Please please resume the meta-rationality book! That book is so dear to me. I am so glad that I found it when I experienced meaning crisis almost nine years ago.
As a software engineer, I feel lucky that you are considering adding more examples from software engineering. I love the idea whole-heartedly!
I love the self-analysis on the book. I've found it great, but also pretty abstract from the get. Maybe you're getting a chance to practice the product management skill subset, figuring out what users really need with the book? 👀
Re. the fusion of product manager, architect, and project manager - IMO figuring out how to harness teams of agents together into a system that does work correctly AND refines itself when it messes up is the goal. In other words, the same org-building every CEO does.
I suspect it starts with looking at what "work" even is using what we've learned from orgs of humans harnessed together into companies. So places like Linear, Notion, Todoist, etc will have big lessons on offer.
I became a meta-rational software developer by accident because the analysts supposed to be doing it were so bad. I was reviewing their work and I couldn’t stand it. So I stepped up and designed two custom systems. I’m not sure how I got the skills - I understood data and I could think and write. I didn’t get much respect from the higher-ups, but the programmers and testers and users loved me because I could answer their questions. All over now, I retired in 2018.
Hey David, I would highly appreciate if you continue the book. I am waiting for your perspectives on how AI interacts with the reasonableness/rationality/metarationality for such a long time already.
So many people are worried what to do in a world where rationality is done by machines. Your book could point out a direction for finding meaning again.
One more datapoint, where I work at Meta engineers are "red-flagged" if they are not generating enough AI agent code! The code I produce has been 100% AI generated for months, and I have also been contemplating that I am now employed mostly for my meta-rational/taste/organic alignment skills (on the contraries, thousands of my colleagues have been fired in the process)! Timely post David :)
There is a school of thought that you don't read the code that is produced, you just set up tests and harness so you can trust it. Even if you read the bottleneck is not code review but is alignment and meta-rationality
Lovely essay; strikes me that meta-rationality plus meta-design plus meta compassion would constitute philo-sophia today, i.e., what the pursuit of wisdom would be if it arose out of the Alayavijnana in 2026 instead of India or Greece or China during the Axial Age.
I'd definitely like to read stories that you believe are relevant for meta-rational principles. The trouble is that they might be a bit dry compared to stories about relationships! To have a a chance of being interesting, I think they need to be based on real stories, including stuff that's not directly relevant, to feel real?
The Parable of the Pebbles worked pretty well though, so maybe that style of storytelling would work too?
As someone who has started using AI tools to code at the end of last year and hasn't looked back I have often thought to myself that the bottleneck that's left is meta-rationality, so I'm glad to hear you're thinking along the same lines. I would like to listen to you explore more on the subject.
One thing I've seen that was a very interesting use of AI was this project: https://banteg.xyz/posts/crimsonland/. The TLDR is that in a month or so he was able to completely decompile a game from 2003 into a modern programming language (with the behavior of the binary as the spec, warts and all.)
I find this fascinating because appears to be something uniquely suited for AI agents but very difficult for even a team of experts. This was a 10 million dollar project done by one crazy person in a month. What does it mean to live in a world where "you have the binary, therefore you're only a few months of work away from having something isomorphic to the original source code" us true? I haven't a clue.
This seems similar to Anthropic's announcement (yesterday? I think) that they could translate large COBOL systems into high-quality modern-language ones. (Which instantly sent IBM's stock price down by 20% or something.)
Yeah, I heard about that. The game one hits me harder because it feels more concrete. Whereas people having been shilling stuff like that for the past 4 years. I'm sure it could be done for COBOL though (the hard part would be designing the harness/testing loop to force agents to double check it actually matches with what the COBOL does).
I personally enjoyed the abstract approach you took so far. And as a non-software engineer, I am not sure i would find the coding based inverted approach as accessible or engaging; although it might be a good chance to dust off my Abelson & Sussman - people still code in Lisp, right? :)
A related question I was not totally clear on is the intended audience and goal of the work. If I understand it correctly, and I am knee deep in Kegan right now to try to improve my understanding, progressing even a quarter stage in cognitive development from rationality toward meta-rationality could be a year(s) long process of lived experience and character building vs a skill that you can pick up from reading any amount of text. So is it correct the audience could be at a cognitive development stage somewhere in the rationality to meta-rationality zone, and the goal is to provide some combination of theory, inspiration and scaffolding, to be referred back to by the reader in the process of climbing ever higher on the rationality->meta-rationality ladder, as a lifetime endeavor? I can definitely see how that could mushroom into a much larger project than a single book, although maybe not necessarily so.
> I was not totally clear on the intended audience and goal of the work.
Yes, I also was not totally clear on this, which may be why I had a lot of trouble deciding what form it should take! Sorry about that.
> So is it correct the audience could be at a cognitive development stage somewhere in the rationality to meta-rationality zone, and the goal is to provide some combination of theory, inspiration and scaffolding, to be referred back to by the reader in the process of climbing ever higher on the rationality->meta-rationality ladder, as a lifetime endeavor?
Please please resume the meta-rationality book! That book is so dear to me. I am so glad that I found it when I experienced meaning crisis almost nine years ago.
As a software engineer, I feel lucky that you are considering adding more examples from software engineering. I love the idea whole-heartedly!
Do you have other recommendations for people encountering a meaning crisis?
Possibly Lulu was referring to this: https://meaningness.com/
Ah yes, thank you. I began reading that a while ago but never got deep into it. Perhaps this is a sign to give it another go.
Possibly this new thing would help: https://chapman-tracker.netlify.app/
(Not by me—written by someone I don't know, who pointed me at it a few days ago. Pretty cool I think!)
