0:00
In the open market, AI is never going away, right?
Search this video
0:01
It's just going to get better and better.
0:03
We have to adapt to that reality.
0:04
60% of kids are just automating their homework.
0:06
And then the teachers are just, you know, using chat GBT to score it.
0:10
I realize like, oh, wait a second.
0:12
Because I drew every line on my floor plan.
0:14
I really understand how architecture works.
0:16
And now when you give me the technology, it's this exoskeleton around me that lets me amplify all of the hard-earned classical skills.
0:26
>> AI isn't going to take your job.
0:26
A human being using AI is going to take your job.
0:33
>> If school doesn't teach the slow, hard, boring, ancient fundamentals, there's going to be no intelligence to amplify.
0:40
When it comes to essay architecture, you can upload your own essay, and it will both say, "Hey, you've actually done very well at patterns 2, 4, and 7, but for pattern 17, you scored a one on that.
0:53
You're not even thinking about pattern 17."
0:54
So the whole idea is this is like a nonlinear curriculum. Well, hello everybody.
1:03
It's Jim Oasi with yet another infinite loops.
1:06
My guest today, I just saw him in real life last week and we had a fabulous discussion.
1:15
My guest is Michael Dean, an architect turned writer and Oshani fellow whose work focuses on finding the hidden structures of non-fiction, especially in his beloved essays.
1:28
He's the former editor and chief at Right of Passage, lives in New York, and writes a Substack on craft, culture, and the future of writing. Michael, welcome. >> Great to be here.
1:44
So, we had a long conversation and we're about to have it again. So, so, so I got primed.
1:52
Tell tell us uh tell our audience what what on earth attracted your uh essay architecture to Oshado Ventures to the point where we're like, hey, we got to give this guy a fellowship. >> Yeah.
2:09
So, uh the word architecture means a lot to me.
2:12
So I started in architecture school.
2:14
Um so that was like my whole background.
2:17
I was entering design competitions and uh I'm sure we'll talk about education a lot today but I totally love architecture school and think it has the kernel for what we need to do to fix education basically.
2:30
Uh but basically architecture school has a or sorry the field has a particular way to break down problems and uh there's all these works like form space and order a pattern language where they break architecture down to these primitive geometries and it says no matter what work we're looking at there's a pattern language to help make sense of this and so I have been writing
2:53
online for for 5 years now and uh through right of passage I uh started the editor program I built the curriculum And it it just kind of bugged me that no matter where I looked in the writing space, there was missing a a framework saying, "Hey, here's like basically all the tools that a writer needs to learn to become fluent in in composition." And
3:11
And I think composition is an important word.
3:16
I think, you know, when you're literate, you learn how to read and write, but composition is actually a very separate separate skill set.
3:21
It's very hard to do to shape an essay, but I I believe that it is a teachable thing and I think we can use technology to make that an incredibly hard process into something that anybody can master maybe in under a year. >> Very very cool.
3:38
Uh I was researching before uh we joined each other and yeah tell our listeners and viewers about your professor Tootszie and and how he inspired some of I I think he had a great line that is like are you an architecture or are you a critic of architecture for the New York Times? >> Yeah. Yeah. Exactly.
3:58
We we tend to romanticize our own process.
4:01
So we fall in love with the journey to come up with something.
4:06
But uh this professor of mine, he was old school, right?
4:08
He really believed that what is on the table speaks for itself.
4:12
And uh he would always just cut you off.
4:15
Said enough of that Like let's just look at what is here.
4:19
Uh he'd even go as far as when you first came into class, he would organize everyone's model in a line from best to worst.
4:27
And then he'd start with the worst one and go, "What are you doing?
4:29
We can all see this sucks."
4:29
And then he would work on it with you.
4:31
and then he could take a really poor design and actually save it.
4:35
Um, and so it's like he understood the pattern language.
4:40
He's been doing this for 50 years and through his class we started to see the pattern language ourselves.
4:44
Yeah, that's an interesting uh segue because uh as you know I I talk a lot about patterns, UR patterns which are kind of the the primitives and an AI person would refer to a model as a primitive that you can build things off of.
5:02
It's kind of the essence, the beginning that you can find patterns in and then you can build kind of really magnificent structures from those patterns.
5:12
And and so your essay architecture is a pattern language arranged across three dimensions with nine elements, 27 patterns.
5:21
We're not stick with us group.
5:25
We're not going to go through every single pattern, but but let's give our listeners and viewers a sense for the map, right?
5:32
the uh idea that you've got various dimensions and you know kind of the idea the conceptual center of an essay, right?
5:44
And and you give scoring systems which I'd like you to uh explain.
5:49
I I'm looking at the the one for uh the idea and I and I love the way it goes.
5:55
One is kind of the worst score you can get and five is is kind of the best.
6:00
and and on the idea you have your your number one is this is a Wikipedia entry which I really really love.
6:08
Um and then you go all the way to two vague illusions to your life up to five which is biographically distinct very self-aware etc.
6:20
Talk about the process and and how you do these ratings and and how they all work together. >> Right. Okay.
6:29
So I think the first thing to start is with the ratings because people are very scarred by scores, right?
6:36
We're we have that high school model.
6:37
It's like, oh, I got a 84 and you interpret scores in a certain way and uh I used to have a 1 to 10 system. It seemed traumatizing.
6:44
So let's do one through five.
6:45
Uh five is mastery, one is missing.
6:48
Three is like, oh, you kind of have the basics.
6:50
Like if you have a three out of five, that's publishable.
6:54
And uh the important thing to note is that this rubric is very flexible.
6:55
So, what I'm not saying is there's only one specific way you can score a five out of five on personal experience.
7:03
Um, so this whole pattern language is based off of Christopher Alexander's system and he sees these patterns as fundamental questions and you can see that pattern solves a million ways and each one will be different.
7:18
So even though we're applying, you know, quantitative analysis to this, each one is flexible enough so that it can uh be different every time and I think that's important.
7:28
That's what an artist is.
7:28
An artist is someone who is not defaulting to their templates, but they're always trying to say, "Hey, I already got a five out of five, but let me reinvent what that thing is."
7:38
To switch now to the idea, form, and voice, this is like the larger scaffolding that all the patterns um fall into.
7:44
Uh and then the way I see it is these this whole language, it kind of stems from the shared psychology of your readers.
7:53
So when we talk about like an idea, all writers, it starts with an idea and and readers generally have limited bandwidth.
8:00
They they can't handle infinite complexity.
8:01
And so a good idea has a center which is the thesis.
8:06
It then has all the material that orbits around that center which is the material and then you want to name that center and and that's your title, right?
8:14
Um and then when it comes to form, uh have you seen the movie Arrival?
8:18
Yeah, just watched just watched it last week.
8:23
>> Oh, so you're familiar.
8:23
So you have these like hepttopot aliens and they could effectively read essays all at once.
8:29
They can see beginning and end at the same time.
8:31
We can't do that as humans.
8:32
We have to read from beginning to end.
8:34
And because of that constraint, um there are ways that you structure all of your material so a stranger without context can get it.
8:41
Uh there is an arch to how you compose the paragraphs, your little units.
8:46
And in terms of tension, you generally want to make it so that it moves from uncertainty to certainty because people can't bear certainty.
8:52
And if you bake that in, sorry, they can't bear uncertainty.
8:57
Uh, and if you have tension, they they are invested in the question and the characters.
9:02
Um, and the last thing about voice, um, the whole idea is that the act of reading is friction, right?
9:09
It takes bandwidth to to consume sentences and paragraphs.
9:12
And so if you imbue them with sight and sound and spirit, the whole idea is you are making something like a linguistic hallucination where you can feel those words in your nervous system.
9:23
Um, so again, these are not like prescriptive solutions.
9:29
These are just the questions and as a writer, it's up to you to be creative in how you solve each one.
9:36
>> Yeah, that I I'm fascinated by Alexander's work and he he mentioned something I wanted to get your read on. Sure.
9:45
>> Uh he he he warned that pattern languages can sometimes devolve into style kits and and I know that you've taken that into consideration with essay architecture. >> Mhm.
9:56
>> You know what's the Oregon experiment uh equivalent for essay architecture where living local patterns instead of a universal template?
10:05
I imagine like might be like me writing a very personal essay about where I live.
10:10
Uh, so maybe the patterns of Greenwich, Connecticut are different than the patterns of Queens.
10:22
>> That's a great question.
10:22
And the way I'd say it is every pattern has a whole dictionary of types, right?
10:27
So if you go into repetition, for example, there are dozens of Latin named concepts like an epizus or an anafa.
10:35
And uh, so you have infinite types.
10:38
And the the risk is the development of a global style where everyone is using the same forms of repetition.
10:46
Everyone's using just the hero's journey, right?
10:48
We take these proven global templates we know work and then it just becomes boring because everyone's using them.
10:53
And uh the equivalent of the orgon experiment is to understand how can you be idiosyncratic in how you approach each pattern.
11:00
Um that could be trying to think with location. Yeah.
11:06
I mean like what how would a a writer from the south who's who's meshed in that history, right?
11:12
Different from a writer in New York City.
11:14
And I think that that's like the critical nuance in how you can have this shared language, but we each still use it very differently based on where we come from and who we're inspired by. >> Yeah.
