jobs are going to go away.
They're not. How do we know that? What tools are going to make that go away? LLM tools haven't
done that. The MIT report from last year studied a bunch of companies, 95% of their gender
of AI pilots failed. They weren't effective. It's complicated. It's not a silver bullet.
So we have to stop thinking about AI is this abstract force. We like to, here's what I think
is going on. People want to have comments. People want to give commentary. The technology is
complicated. They don't want to give commentary on how the technology works. That's complicated.
And so they say, let's deal instead with the thought experiments, the abstract thought
experiments. It's getting more powerful. Whole industries will be automated. Now I can give my
feedback on what we should do about that. You avoid having to deal with the technical details,
but often those abstracts and thought experiments and those assumptions about what the AI is going
to do is not right. And I get it. It is hard to get into the technical weeds. Like I wrote last
week an article where I was trying to explain some of the more complicated or sensationalized
details around like the hugging face attack. And I was explaining how multi-agent swarms
actually work. These are pretty common in coding harnesses. And it's not some scary new form
of intelligence. It's really more about prompt management, keeping the context smaller to get
more out of your inference time. And then I got a lot of feedback. It was like, yes, but actually like
in this hugging face attack, it was slightly different. It wasn't a small number of prompt
loops generating many secondary prompt loops. It was actually a huge number of prompt loops generating
almost no secondary probably, which is a very unusual setup, but open AI was doing something unusual.
And so I had to go back and add an update to my essay about this on calnewport.com.
This stuff's complicated. I get it. Like that's not fun. It's more fun to give me the thought
experiment. So I can come in and be so is the sober statesman and talk about what we need to do
in these hard times. But we have to get into the technological weeds because when we do, we find out
these technologies are incredibly cool and incredibly normal. And all this is going to be way more
jagged than people say. And the more we're reality based, the more we have a say in what we like
and don't like, the more we can hold companies to account, the more we can stop like having our
our stock market be hijacked in the massive amounts of unnecessary global anxiety that's
being caused right now out of these more rationalist AI safety narratives can be tuned down.
How much like productivity and human happiness are we losing to unnecessary anxiety because some
of the big forces talking about AI spent too much time on less wrong when they were younger.
And love the narrative of super intelligence and their neo and the matrix because now they can
wear that weird like neoprene leather stuff. All right. So anyways, I don't mean to rant, but I
thought that was useful. AI is an exciting technology, but it's a normal technology and it's messy
and it's complicated. And the more we recognize this, the more we can chill out and the more we can stop
playing AI companies get away with stuff, they shun it. All right. Let's close the inbox for now.
I like to end my episodes on Monday with talking about what I am up to. Again, Jesse is not here
today. He's in some family health issues. We wish him the best. And I'm doing my best not to
bring Jesse skeleton into the studio. And if you don't know what that means, you're welcome.
You don't want to know what that means. That's a whole bunch of nonsense. I finished a book last
week. I read The New Dark Age by Nigel B. Gar, B-I-G-G-A-R. The should not be confused with James
James Marriott's upcoming book The New Dark Ages. The New Dark Age Nigel is an ethicist, I believe,
at Oxford, sort of a storied British academic type. And this book is about academia.
It's really like a critique of woke one as AOC now calls it the sort of initial woke period that
has sort of died off. Nigel was sort of caught up in this because he wrote a book and I believe
had a big grant that was looking critically at colonialism and saying there's a lot that was
bad about colonialism. There's other things that was good as complicated. It was called colonialism
and moral reckoning. But post-colonial studies had emerged as one of the massive ideological forces
in woke one and they do not take kindly to any pushback against the way they
think about colonialism. So it caused a lot of problems. So now he wrote this sort of
clinical book about, uh, woke one was it's ruining the universities. We have to push back.
The thing I admire, I always admire when I read English, especially like Oxbridge academics
is, man, they have this, this tone when they write that is somehow both conversational,
brilliant and cutting. Americans, we can't do this well. It's like, I'm being conversational
and you can almost imagine the British accent, but like their references are so air you
died. And the underlying logic, it's, you know, um, like, uh, uh, uh, a wicking style
in action. This piece is all clicked together and it's devastating in the end. So, um, I
always like reading English, you know, having an English accent already gives you a one-up
in debates, but it's not just the accent. The Oxbridge style, I find to be very effective.
I, Americans don't, I guess we would rather just have someone yell or be more passionate
and, you know, hey, this guy is on my team or whatever, but I like the Oxbridge style.
So I, I did enjoy the book though, you know, I feel like everyone acknowledges now woke
one. When you have AOC saying like, oh yeah, we got kind of crazy then you're sort of
out of, you're starting running all the people left who are still endorsing all of the
extremes of woke one. I think they're all on blue sky right now, um, trying to figure
out who to cannibalize next. All right, but it was interesting book. Um, I will leave it
there for today. Thank you, everyone for listening. You know, again, um, I always have a Monday
episode of this show, typically on the human flourishing side of the, the, of my flourishing
and technological world. Um, but I try to put up more episodes when I can, including tech
decoded episodes where I can look closer at big stories like involving AI, but I have
some other episode types that might be coming as well. So, you know, stay tuned to this
feed or channel. And of course, you can always subscribe to my newsletter at calnewport.com
to get these dispatches right into your inbox. Until next time, as always, stay deep.
PodcastTranscript AI