Yes! I'd be very excited to read about more explanations and examples of meta-rational software development!
I love the self-analysis on the book. I've found it great, but also pretty abstract from the get. Maybe you're getting a chance to practice the product management skill subset, figuring out what users really need with the book? 👀
Re. the fusion of product manager, architect, and project manager - IMO figuring out how to harness teams of agents together into a system that does work correctly AND refines itself when it messes up is the goal. In other words, the same org-building every CEO does.
I suspect it starts with looking at what "work" even is using what we've learned from orgs of humans harnessed together into companies. So places like Linear, Notion, Todoist, etc will have big lessons on offer.
I became a meta-rational software developer by accident because the analysts supposed to be doing it were so bad. I was reviewing their work and I couldn’t stand it. So I stepped up and designed two custom systems. I’m not sure how I got the skills - I understood data and I could think and write. I didn’t get much respect from the higher-ups, but the programmers and testers and users loved me because I could answer their questions. All over now, I retired in 2018.
Thank you! Yes, this is a classic experience :) Usually how it happens. Thanks for telling your story!
Hey David, I would highly appreciate if you continue the book. I am waiting for your perspectives on how AI interacts with the reasonableness/rationality/metarationality for such a long time already.
So many people are worried what to do in a world where rationality is done by machines. Your book could point out a direction for finding meaning again.
Thank you for all your hard work.
One more datapoint, where I work at Meta engineers are "red-flagged" if they are not generating enough AI agent code! The code I produce has been 100% AI generated for months, and I have also been contemplating that I am now employed mostly for my meta-rational/taste/organic alignment skills (on the contraries, thousands of my colleagues have been fired in the process)! Timely post David :)
How much do you review the code that's produced? I'd imagine you'd be bottlenecked by that pretty quickly?
There is a school of thought that you don't read the code that is produced, you just set up tests and harness so you can trust it. Even if you read the bottleneck is not code review but is alignment and meta-rationality
Another line to express this "when implementation is cheap, alignment is costly"
Lovely essay; strikes me that meta-rationality plus meta-design plus meta compassion would constitute philo-sophia today, i.e., what the pursuit of wisdom would be if it arose out of the Alayavijnana in 2026 instead of India or Greece or China during the Axial Age.
I'd love to read more essays about practical software engineering examples of meta-rationality (and any other wisdom you have about the field)!
As a programmer I'm a very hands-on person, so it's hard for me to learn something if I can't immediately apply it
Indeed, please! I’ve been following the project a long time and love it.
I'd definitely like to read stories that you believe are relevant for meta-rational principles. The trouble is that they might be a bit dry compared to stories about relationships! To have a a chance of being interesting, I think they need to be based on real stories, including stuff that's not directly relevant, to feel real?
The Parable of the Pebbles worked pretty well though, so maybe that style of storytelling would work too?
As someone who has started using AI tools to code at the end of last year and hasn't looked back I have often thought to myself that the bottleneck that's left is meta-rationality, so I'm glad to hear you're thinking along the same lines. I would like to listen to you explore more on the subject.
One thing I've seen that was a very interesting use of AI was this project: https://banteg.xyz/posts/crimsonland/. The TLDR is that in a month or so he was able to completely decompile a game from 2003 into a modern programming language (with the behavior of the binary as the spec, warts and all.)
I find this fascinating because appears to be something uniquely suited for AI agents but very difficult for even a team of experts. This was a 10 million dollar project done by one crazy person in a month. What does it mean to live in a world where "you have the binary, therefore you're only a few months of work away from having something isomorphic to the original source code" us true? I haven't a clue.
Cool, thank you!
This seems similar to Anthropic's announcement (yesterday? I think) that they could translate large COBOL systems into high-quality modern-language ones. (Which instantly sent IBM's stock price down by 20% or something.)
So what you are saying is I should learn COBOL and how to use AI?
Yeah, I heard about that. The game one hits me harder because it feels more concrete. Whereas people having been shilling stuff like that for the past 4 years. I'm sure it could be done for COBOL though (the hard part would be designing the harness/testing loop to force agents to double check it actually matches with what the COBOL does).
I personally enjoyed the abstract approach you took so far. And as a non-software engineer, I am not sure i would find the coding based inverted approach as accessible or engaging; although it might be a good chance to dust off my Abelson & Sussman - people still code in Lisp, right? :)
A related question I was not totally clear on is the intended audience and goal of the work. If I understand it correctly, and I am knee deep in Kegan right now to try to improve my understanding, progressing even a quarter stage in cognitive development from rationality toward meta-rationality could be a year(s) long process of lived experience and character building vs a skill that you can pick up from reading any amount of text. So is it correct the audience could be at a cognitive development stage somewhere in the rationality to meta-rationality zone, and the goal is to provide some combination of theory, inspiration and scaffolding, to be referred back to by the reader in the process of climbing ever higher on the rationality->meta-rationality ladder, as a lifetime endeavor? I can definitely see how that could mushroom into a much larger project than a single book, although maybe not necessarily so.
> I was not totally clear on the intended audience and goal of the work.
Yes, I also was not totally clear on this, which may be why I had a lot of trouble deciding what form it should take! Sorry about that.
> So is it correct the audience could be at a cognitive development stage somewhere in the rationality to meta-rationality zone, and the goal is to provide some combination of theory, inspiration and scaffolding, to be referred back to by the reader in the process of climbing ever higher on the rationality->meta-rationality ladder, as a lifetime endeavor?
Yes, that seems a good way to approach it.