11:28
I'm reading uh a biography of Mark Twain right now and it uh it's really interesting although the author doesn't really seem to like Mark Twain and and I and I I love Mark Twain.
11:41
>> That's an interesting concept for a biography. >> Yeah.
11:44
Uh well, lots of interesting things I didn't know about him.
11:47
So, I am enjoying the book.
11:49
Although it does annoy me when the author clutches his pearls and and talks repeatedly about Twain's failures to become, you know, a business tycoon and and you know, some of his idiosyncrasies.
12:02
I I think that's what makes him kind of an interesting character.
12:07
>> But one of the things that what you just said brought up is Twain changed pretty dramatically over the course of his literary career.
12:16
um he became as many very famous authors said, you know, the true first American style novelist, right?
12:27
Uh but then as he as he let more of his um thinking about issues outside of humor in his his writing changed pretty dramatically as his experiences changed.
12:41
Um, and I I one of the things that I like about your essay architecture is you're all set.
12:47
If somebody starts out if if we put essays that I maybe I wrote when I was, I don't know, 18 uh into essay architecture and then compared them with now, I would expect that they're going to hit a lot of different scores on your uh your various factors.
13:10
uh how do you how do you or what do you think about the journey an author takes?
13:17
Because when we were together, I thought like one of the cool things uh that essay architecture could be incredibly useful for would be like substract writers who, >> you know, just want to maybe experiment with a new type of style, you know, in their in their personal journals.
13:37
they're finding themselves, wow, I I'm writing completely differently now.
13:39
Uh how how how would essay architecture help them along? >> Yeah.
13:46
So, well, there there's two parts to this, I think.
13:51
And the first is it's almost like when Twain's audience changed, his general style warped with it.
13:56
Uh and basically, your audience is a huge factor in in changing how you might use these patterns differently.
14:03
And when you're writing in your journal, you're much more willing to push the boundaries and experiment than if you're writing to say your whole professional network.
14:10
Um, and so when it comes to essay architecture, it it doesn't really have the audience in mind, which is interesting, right?
14:22
Because it it kind of it tolerates all forms of style.
14:24
It's looking just at the primitives of composition.
14:26
And so this is software that I'm building right now.
14:30
So you can upload your own essay and it will both say, "Hey, you've actually done very well at patterns two, four, and seven, but for pattern 17, you scored a one on that.
14:39
You're not even thinking about pattern 17."
14:41
Um, and you know, usually writer will be like, "Well, you know, I wasn't going for 17. That's not my style."
14:47
But then my tool can recommend here's a classic essay where it takes 17 and it takes the patterns you're already great at and it confuses them into a single thing.
14:57
Um, so the whole idea is this is like a nonlinear curriculum.
15:03
I'm not telling everyone to go read my textbook cover to cover.
15:05
The whole thing is you upload your drafts and it'll say, "Hey, here are your three weakest scoring patterns.
15:11
go study these and we're going to teach you how to weave that into what you're already good at.
15:17
And uh I think this is a little bit different than the the whole model of of these sickopantic chat bots today because because right now these things are designed to conform to you and it tries to understand your taste and make you feel like oh yeah you're doing a great job with what you're doing.
15:32
But this system has like a stubborn unchanging quality standard and no matter what it's going to always kind of uh yeah champion each of these 27 patterns as having equal importance, right?
15:46
because some people say, "I only care about these 15 patterns."
15:48
And this is a great way to upscale in the things that you haven't thought about.
15:54
>> And one of the things that I find interesting is how does how would it deal with like a James Joyce?
16:00
We we talked about this when we were together.
16:05
Um, you know, when Joyce came on the scene, his type of writing was really really new and people didn't know how to read it.
16:14
literally they didn't know how to read it.
16:17
And so if you go back and like read some of the um original reviews of his work, they weren't they weren't glowing, >> right?
16:29
And so is there a mechanism uh for you to incorporate that or let's say let's use Joyce let let's say Joyce or a Joyceike character today wrote an essay in a style that was really really different that uh that didn't fit any of the 27 patterns that that you've got in the software.
16:54
How would you how would you approach that? Yeah.
16:58
So my understanding if you look at Joyce or even Picasso or the Beatles or any artist in their earliest phases they do master the the fundamentals before they break into very experimental territory.
17:13
Alex Kbusier is an architect in a similar state.
17:15
He started with very machine-like boxes and he became known for these very expressive curved buildings.
17:21
And um I I think if you look at portrait of an artist at a young as a young man, I believe that is much more conventionally structured than something like Finnegan's Wake.
17:32
And so it's almost like once you master composition, you kind of get bored.
17:35
It's like I want to do different things.
17:37
And if you look at something like Finnegan's Wake, uh for anyone who hasn't read an excerpt, it's almost like uh to me I have to read it out loud and I can't force myself to logically understand it.
17:50
And it's almost like the sequence of dream images that make sense at a layer that I can't quite understand, but it's fascinating.
17:55
And uh you could say that Joyce is actually far or put this way, he's he's maybe scoring a five out of five on imagery or five out of five on rhyme, but he's doing it in a way that no one has ever done before.
18:12
And I suppose you could if you had a big enough data set, you could attempt to categorize all the different ways someone gets a perfect score in imagery.
18:19
And then you can be like, "Oh, Joyce is doing something here no one has ever done."
18:22
Um, I'm not sure if if that would be useful because I think the number of people who are doing that level of pioneering, I don't know if they need numbers to to quantify it.
18:31
I really think this system is um it is really focused on composition.
18:37
And I think quality is this larger thing.
18:44
And there's multiple reasons why something is high quality.
18:48
Composition is one of them.
18:48
And I feel like once you master it, you then have license to break it and have a lowcoring essay because because you know you you've earned that and there can be other reasons why your piece um shines.
19:00
And so this kind of ties into the fact that I'm using my tool to host an essay competition coming up uh September 15th and uh we are using the tool to kind of score everybody on composition.
19:11
But then we also have human readers and some and and some batch of readers they're they don't care about composition.
19:17
They're going to say, "Is this essay timeless?
19:20
Is there a singular aura? Is it strange?" Right?
19:22
So, these are some of the more intangible degrees of what makes quality.
19:25
And then we want a guest judge who's a wild card, who doesn't even tell us what their standards are.
19:32
They're just ranking the essays in order of of how it hits them, say, emotionally.
19:36
Um, and so yeah, hopefully that articulates that that composition is is like kind of just one part of the quality puzzle on what makes something great.
19:49
>> Yeah, I as you were talking I was thinking about uh, you know, get the basics first.
19:53
And I think it was Charles Mingus whose son was studying to be an artist and and wanted to go right to abstract painting and and Mingus said maybe even quoting an artist uh something like hey man learn how to paint an apple that I can understand that it's an apple.
20:12
After you've mastered that then go ahead and get as abstract as you want.
20:16
Um, how would you coming back to Christopher Alexander, he he has this concept of quality without a name >> and like how would you deal with a a quan uh to make it an acronym uh in your in your architecture of your software in your book? >> Yeah.
20:38
So, so his system it covers these 253.
20:41
So, his is 10 times more complex than mine.
20:47
253 patterns that you can use for uh from everything from the design of towns then to buildings and to details and it's all this hyperl maze.
20:55
So if you're designing a garden, well then it points you to the 10 other patterns uh you need to design a garden.
21:02
And so the design process is this, you know, it's kind of like a sudoku where you're going in and out of different problems and you fix it in one area and then it changes somewhere else.
21:12
And you can be logical in how you think at these very little isolated parts of your design problem, but then all together they come together to create this thing that the mind cannot behold. Right?
21:25
It's it's almost like sacred where you've done your design work, you've pushed the boundaries of how you can think about it and then there's this emergent thing.
21:34
And he he kind of gets a point where we can't really name it, we can't describe it, we can't quantify it.
21:38
uh it's almost like you can't even aim for it or desire it.
21:42
It just emerges, I think, when you start to uh solve all of these design constraints earnestly.
21:51
And uh now we're kind of talking about a level of writer who, you know, maybe that's their occupation.
21:58
They're an essaist or they have a Substack or whatnot.
22:04
I know from having looked at your software and and chatting with you that you can serve that artist very well, >> but I personally think one of the areas as we talked about when you were here where where this might really be something that you could not only what's the old joke about missionaries that they went to do good and did well.
22:26
Uh where where you could do good and well at the same time.
22:32
This wasn't my concept in our conversation.
22:34
This was Jimmy Sony's, my editor-inchief at Infinite Books.
22:39
But, you know, it's kind of like what's the one time in life that everybody is required to write an essay, >> right?
22:47
>> When they're going to college, right?
22:49
And college doesn't, at least in in my review of the discussion, let's call it a discussion uh that academics, colleges, etc.
22:59
are having with I'm doing air quotes here for our listeners uh that uh that kind of is fearful of AI >> uh in a way like don't let it write for you blah blah blah blah but your thing doesn't write for them essay architecture does not write for them but it's a really great learning tool and you know right now college admissions It's not a level playing field.
23:29
As as Jimmy brought up at our lunch, he he knows people in Manhattan who are literally paying consultants $1,000 an hour to help their their child write an essay that will get them into, you know, the IVs or or whatever.
23:48
the IVs or or whatever. and and I see your tool as something that could level that playing field if you could get buy in at the college level and and not call it AI but as I mentioned when we were chatting call it what I think it really is which is intelligence amplification uh you would have a way it's not an AI detector but it's also not a ghost writer it isn't it isn't writing anything for the student and you know it
24:24
would give information to colleges that they simply don't have right now and that is how did this student progress let's say I'm uh you know u any university in Vermont or whatever state and and I say you know what I'm going to
24:42
embrace this I'm going to use Michael's essay architecture and you have every student use the same template to to write their essay and you get to see their progression because again let's emphasize an underlying essay architecture is not a ghost writer. It
25:00
It is something that helps the student get better at the craft.
25:06
And you you would get a really high signal score by looking at the delta between what they did originally and then their final essay that they submit uh with the scores from essay architecture.
25:23
Uh I I I think it's like a really great idea. What what do you think?
25:30
Yeah, there's there's so much here and I looked it up and I think I read that 25% of the incoming students going to Harvard are using professional admissions consultants.
25:42
Uh, and they're spending up to $3,000 for this process.
25:44
Um, and you could argue that a tool like this, like what I'm building, if I'm trying to build a world-class editor that anyone can access for under $10 per essay, right? Right?
25:57
So, a tool like this could suddenly give everybody access to that essay expert to articulate their thoughts at their highest potency.
26:03
And uh there's a funny what's the word?
26:09
There's a discrepancy where even though I have a textbook and it articulates all 27 patterns and it describes what makes it a five, if you upload your draft and uh and you say, "Hey, make it a five out of five."
26:21
It still can't really write an essay on its own that exceeds a 3. 5. Right.
26:27
So there's some uh difference in capabilities where even though we can analyze precisely and accurately, we can't yet have a chatbot generate from those same instructions.
26:36
Um and so yeah, like you're saying, we could manage uh we could draft student progress over time and it's not writing for them, but it's giving them all the right questions and ideas for them to rewrite the draft at a higher form.
26:52
And so if you have a student who now is getting past a four, it's very hard to get past a four.
26:56
Even for me, I struggle to get past a four as creating the system.
27:00
It takes a lot of deep thinking, synthesis, and and time, right?
27:05
You just really need to go through multiple drafts on your idea.
27:09
So, uh the hope is that this tool can develop a quality standard that is a proxy for has this student actually done the thinking, right?
27:18
And have they actually worked through all the different challenges of composition?
27:23
Yeah, I I think it ensures integrity and encourages improvement, right?
27:26
And rather than, you know, hide their heads in the sand, I think that colleges, universities certainly have to come up with a game plan, right?
27:38
When when I was a kid, uh I nerd alert, uh I I really wanted a calculator which were just coming out.
27:49
uh you know the usable ones back in 1970 when I was 10 and uh I got one for my birthday and I was really excited and I brought it to math class and uh my my professor had seen them but a lot of the other kids hadn't seen calculators yet and like everybody was like oh my god this is so cool then what did the school do about as they became more popular right as they went down in price.
28:19
They ban calculators and I'm kind of like I've always been had a bit of a rebel streak in me.
28:24
I'm like what the Like you you want us to go back to abacuses and slide rules?
28:33
I I see that in today's debate about this incredible tool of intelligence amplification AI and and it's like it's not going away colleges and universities and to instantly label it as ooh that's bad that that harms human creativity. I I completely disagree.
28:51
If used in the context like we're discussing here, >> right, it it allows you through technology to level that playing field because it scales, right?
29:04
If if universities embraced SA architecture and it became the platform, then you can have comparisons from universities, then you can have, oh, what can we learn and then iterate?
29:19
it would become kind of a a selfdefining great loop where you're going to learn a lot from the students that are using it because you're going to have a much larger sample size.
29:30
Also, the quality of the writer, right?
29:33
If if you're only dealing with people who already are professional Substack writers and that's their job, I would assume that their level of composition and all of that is going to be higher >> than your average 17 or 18year-old who is facing this daunting task.
29:51
And mom and dad can't afford that $1,000 an hour consultant.
29:57
And to me, that just doesn't seem fair.
29:59
And right like I I'm I'm a full believer in free markets and the way they shake out.
30:10
But if you do have a technology like yours where every student could have access to it, presumably every student is going to learn from it.
30:19
You're not doing ghost writing.
30:21
You're not doing any of that. Wow.
30:23
I mean, that seems an auditable thing that universities could be like, "Wow."
30:30
And it would also allow them to not just hide their heads in the sand, right?
30:35
And and like this idea that they're going to ban every new technology just seems beyond idiotic to me.
30:43
It's like, you know, who was the most distraught when photography became popular?
30:49
Portrait painters, right?
30:49
I just read Morgan Hel, my friend, who I call the sage of Seattle, just wrote uh a piece of some of his thoughts.
30:58
And I didn't know this, but um in the late 1800s, as cars were first coming on the scene, Washington banned the Washington DC because they would impact the employment of horses. >> No way. >> Yeah. Yes. Yes way.
31:14
And and you know, you can see it in every innovation.
31:18
Like ice boxes are called ice boxes because they used to have ice at the top that kept you cool.
31:26
And it was a huge profession here in New England, ice harvesting.
31:31
People would go into the river and carve out big blocks of ice, haul them to people's homes, and deliver the ice.
31:39
Well, then refrigerators came along and guess what?
31:41
Wiped out the ice harvesting industry.
31:45
But all of those people probably found much more enjoyable jobs for the very least.
31:50
Why do you think that there is this innate hostility toward innovation technology that in my view and by the way I'm not naive.
32:04
I know it can be used in a bad way as well, right?
32:08
But like >> it seems to me that the immediate response is always hostile.
32:13
No, that's going to break us as a species.
32:16
And we can go all the way back to Socrates who said writing is a bad idea.
32:21
This seems part of our human OS. What are your thoughts? Yeah.
32:26
So, I think it's good to break it into two buckets where you have the open market, right, where the best ideas win, and then you have schools which are supposed to kind of like train the citizens to have all the skills needed to live a fulfilled life.
32:40
And uh in the open market, AI is never going away, right?
32:44
It's just going to get better and better.
32:46
We have to adapt to that reality.
32:47
I think what schools are seeing now is that some of that technology is seeping into our schools and now 60% of kids are just automating their homework and then the teachers are just, you know, using chatbt to score it and it's exposing the I think the shallowess of accreditation at all the gates of our school system.
33:04
So the first flash reaction is like we'll just ban it, get the technology out of the schools.
33:09
And that's a temporary fix.
33:13
It's it's not really going to stop it.
33:15
Um so so I think the challenge here is we need to think more nuanced about how we think about um like what is the point of school, right?
33:25
Um under I think at the basic level it should be trying to motivate students to learn.
33:30
I I think it's failing at that in a lot of ways. But um I have a story.
33:33
So, I went to University of Miami.
33:36
I transferred into architecture school and I found out, hey, by the way, we ban technology for the first three years.
33:42
I was like, what are you guys talking about?
33:44
Like, like, you're going to make me handdraw all of the stuff and it's going to take 20 times longer because I can't use AutoCAD or SketchUp. And like, yes, exactly.
33:54
Like, like you're going to learn by drawing those lines.
33:55
Uh, and so this this map behind me is something that took me 120 hours and it's a a depiction of Manhattan drawn with only dots, right?
34:06
Effectively, this is a job for a printer.
34:08
A printer should do this. Why am I doing this?
34:09
So, I got pissed and I transferred after one semester and I went to a modern technology school.
34:14
And when I was there, I realized like, oh, wait a second.
34:18
Because I drew every line on my floor plan, I really understand how architecture works.
34:22
And now when you give me the technology, it's this exoskeleton around me that lets me amplify all of the hard-earned classical skills.
34:32
Um, I think the problem is if school doesn't teach the slow, hard, boring, ancient fundamentals, uh, there's going to be no intelligence to amplify.
34:43
And I think there's a way to fix this.
34:45
And it and I think it just comes from like sports, like how you train athletes.
34:48
Uh, and so I have one more story where in ninth grade I did not make the baseball team. I was a water boy. I was devastated.
34:54
I said, "Let me get a hitting coach." So I got this guy.
34:58
He's like affiliated with the Yankee minor leagues.
35:00
Okay, I'm going to I'm going to learn how to hit.
35:02
And like on the first day, he says, "Okay, the first drill is you're going to put the bat behind your back."
35:07
I'm like, "What are you talking about?
35:08
Why would I put the bat back there?"
35:09
He's like, "I'm gonna pitch to you and your only job is to watch the ball and and pivot your back leg, right?" Right?
35:15
And he's like, "We need to isolate that one skill."
35:17
And I learned from him that there's these like 10 different parts of a baseball swing that you have to isolate part at a time.
35:25
And like none of them involve or a lot of them involve not using the bat.
35:27
So you're taking away the main tool, right?
35:31
But then when you go to actually do it all together in in context in the game, all of these little micro movements are just fluent to you, right?
35:40
And so how this ties to education is we need to figure out how these two different sandboxes can coexist in one sandbox.
35:48
We're isolating mini games where you're not you can't use all the technology.
35:53
Um I've designed exercises a whole curriculum of exercises and like one of them would be okay you're going to write a paragraph and you can only use one syllable words >> and then do it again.
36:04
you can only use two syllable words and and like again these aren't um trying to produce great writing but they're teaching your mind to think in a certain way and then we need another sandbox where we say now we're going to give you all the best tools in the world and you can take your fundamentals and leverage them.
36:19
So I think yeah I just think education could think along those two lines like mini games for skill training and then understanding the cutting edge of of what's possible.
36:31
>> Wow that is a really great answer.
36:31
Uh because um as as I was listening to you, I remember when I wrote uh the book What Works on Wall Street, um the data provider uh gave me a an automated tool that would allow me to say, "Hey, how did the 50 stocks with the lowest PE ratios in 1995 do in the ensuing year?"
36:57
And as I was doing it, I was realizing that really doesn't look right.
37:03
And so I ended up having to do it all by hand.
37:06
The tool did not work as advertised.
37:11
>> And the doing it by hand, I oh man, I was so grumpy.
37:14
It was just like, I can't believe I'm actually doing this.
37:16
Uh but man, it really gave me a foundation for understanding the way data works, the way errors creep in for for years.
37:30
Uh, I was their best error finder um in the compat database because literally I had to like you doing the map of Manhattan, I had to by hand walk the data for every year between 1952 and I was writing this in the early 90s.
37:51
Uh, and the things that I found on the way were really really cool and ideas that I had never had. right?
38:02
Like um boy this this data is really dirty.
38:06
In other words, there are a lot of errors.
38:10
It's error uh written and if I hadn't had to do that by hand, I would have never known that.
38:16
So that's a really important concept.
38:20
H how how do you think universities could go about the two sandboxes?
38:26
I like that metaphor as well.
38:30
Yeah, I think the challenge is that these little granular skills that you need to do brute repetition on to an untrained mind it feels like torture.
38:41
It felt like torture making this, you know.
38:43
Um but and and similarly, if you were to talk to a new writer and I tell them, hey, like I actually go and I I read essay and I pause in a paragraph and I score it one through five on 81 dimensions.
38:55
That could sound like torture to somebody, but for me that brings me so much joy, weirdly enough.
38:59
Like that's actually where I'm in the pocket uh at the edge of my thinking, right?
39:03
And I think it's because I'm an essay writer myself.
39:08
I I'm trying to, you know, grow an audience on Substack.
39:10
I understand the value of practice and of writing and there's a larger story that motivates me and it makes it worth kind of going through these little drills.
39:20
Um, so I think two things.
39:22
is I think school needs to figure out how you actually motivate people.
39:28
This is another discussion we could have because I think our teacher school nailed this but you need to give students autonomy.
39:33
There needs to be social stakes and they need access to resources where they can actually see improvement quick enough.
39:40
So they need to get like the motivational OS right.
39:42
Um and I think the second thing is you need to gify it, right?
39:47
Like if you just say, "Hey, you're just going to go in and you're going to um I don't know, root like just handwrite the same thing a thousand times on a blackboard."
39:56
Like that that I mean that is more like a torture than a game.
40:00
But if you make it a competition where there's a leaderboard and and some kids win, I think uh I mean synthesis does that right like you can use the structure of games to get them to inherit the right skills.
40:10
So I think it's motivations and and games.
40:12
I think also part of it is you know this this every there's a great quote I think it was Tony uh Dlo who said uh everybody wants progress but nobody wants to change >> and like I think that fits the what's going on right now at the educational level right like on the one hand the I think it's hard to argue that say K through of is prosper. Right. >> Right.
40:46
>> And and the the negative outcomes.
40:46
I was just reading a piece about a kid who went all the way through college and he's functionally illiterate.
40:55
He's literally functionally illiterate.
41:01
>> So, um the institutions to a large degree are failing their students.
41:04
And that doesn't mean I hate the institutions.
41:09
It means, hey, you got to reform.
41:12
You've got to try new things.
41:15
And that's why I was so excited about this idea that Jimmy Sony had uh because your software is kind of like the perfect place to do it because you they they can overcome their fears that an AI is just writing everything because that is not what happens with your software.
41:34
Uh but it also levels the playing field so that the rich kid from Greenwich uh is operating under the same kid uh same uh circumstances as the kid from wherever the Bronx.
41:51
>> And to me that's very exciting.
41:51
Another thing as I mean I as you can tell I'm very enthusiastic about about essay architecture.
42:00
Another thing that would be really cool is your idea of a contest, but also what what about a tool where you simply ingest through RSS feeds every Substack essay, run them through your system that allows you to discover, wow, there is a great essay writer who nobody's ever heard of.
42:23
and then kind of like a hacker news uh spot where you elevate and amplify this writer.
42:32
Hey, you really got to check it out.
42:34
Like what a great way to find great new talent.
42:37
What do you think about that use case?
42:41
>> Yeah, that is probably what I'm >> It's hard to say.
42:45
I love the vision of of education and helping people discover the process of not just learning but learning through writing.
42:54
I really think writing is the the wedge from which you can become a self-driven learner.
42:57
But the other uh you know implication of having a quality standard that's baked into AI is that suddenly we now have the ability to discover uh where a great writing is.
43:11
And the internet right now can't do this very well.
43:14
um the only data we've had since you know 2005 or so we have engagement based metrics and all of our algorithms are based on the volume of people who are engaging with something.
43:27
Um and so I think it amplifies a certain kind of of content and it's not necessarily the the best written or the most visionary stuff.
43:36
Um and I think AI is going to make this even harder both because generation and engagement hacking can be done at scale and uh we will need effectively a quality rhythm to sorry a quality algorithm to find the essays that are that are worth reading.
43:53
Um yeah and uh I think there's value to doing this like in real time.
43:59
I would love to have a newsletter where every week it's like, "Hey, we've we've read 100,000 essays this week and uh but you've got to read these these three, right?"
44:10
So, that's kind of one angle.
44:11
And then I think there's something to this like big uh annual essay contest.
44:17
Uh there's there's a long history of of essay competitions and they still exist in modern time, but I think they're much narrower in their focus.
44:25
Like you have literary magazines and they have the best American essays anthologies.
44:28
Um, and I'm just asking the question like how would Mr.
44:32
Beast uh run an essay uh prize today, right?
44:35
Because this whole thing is he gives away huge sums of money and he creates these like attention games and could you use that same mechanic to rebrand the essay almost like a Trojan horse, right?
44:45
It's this it's this thing that everyone has a certain uh perception of and can it actually be fun and gamified?
44:52
And I think the difference with this is if you are someone who has this tool and you're in this competition, you're most likely not going to win, right?
45:02
If there's thousands of entrance and one winner, but every loser now gets this essay on an important idea to the edge of their ability and they still have full ownership of that idea and they can publish it themselves.
45:14
So a lot of these like literary compositions, they really try and like own the submissions.
45:18
Uh, and I think in the spirit of the internet, you want to let people, you want to lean into self-publishing, and I think you could really build the community uh around this kind of like prize quality standard.
45:31
Um, there's even like a a textbook that comes with it.
45:36
So, that's kind of my question, like what does it look like to build a modern kind of like writing institution?
45:40
Um, both to teach writing and to celebrate the best works that are coming out of it. >> Yeah.
45:47
Um, as you know, I've written all of my adult life.
45:50
I have, you know, more than 45 journals sitting around me here.
45:56
And, uh, writing is I I could not agree with you more.
46:01
Writing is maybe the most important aspect of learning kind of what you think, right?
46:13
We we we we all have a bunch of ideas jumbled up in our heads and oftent times people will say ah I think that's a great idea and of course that's happened to me many many times.
46:24
I've had lots of crap ideas but the way I discovered they were crap ideas was I tried to write about them and as I was writing I was like you know the the gif where the guy is like he's starting to write and then he goes I got nothing. >> Yeah.
46:41
And unless you take that transfer from your head to pen and paper or your keyboard, uh I happen to believe pen and paper is even better.
46:51
And apparently uh there are some studies that show that that is true.
46:57
We won't devolve into that right now, but >> I believe it.
47:00
Yeah, you're absolutely right about just the the value of writing to improve your thinking, to improve your communication, to improve like there's just so many aspects of it where it just really sits at the center of everything.
47:17
And so I am certainly not one to advocate just having, you know, the AI do everything for you.
47:22
I think that's going to lead to the tsunami of slop.
47:28
Uh, but that makes me super interested in apps and uh and uh structures like yours because I I think what's going to be big and something we'll be building at Oshana Adventures is we're going to need curation apps.
47:45
Here are people who are just like, you know, I don't have the time to read a 100,000 essays, but I I would like to read really good ones. Right. >> Right.
48:00
Show me the best one and then give me summaries of the other ones that I might want to read after. >> Exactly.
48:06
and and and I think that the the the more you zero in on ways to do that, in your case, this is a way to rise the best essays to the top.
48:18
Um, but we also are big believers in what I call what is generally called the senotaur model of how to use AI.
48:28
And that means man, woman plus machine, not machine alone, not person alone.
48:36
It's it's like the the uh quip that AI isn't going to take your job.
48:44
A human being using AI is going to take your job.
48:49
>> And so I think that listen, I I've been lucky enough to see tons of innovations in my life.
48:55
Uh, but I think I think this is right up there with Gutenberg.
48:59
I think that the especially the intelligence amplification reframe of it.
49:07
I I agree with you by the way that you got to get the basics and and schools should be uh trying to uh focus on that as well.
49:17
But I I definitely see a cognitive chasm opening opening up between people who are using these intelligence amplification tools and people who are not.
49:30
And and so I like your idea very much that at the student level make it fun, right?
49:37
Like if you look at almost all of human history basically and by the way not just humans most of them learn through play right and you know we're highly mimemetic creatures but that can lead to downsides but it's really good for learning and but if you if you make it fun you're going to get I think much better results.
50:05
I think one of the big mistakes we made was borrowing the Prussian educational system.
50:13
>> It was fine pretty much when they did that, but they lacked it, right? It didn't evolve.
50:19
And you know some there are some theories that really the the reason we have public schools the way we have them is because during the uh guilded age right when everything was becoming industrialized the industrialists said to the uh politicians yeah we we want people who can sit calmly in a room for 8 hours doing what we tell them to do.
50:44
And the fact that we are still kind of doing that to our kids through 12 kids, it's to me that's insane.
50:55
It's absolutely insane because the the the the time of life in childhood when you should be full of wonder, right, and full of the desire to explore and learn and play is being almost beaten out in the current system.
51:13
It's it's it's like like I saw somebody who said uh what group of uh in society has even fewer rights than prisoners and the answer was school children.
51:30
>> Jeez, >> that's just awful and it's so normal.
51:33
It's just like the thing you have to do.
51:35
And and I think it's it's this whole homework apocalypse is hopefully a blessing because it's making us realize that we've been stuck in this old system for for 200 years.
51:45
Uh there's an essay by Paul Graham called uh towards a golden age of the essay.
51:52
I think he wrote it in 2004. >> I read it. It's good. >> Yeah.
51:57
and he noted that the five paragraph essay came from medieval law schools.
52:03
So in the middle times it wasn't just uh uh seminaries it was mostly law schools.
52:08
And so that's why the five paragraph essay reads like a legal defense.
52:12
In this essay, I will argue this, right?
52:14
And so to tie this to your framework of the thinker and the prover, our whole method of essay writing, which should be the tool to unlock curiosity, is now basically lawyer training, right, from sixth grade on.
52:29
It's like it's completely backwards.
52:32
And uh and so I think this came to the US in like the 1830s, 1840s.
52:38
first compulsory school was Massachusetts in 1850.
52:41
And even Emerson was like, "Guys, what are we doing?"
52:45
Like out of like 10 years out of the gate, you have Emerson saying, "This is ridiculous. What are we doing?"
52:50
And he says, "Every student needs to be inspired to uh a lifelong process of learning and they need private tutors." Right?
52:58
It's like he had the vision, but it was ultimately so impractical, right?
53:03
In 1850, they were just desperate to get anybody in the room to look after these um 30 kids, right?
53:09
And so it's much easier just to teach a formulaic five paragraph legal essay than to like sit one-on-one with each kid and be like, "Okay, here's Montaigne.
53:17
Here's how you actually question authority."
53:18
And there's been no solution on how do you actually equip each student with what they need to learn to write.
53:26
And arguably AI is that moment.
53:28
We finally have the technology to realize Emerson's vision of education.
53:33
Um, and it's like there's this very urgent moment where like you're saying, uh, there could be a radical gap in those who know how to think and use AI and those who don't know how to do either.
53:44
Uh, but within this decade, we have the tools to suddenly teach essay writing as basically the the wedge for someone to start actually caring about learning.
53:59
And it's just amazing to me how much I love essays now.
54:01
I completely hated writing in high school because it was all these legal defenses and it's the provers, not the thinker.
54:06
And um it wasn't until much later when I'm writing to make sense of my own ideas uh that I I really love the process. >> Yeah.
54:16
Um, and you know, just to get off the soap box, but with one last blast, like, you know, clearly we have a system that kills as opposed to encourages that love of learning, that exploration.
54:34
And you know, Robert Anton Wilson was talking about his time in in Catholic schools in the 1930s and 40s and he said, you know, it seemed to me that their only real objective was to install the only one correct answer machine in my brain.
54:52
>> And that doesn't lead to people being really excited about self-learning and self-direction.
54:58
But certainly in my own case, like almost everything I've learned, it was not in the schoolups, right?
55:06
It was because I was curious and and went down that path.
55:09
I another idea though is like nobody like because maybe because of that five paragraph legal.
55:17
I mean, who wants wants to read that? I mean, nobody.
55:23
Um, let's talk a little bit about some some great essays.
55:26
You you know, you've done a list with showing the the greatest essays you scored and and I'm going to lead in with the question of the top 10.
55:34
Why does David Foster Wallace take three of those top 10 slots?
55:42
>> Yeah, he has uh he has three of the top 10.
55:44
He also has some essays that are in the like seven out of 10 and not high nines. So, he has variability.
55:49
Uh, and I also find that he has two versions of a lot of his essays.
55:54
He has one in his essay book where he basically didn't have an editor.
55:59
He could throw in all the tangents he wants and then the public version that had to go to a magazine.
56:03
And I've tried scoring both and the ones that have editors working with him do score higher.
56:08
Uh, because it's more like we're trying to compress the insight versus see every nuance of his mind.
56:14
But I just get the sense that he um he just had this intuitive grasp of that pattern language.
56:21
Like his essays just pushed the boundary on every single dimension.
56:27
dimension. Like if you read that book consider the lobster, it's just like a perfect example of what a microcosm is where, you know, he could have written, you know, a treatise on animal ethics and would have been very
56:39
boring and comprehensive, but instead he brings you to him at a specific lobster festival with his parents and girlfriends and the chaos around him and it's like fun and psychedelic and weird and then he gets into all of these like philosophical dimensions. So, it seems
56:51
So, it seems like every essay in that book, because he has a whole book called Consider the Lobster, every single one is like a masterclass in how you can pull off a microcosm, which is another reason I'm excited about SA Architecture is you provide a path for people to understand how they with their passions and their particular interests can can do the same kind of thing, right?
57:16
they they can learn through continual repetition of of interaction with SA architecture. Oh wow. Wow.
57:25
I'm getting a super low score over here, but I love that.
57:28
Maybe I maybe I need to put that back in.
57:30
And uh as someone who's written four and is currently writing a fifth book, um I they would be just pure if they didn't have editors.
57:42
Um, I was very lucky uh because my wife graduated Sumakum Laad in journalism and and had the reddest of red pens.
57:52
And I I I was first really angry when I would write something and it would come back uh you know with just a sea of red all over it.
58:03
But as I as I got over myself and I realized, wow, she's she's right.
58:11
And and so the I I just on the fly that's another great use case for how AI can be helpful, right?
58:24
Not everybody's going to be lucky like me and and be married to someone who graduated Sumakumlad in journalism.
58:27
uh the editorial function of AI.
58:31
Again, let's underline not the ghostwriting aspect of AI, but the editorial aspect has great promise.
58:41
As a matter of fact, at Infinite Books, we have a thing we call the portal, which we put manuscripts in and it ranks it across many different categories >> and then our human editors take look at it and fine-tune it, right?
58:55
like, "Ooh, got this one wrong."
58:58
You know, Dylan thinks that this is a masterpiece.
59:00
Let's train it on why that is.
59:03
So, another important reason from my point of view why you want the senator model, the human plus the um the AI, but you know, then then the quant in me uh is basically one of the reasons I became a quant was because precision, right?
59:25
Quantitative models work not because there are these incredibly complex things, but because of they are uniformly applied every single time.
59:36
>> And when that is the case, guess what?
59:39
Their batting average on balance, in other words, on all of the forecasts or decisions they made tends to be higher than the batting average of a human. Right?
59:51
Because and by the way this is true not just for investing there's a uh a great book written back in the 70s uh the limits of scientific reasoning I think is the name where where he looks at literally I think the author's name is David Foust but literally it's everywhere he looked at you know uh what I I would call quantitative models others call algorithms or whatever and a huge variety of tasks Right.
1:00:20
Horse handicappers handicap handicapping a horse race, admissions councils at universities deciding whether somebody gets in or not.
1:00:30
Uh doctors making diagnosis, etc.
1:00:35
They were all outperformed.
1:00:35
They thought that these quantitative models would be a floor in which the human uh would would soar above.
1:00:43
They proved to be a ceiling that the humans very rarely touched.
1:00:50
But but the thing that I found in there was there was a persistent group of subset of humans >> exactly >> who were like amazing.
1:01:03
And then so what I did when I was developing all of the investment algorithms was incorporate those incredible human insights as best I could in in the um in the model.
1:01:11
could in in the um in the model. So I I I wonder you've said you know quality is statistical and and a lot and a lot of people like we have this math anxiety that the minute you bring in statistics or whatever it's like
1:01:31
>> no no I'm like no it's not and we get violent reactions like if you say things like that but you know it's it's objectively fairly true right you're getting I think correct me if I'm on 99% precision on the various essays and and I think the closest human is at around 73%. That is not to say we should not use
1:01:56
That is not to say we should not use humans.
1:01:58
Let me be very clear and underline that.
1:02:01
>> I love your wild card idea by the way for your judge.
1:02:04
>> And I think it's this combination that will lead to a lot of the things we've been discussing.
1:02:09
Um, can like what what are your thoughts?
1:02:14
Is this going to be a slow slog?
1:02:17
Is everyone just going to keep burying their head in the sand and saying, "No, no, no, no, no."
1:02:22
And let's go back to calculators, right?
1:02:25
So, there was this period, calculators were new.
1:02:27
They didn't get banned yet. Then they got banned.
1:02:31
Then we had this this stub period where I was trying to figure out how to use a slide rule.
1:02:37
And then of course like literally now they're in every classroom.
1:02:43
They're required in in many math classes.
1:02:46
Do you see a similar adoption curve here?
1:02:52
>> I think this decade historically could be remembered as one of extreme like existential anxiety for writers.
1:02:57
I think it's very natural that people are pushing back against AI.
1:03:00
Um some people will just always have a blanket no AI stance and I think that could change over time.
1:03:07
Uh my hope is that by saying hey this tool is not going to write for you that signals that this is a tool that is for the earnest writer.
1:03:14
We're not using statistics to replace writing and it is not the only thing that leads to great work but it is the thing that could identify your weak spots and help you improve so you can articulate your your ideas better.
1:03:28
Um the the the reason that I think this has to be statistical is because we need some sort of objective way to to kind of it the AI needs a framework on what makes something good because even though it has seen trillions of words, it has seen all of the great works, it doesn't have a model of what makes any of them good unless that is written.
1:03:52
Um, and so the goal is to kind of take words and then turn it into math, right?
1:04:00
And so this system is very much thinking in terms of of numbers and I'm there watching how it thinks and at first my model is like very off, right? I had these 27 prompts.
1:04:11
It's like, oh, it's not working at all.
1:04:13
And so I find all the nuances, then I make it more complex.
1:04:16
And I said, okay, now it's using 121 prompts and it's like slightly better.
1:04:21
But the whole thing is you you translate words into math.
1:04:26
>> But then you can't just give the writers a bunch of statistics.
1:04:28
That's not useful for them.
1:04:30
You have to make human readable compressed summaries on like and it has to be like fun and motivating too.
1:04:36
So you have to turn it into feedback too.
1:04:40
And and then I kind of realized like oh I got all the math right but it's like the feedback's not great.
1:04:44
So you have to keep changing the model.
1:04:46
And so now I'm at a point where for every draft you upload, it runs between 1 and 5,000 evaluations on your draft and and that I have the systems to evaluate and see the specific line level feedback uh it's giving you.
1:05:00
And uh the way I see it is writers love to romanticize writing.
1:05:06
They never romanticize writing. Writers hate editing.
1:05:08
It's like one of the most they people construct elaborate world views on like I mean think of Jack Carowak.
1:05:14
first thought, best thought, like editing is the devil.
1:05:16
Like people will come up with any excuse to avoid it.
1:05:19
And the hope is that hey, we can actually use math to have a very stable and simple way to help you understand how to edit.
1:05:27
Um, and the difference between my tool and a tool like Grammarly is that it's not just giving like little line edits like, oh, fix the tense of this verb.
1:05:34
It's trying to give you the the big picture sense of how you want to improve the composition of this piece.
1:05:41
And that often evol uh involves doing a complete rewrite. Right?
1:05:44
So the last piece I published it went through seven complete rewrites from scratch.
1:05:48
Uh and I think like writers avoid that even more, right?
1:05:56
Because there's like the sunk cost fallacy.
1:05:57
It's like why I worked so hard on this draft?
1:05:58
Why would I start from scratch?
1:06:00
But in my mind, when you rewrite, that is really when you push the edge of how you think about an idea.
1:06:08
If you're just writing a first draft and you're doing little line edits, you're not actually thinking because in my mind, writing is the process to reveal the thoughts that are in your head.
1:06:18
So, you can see what you think, but you can't necessarily change what you think unless you edit it.
1:06:24
Um, and so it's through editing that you you change how you think.
1:06:28
Uh, I love the quote uh by I think S.
1:06:31
Kelly Harold, which is that editing is not rewiring words, it's rewiring synapses.
1:06:35
uh which is exactly what this whole essay process is supposed to do.
1:06:40
So it's like hey you know let's let's try let's try using math to help lead to this very transformational benefit of of editing that everyone is already avoiding. >> Yeah.
1:06:52
Um I I certainly I I am not one of those writers having written four books that that uh uh you know romanticizes writing.
1:07:05
I I love the uh Dorothy Parker quote. I hate writing. I love having written.
1:07:11
>> Um and that was kind of me like but but >> it's a workout.
1:07:19
>> Yeah, it is a workout.
1:07:19
But listen, I owe most of my successes in my career to writing to to books.
1:07:24
And so I personally have definitely seen the power that writing can unlock.
1:07:34
Um, and one of the things I'm experimenting now with is I'll write a first draft of something and then reread it. I always read it aloud.
1:07:45
That's a a trick I'm sure you know. >> Yeah, definitely.
1:07:49
>> And like if it doesn't sound right to my ear, I'll rewrite it, etc.
1:07:51
But then I'll we have a our own AI lab at Oshara Subes.
1:07:56
Um and I'll put it in there and I'll give specific instructions to be like hey look into the liinal spaces here which we humans have a really hard time making those connections right uh and and tell me what you see right just as feedback. Mhm.
1:08:15
>> And I got to tell you, sometimes I have been like my jaw drops open like, "Oh my god, I I certainly didn't think of that."
1:08:25
And then that kind that kind of leads to the next jumping off point >> uh from from what it had surfaced as as a good editor, right, and contributor would be able to do.
1:08:38
>> I also completely agree with your idea.
1:08:42
Uh, again back to Mark Twain, right?
1:08:42
The the difference between the okay word and the perfect word is the difference between lightning and lightning bug. >> Yes. >> Yep.
1:08:56
And and the other thing that I love to read as well and and I and I think look, we're not we're not going to we're not going to re-engineer human beings here, but I think what we can do, especially with uh systems like yours, is at least give them a chance to have a fun way to to get it right.
1:09:21
Like >> I I don't know about you, but a lot of my friends back in the dark ages when we were we didn't have writing coaches.
1:09:27
We only took the SAT once.
1:09:29
We didn't have any study guides, and we only wrote that goddamn essay once.
1:09:36
And uh some of us were lucky enough to have friends or parents or family members who would take a look at it and and rework it a little bit.
1:09:43
But like I remember at least it was kind of like universally hated.
1:09:49
I remember when when we were all writing our college essays, my classmates were all like, "Uh, and I I was the editor of the school newspaper, so everybody came to me, will you read this essay?"
1:10:01
And I'm like, "No, >> you're the guy."
1:10:06
>> But I did read a few of them.
1:10:06
And like it did strike me even back then.
1:10:10
I just wonder there's there's the a theory that has been developed that um we've got the software pre-installed in our brains for learning how to talk, right?
1:10:22
You you take any child of any nationality or ethnic origin and drop them in a place, guess what?
1:10:30
Even if you borrow a child from Japan and move them to Queens, they're going to be speaking unacented English by the time they're five. Right. >> Right.
1:10:41
And vice versa, if you take a child born in Queens and move them to Japan and at birth, right, five years later, they're going to be very fluent without an accent in Japanese.
1:10:53
Some of the theories that I've read said that writing, on the other hand, very, very different because we developed it much later.
1:11:03
And it's one of the reasons why learning how to write takes a lot longer than just the natural ability to use your pre-installed software in your brain to pick up language.
1:11:17
And and in fact, I think it's the book uh the weirdest people in the world uh that that calls out that if you take brain images of highly literate person versus illiterate people, the shape of the brains of the highly literate people are different than the shape of the brains of illiterate people.
1:11:39
And what they contend in the book, The Weirdest People in the World, is that that happened because the uh ability for us to learn how to write colonized a different part of the brain.
1:11:54
If memory serves me, it was the one for visual acuity, >> which which were obviously pretty important to our hunter gatherer uh forefathers.
1:12:05
>> That's like edge detection, right? >> Yeah.
1:12:08
But do do you think that that might be like one of the lasting remnants why people are just like not not just naturally drawn to man I got to be a good writer.
1:12:19
That's that's going to be so um helpful in everything I do in my life. What do you think?
1:12:27
That's a that's a great question and and I'm of two minds on it because I really think that if you want to learn to compose better whether it's your idea, form or voice, one of the best things you can do is just read great essays. Yes.
1:12:42
>> Like like so part of this project is well I now have a list you can read these top 10 essays and you'll become a better writer just through osmosis.
1:12:47
And it's a similar thing where um >> you know what's the story Hunter S.
1:12:53
Thompson rewrote um was it Hemingway like word for word?
1:12:56
>> He rewrote Hemingway word for By the way, I was doing a little research and it wasn't just Thompson.
1:13:00
Many writers made a claim not for Hemingway but from taking a writer that they absolutely adored and literally rew writing it out or typing it out. >> Yeah.
1:13:14
>> But is that where copywriting the phrase comes from?
1:13:16
Like you were actually copying even though it's known today as advertising.
1:13:21
I think they were actually copying advertisements to know how they worked. I'm not sure.
1:13:23
But the other side of that, right, one, so one angle is, oh, writing can be learned through osmosis, but I think there's a similar to just spoken conversational language, but part of me believes that writing has to be studied to be mastered, right?
1:13:41
And it's like a lot of writers default and say, "Hey, just trust your intuition, right?
1:13:44
It's the answers are inside of you."
1:13:45
And I kind of think that's BS.
1:13:47
And most people just have bad intuition, right?
1:13:49
Especially when it comes to writing, that's why we all use cliches.
1:13:52
That's what it means to follow your intuition.
1:13:54
And it's through very conscious and deliberate practice in studying the different elements of writing.
1:14:00
Through doing that over and over, you eventually internalize it where you then you've trained your intuition, right?
1:14:06
Like the the end goal of essay architecture is to write through stream of consciousness.
1:14:11
You forget all the rules and it just comes out without you thinking.
1:14:14
But there has to be this intermediate process in how you actually become aware of these tools.
1:14:21
And so basically for example the concept of modularity and structure most people are unconsciously incompetent about that.
1:14:28
They don't even know that it exists.
1:14:29
And so now through SC architecture it's like oh well I know it but I no way I can do that.
1:14:33
And then through slowly editing you can gain competence and then eventually you forget it with enough reps you internalize it.
1:14:41
And and I get the sense that one of the reasons writing has never clicked is we've never actually built a formal system that maps out all of the compositional rules that we want to practice and then and then forget. >> Wow.
1:14:55
I that that's very insightful and I and I agree with you.
1:14:58
Uh, I'm a huge fan of what I call imbued or saturated intuition, which is quite simply like in my old job of asset management, I I got to a point where I had seen the same pattern so many times in markets that I just intuitively knew what was coming.
1:15:20
Right now, I tried to keep myself honest by anytime I would get an intuition.
1:15:26
I still do it to this day.
1:15:28
If I get an intuition about something, I always immediately put it in an Apple note so that I can see my real batting average.
1:15:36
You know the people who are always like, "My gut is always right." No. No, it's not.
1:15:43
>> They haven't wrote down all the bad ideas. >> Yeah. Yeah. No, it's not.
1:15:45
You think that and then you convince yourself of that.
1:15:50
And as you mentioned about the chat bots, right, confirmation bias is kind of the king of the biases, right?
1:15:54
Like of course people are going to love the chatbot that says, "Wow, what a great idea."
1:16:04
There was that the recent South Park about one of the characters always putting his business ideas into a syncupantished uh chatbot. Yeah.
1:16:14
>> Like what what about turning bacon into salads? Wow, that's great.
1:16:22
And so one one of the other tricks that I have found is very useful in our AI lab is I I I will say you it is your job to act as the most highly literate but vicious literary critic.
1:16:38
Take this chapter apart piece by piece. And guess what?
1:16:46
They follow instructions. Mhm.
1:16:49
>> Like I had a chapter for the book I'm writing now which is a fiction book which is very new for me um that I really loved quite frankly and then I put it into the our AI editor and made it that explicit instruction.
1:17:02
You are uh embodying the most literate the but the crulest and meanest and vicious critic. >> Yeah. >> Do your worst.
1:17:14
And oh my god, like it it did, but it was great because I'm like, you know what?
1:17:23
I hadn't I hadn't thought about that because I'd gotten so in love I'd gotten so in love with the way I was reading it that my confirmation bias was excluding all of those >> after I read its its vicious attack on me.
1:17:39
>> It's like a kill your darlings GPT. >> Yeah, totally. Exactly.
1:17:41
And you know that's one of our watch words at Infinite Books and has been a watch word in my life.
1:17:47
My wife was the first to say it to me.
1:17:50
You got to kill your darlings.
1:17:54
>> And I was very reluctant to at first. >> Sure. >> Yeah.
1:17:59
Sometimes I'll I'll I've done that before and I'll say, "Do it in the style of Hunter Thompson.
1:18:03
Be really funny and make me feel bad about it."
1:18:06
And like when it goes that over the top, it's like you're it's almost more palatable because you know it's like a joke, but then it's like it incepts the idea.
1:18:12
It's like, oh, he's actually kind of right when when he put it puts it that way.
1:18:18
And uh it kind of shows me that these chat bots aren't inherently sophantic.
1:18:24
That's actually part of the programming and the the product system prompt that they put into it.
1:18:28
And I think it's funny because I I see this whole, you know, I mean, this is a 2025 issue of like these chat bots are just like so o overly validating and that's never happened to me and I didn't get it.
1:18:40
I'm like, what's going on?
1:18:42
And it's because if you have a system in in a one system prompt, you can prevent that whole effect.
1:18:46
And uh what I have in mind is it says before every answer, state your assumptions so I know what you're basing this off off of.
1:18:55
And then at the end, like tell me where I might be wrong.
1:18:57
like give me the exact opposite perspective so I can at least um consider that.
1:19:00
And uh that's very different from the current system prompt because now it ends every response with saying well do you want to know more?
1:19:09
Like it's just trying to beat you into endless conversation, >> right?
1:19:12
Um and and I I love that you do that.
1:19:16
Um, uh, another habit I've gotten into is whenever I read something that I'm just sort of naturally in favor of, right?
1:19:26
I will now dump it into our AI lab and ask multiple large language models to make a steelman argument against the subject.
1:19:38
>> And wow, is that ever good.
1:19:38
And it just reminds me endlessly that we are kind of slaves to our own internal assumptions >> and our own internal preferences.
1:19:54
And you know it it it devolves at at scale into these us versus them tribalism that we see which is not great for society in my opinion. Mhm.
1:20:07
>> And and yet you see it when you see it yourself, you're like, "Okay, I get it.
1:20:12
I I get like the fact that you know this is a natural inclination if you're a human being."
1:20:18
Um do I do I I I I don't remember.
1:20:23
Do you have anything in your system that Yeah. Okay. Tell me about that.
1:20:28
So, that's the argument pattern.
1:20:30
And, uh, there's a few different criteria, but I think the worst kind of essay is when it starts by saying, "Here's what I believe, and anyone who disagrees is an idiot."
1:20:37
And then there's just like you spend 2,000 words just like unhashing this like this this stubborn belief.
1:20:44
And the best essays actually open with a question where it says, "Huh, like I'm typically on this side, but this extreme opposite point of view actually has some validity to it."
1:20:55
And then you actually the whole essay is about exploring the tensions in and this is like basic basic dialectics.
1:20:59
You have your thesis and you have your antithesis and by the end you arrive at some nuanced >> synthesis. Yeah. Good old K for the win. >> Right.
1:21:11
>> So tell us a little bit about the essay competition.
1:21:15
Uh I know that we uses and you were chatting about some other use cases.
1:21:21
Uh but the um the essay competition sounds fun.
1:21:25
Uh tell our listeners, viewers about that.
1:21:29
>> Yeah, so the inaugural competition for essay architecture, it starts uh September 15th and it'll run uh I think like 6 weeks uh through the end of October and it is a way to both try the tool and to see how far you can push this like recursive uh feedback loop.
1:21:48
Um, and one of the things that inspired me is to make this an open essay prize.
1:21:56
So, uh, there are all sorts of essay prizes there.
1:21:57
There's a $25,000 essay prize if you're a college student who will write an Einrand book report, right?
1:22:04
Or you can get a $50,000 essay prize if you are a, you know, consciousness researcher from a particular university.
1:22:09
So, we have all of these like hyper specialized high value essay prizes.
1:22:14
um in terms of open prizes and in my mind open is something that is not restricted to Canadians, right?
1:22:22
It's it's anybody can submit and uh the prompt enables any writer to enter.
1:22:28
The highest prize is is $5,000.
1:22:31
Um and so the idea here is for $10,000, this is the world's biggest open essay prize.
1:22:35
Um and the prompt for the competition is simply the year. So it's 2025.
1:22:41
Write an essay about a personal experience of yours that is microcosmic of something that unfolded in this year.
1:22:49
And so the goal is if you have a thousand people trying to write a world-class essay on this theme, the best 10 or 20 essays, you'll have a great anthology of sharp thinkers with highly uh refined essays to help us make sense of the weird times that we're always in. Right?
1:23:08
And so I love the idea of closing the year and saying, "Okay, great.
1:23:11
We've got all the best essayists and here are just different meditations on on what's happening." >> Very cool.
1:23:17
And you say the top prize is 5,000 >> $10,000. >> $10,000.
1:23:22
And then are there prizes beneath that?
1:23:26
>> Uh still working out the sub prizes.
1:23:26
Uh there could be a world in which if we publish the book through Metal Label, uh there could be some sort of royalty share between all of the finalists.
1:23:34
Uh so there's there's different models still working that out.
1:23:38
Uh in terms of the judges, so the idea is to have I think we mentioned this, but human readers and also a guest judge.
1:23:43
So in the end, it's not just AI doing the scoring.
1:23:47
You want to write an essay that is going to appeal to a human, right?
1:23:52
There's no way to cheat the system just through coming up with some prompt injection, which I'm I'm building defenses against, but just sharing that that this is like who can put in the earnest effort to write a great essay. >> Yeah.
1:24:02
And it's a great way to highlight um uh upandcoming authors.
1:24:07
Uh >> exactly >> who that that's one of the things that I'm most excited about not not just the connectivity of the internet but also the ability of AI to surface talents that like in earlier eras just like nobody would have ever heard of, right?
1:24:29
um because of where they lived, because of their circumstances, because of a host of reasons why they, you know, were were just lived in their little village.
1:24:42
So, >> right, >> I I think it's a a great idea and obviously one that we will support enthusiastically uh from our various uh divisions at Ocean Ventures. Uh what's next? What's next for you?
1:24:57
What do you you just want to keep improving essay architecture?
1:25:03
Um or do you have something new and different planned?
1:25:07
>> I think right now essay architecture it's just like I have all of the prototype versions of a platform, right?
1:25:14
Like there's a curriculum, there's data, there's a tool, and then there's media.
1:25:18
And so I think just focusing on uh this project could last me many years.
1:25:22
So I feel like I'm just starting this thing.
1:25:25
Um, simultaneously I'm expecting a daughter uh end of this year.
1:25:28
So that's going to be another project to work on.
1:25:31
And uh I have this idea um and this is again another tangential idea related to writing but it's an idea to redesign the dictionary.
1:25:41
It's really more of a thesaurus but it's trying to give a new architecture to our language.
1:25:45
Um, and so the problem with the thesaurus in my mind is that um, you know, it gives you a whole cloud of synonyms, but it doesn't actually give you a very concise definition on why you would use one synonym over another.
1:25:59
You know, when do you use elated versus jubilant, right?
1:26:04
Like there like that's useful to know otherwise you're just bullshitting and throwing in random words from your synonym checker, you know?
1:26:09
Um, and so I think that's one dimension, but I think the way to organize a language is to figure out what might be 500 core synonym antonym pairs.
1:26:20
So for example, you have happy, sad, rich, poor, um, tired, energetic, right?
1:26:25
What are the core synonym pairs?
1:26:27
And then what are all of the synonyms below them with their um, corresponding like unique uh, qualifier?
1:26:35
And my theory is that if you built out a thesaurus dictionary in this way, it would be much easier to uh advance your vocabulary, right?
1:26:44
Because there's all these studies where if you just throw a human in a swarm of complexity, they retain almost nothing.
1:26:51
Whereas if you package it with an architecture, they can actually not just retain it, but leverage what they've learned.
1:26:57
And I think the same applies to vocabulary.
1:26:58
And my thesis, and this ties into having a kid, like I have five years.
1:27:03
Can I build a new thesaurus that can help kids um understand the lu nuances of a language?
1:27:08
Where someone who's graduating high school, instead of being illiterate with only a 15,000 vocabulary, might they be able to know 60,000 words just because we have a better design dictionary?
1:27:18
Um, so in my mind you have essay architecture and then there's another crazy project that goes down to the word and then there's another project that goes up to the scale of ideas.
1:27:29
So I just have one more project.
1:27:32
Again, this will take 10 20 years to unfold, but it's like trying to actually map out the the core archetype archetypal ideas that have occurred through history and trying to say like, oh, maybe there are actually like a hundred lineages and we should understand how from ancient times to modern times, the same idea has been evolved.
1:27:51
Um, but yeah, I think like once you have that full stack, it's like you're giving someone mastery of composition, of of diction, and of philosophy.
1:28:00
And like that is almost like a core sense of what it takes to be educated. >> I love that idea.
1:28:06
Um, when when you flesh it out more, definitely share it with me because uh I think that's that's really you you you might be the Dr.
1:28:17
Samuel Samuel Johnson of the 21st century, >> right?
1:28:21
He essay writer made a dictionary.
1:28:26
>> Well, Michael, this has been really, really fun.
1:28:28
We are very, as you know, enthusiastic about what you're building.
1:28:34
Uh, we think it's turned out super well.
1:28:36
We think there are multiple use cases for it.
1:28:40
>> Um, and it's a very robust and powerful tool for learning, which we find really interesting.
1:28:47
interesting. uh in terms of you know it's not going to do it for you but it's going to help you learn how to become a better writer which will in my opinion obviously unbiased uh will uh impact many many other areas of your life uh positively
1:29:08
uh as you know uh at the end of uh each episode we make you the emperor of the world you >> I've been waiting for this to be to be emperor Well, it's not going to be a really fun gig, Michael, because you can't kill anybody and you can't put anybody in a re-education camp. But what you can do
1:29:25
But what you can do is we're going to hand you a magical microphone and you can say two things into it that will incept the entire population of Earth, all 8 billion plus humans.
1:29:40
The next morning they're going to wake up and they're going to say, you know, I just had two of the best ideas and unlike all the other times where I didn't follow through on them, I'm going to actually act on these two.
1:29:53
What are you going to incept in the world's population? >> All right.
1:29:59
So, I think I'm going to do one writing related one and one not writing related. If I have two. Okay.
1:30:05
The first one is like a pretty practical challenge and it would be for tomorrow to write down 100 of your thoughts.
1:30:10
It's an experiment that I've done a lot. I don't do it every day. That'd be crazy.
1:30:16
But I probably do write 10 thoughts, passing thoughts per day, but a hundred is pretty remarkable because if you do it for eight hours, that's uh basically a thought every five minutes.
1:30:26
And uh my argument is that that could be the best form of of journaling, right?
1:30:31
to actually see to use writing to see what's in your head.
1:30:35
And I know morning pages are very popular, but usually it's like, oh, I'm going to create this separate space in my life where I get all my feelings out and it's disconnected from what I'm actually doing.
1:30:44
But if you have like an output log of what you're doing every 5 minutes and what you're thinking, you uh you have the ability to actually see the patterns in your in your thinking and improve from that.
1:30:53
You have material for essays.
1:30:55
You have the training corpus for an AI.
1:30:57
And when you just watch yourself think, it's like a new lens to to reality.
1:31:02
So many people are like, "Oh, I'd love to write, but I don't have ideas." It's like, "Oh, no, no."
1:31:05
Like, your head is filled with ideas.
1:31:07
You're not paying attention.
1:31:09
So So that's the challenge.
1:31:10
Write a hundred thoughts tomorrow. >> I love that one. I I'm going to do it.
1:31:16
So what what what's number two? >> Okay, number two.
1:31:19
Uh this is a little more abstract and I wanted to make it about the uh the human OS.
1:31:24
If I can make one contribution to the add to the human OS or to you know augment it would be to uh become a master of conditional logic around your own identity.
1:31:37
And what that means to me conditional logic if you ever written software is just a very basic if then statement.
1:31:43
I think everybody knows what that is.
1:31:45
But when it comes to our identity, I think we tend to think very like in singular things like, oh, I'm a New York Mets fan or I belong to this religion or to to this uh, you know, political party.
1:31:57
But conditional logic says, I'm not just going to bulk subscribe to all of the different tenants of this identity.
1:32:04
I'm going to pick them apart and say, well, in some conditions I do this and in some conditions I'm not about this at all.
1:32:11
And uh I just think we avoid that style of thinking because of fear of you know complexity of being illegible.
1:32:16
No one gets you uh fear of just ambiguity like we hate uncertainty.
1:32:21
But if you really take conditional logic seriously it's like the kernel of of cognitive liberty.
1:32:29
You're actually thinking for yourself.
1:32:30
You're not just part of some thing you absorbed.
1:32:32
Um, and yeah, conditional logic and and I think that to come bring it full circle, that's what the essay is about, right?
1:32:41
You have some tension and you get into the nuances and you get some new lens to the world that it's complex, but it's distilled and made simple so you can just live through it. >> Wow. I love both of those.
1:32:52
Uh, I'm a huge fan of conditional logic when thinking about myself.
1:32:59
I I the uh I I've never really understood our our just basic almost at our base code this need for certainty >> like I the illusion of control because that's what it is. It's an illusion.
1:33:17
I always I always laugh when people are talking about oh those large language models always hallucinate.
1:33:24
Hey, guess who was the original confabulators and hallucinators? We human beings.
1:33:29
And and so the idea I I personally think that certainty is a kind of death >> because like you know okay I'll give you I'm fairly certain I'm 99% certain that the sun's going to come up tomorrow and like so for the 1% error factor I'll never know if it doesn't if it doesn't come up tomorrow.
1:33:58
So, but but people apply the certainty way way too much and I think that it shrinks their life because it closes them off to many many things that they might like really really enjoy or really really grow or or whatever have fun it all of it when when you absolute certainty is absurd in my opinion.
1:34:25
Um, you know, I often say one of the greatest challenges is we are deterministic thinkers living in a probabilistic world and hilarity or tragedy often ensue.
1:34:42
And the more you can change your thinking like you've suggested for your second u inception uh through making it a little more conditional, you're going to I think you're going to find wow there's a whole other world out there that I I now have access to.
1:34:59
Michael, this has been Yep.
1:35:02
You you had >> if I just add one quick thing is I think it's almost inefficient.
1:35:05
It would be inefficient to to run through a whole experiment if the sun's gonna come up tomorrow, right?
1:35:10
When you when you think like that, you're you're effectively doing the proofs on your own from scratch.
1:35:16
You're rederiving Christianity or whatever it is.
1:35:17
And it's very inefficient.
1:35:19
But in the end, by doing it, this is the whole thing where why it's worth doing things manually.
1:35:23
When you actually think through it, it becomes more efficient because you now have an upgraded lens that colors everything.
1:35:30
So in the end, it is worth it. >> Totally. Totally agree.
1:35:33
Where can everyone find you, Michael?
1:35:35
I know you have a substack. >> Yes. So it's michael.
1:35:38
site site for my Substack and saarchitecture.
1:35:40
com if you want to read the book, try the tool or enter the competition. >> We are huge fans.
1:35:47
We will do everything we can to amplify it so you get a lot of uh of uh submissions to the competition. Uh thanks so much. This was super fun.
1:35:58
>> Yeah, this was great, Jim. >> Cheers. Bye.