Theorizing an ideal intellectual career
An intellectual career is comprised of a series of projects over the course of a lifespan. Intellectuals include philosophers, scientists, artists, and many other types, but their one defining characteristic is that they seek to optimize for truth production. Implicit in “production” is both the discovery of novel insights and some form of public manifestation or impact on the minds or behaviors of others. If the intellectual effects zero updating in the minds or behaviors of other people over the long-run future, their total truth production is zero — no matter how many genius discoveries they might have made privately. The purely disinterested goal of maximal truth production requires at least one short-term goal: meeting some minimum of resource sustainability, namely the time, energy, and money required for truth production.
Consequently, intellectuals should seek to build a system that is both effective at producing truths, but in such a way that is self-funding. Executing non-intellectual work for money, to pay for intellectual work, is unlikely to maximize long-run truth production for two reasons. First, obviously, precious time will be wasted on non-intellectual work. But more insidiously, the social compromises required to maintain non-intellectual income streams impose hard ceilings on the intensity of a recipient’s truth production. This is for the simple reason that normal people avoid and hide truths for the instrumental benefits of doing so. Academic researchers are the most dramatic illustration of both problems: They spend most of their time doing non-intellectual work (meetings, paperwork, fundraising, etc.) and even the intellectual work they do is typically proof-of-work fitness signaling, where real erudition and genuinely painstaking scholarly rigor nonetheless function primarily as costly signals proving the scholar’s relative superiority to other scholars.
To be clear, it is perfectly rational and arguably good for people to avoid the full truth of reality, for this is how civilization purchases its extraordinary utility-generating norms and institutions. Science, upon which modern progress is based, is the aggressive simplification or reduction of the world’s exceedingly complex dataset. This reduction or strategically erroneous departure from the whole truth of reality is incredibly useful, and yet it is departure from the truth nonetheless. The everyday lies and errors enjoyed by normal people are only the quotidian, personal version of the strategic reductions employed by scientific rationality.
Thus, the historical task before us is to consciously create systems that better align the everyday work of truth-seeking and truth-distribution, with the accrual of resources necessary to support this work.
I’ve spent the past few years doing my best to develop such a system. Eight months ago, I had enough confidence in my fledgling system to quit my academic career. Now that I’ve been living on my system full-time for almost a year, I consider it sufficiently battle-tested to start sharing it.
Building an Intellectual Content Machine
One lifetime intellectual career is composed of what I will call a machine, which “runs” over the lifespan. My machine is composed of multiple systems. Creative intellectual work organically sediments raw content, which is deposited in buckets. A system is a process that takes content from buckets, executes a set of standardized steps called a routine, resulting in publication of consumable outputs at various sites. Crucially, every system runs on a recurring schedule (a system without a schedule is only a vague plan unlikely to get done).
In my ontology, there are three main types of content. Other kinds of intellectuals will need to alter this — if you are a painter, for example, presumably “painting” would have to be one of your content types. But I would guess that my primary content types are also the primary content types for the overwhelming majority of intellectuals today:
One crucial question an intellectual is constantly confronted with is, “Upon conceiving a thought, how should I record it?” I would submit that there are two key variables to consider for the optimal answering of this question.
The Hierarchy of Content Types: Video > Audio > Text
The first key variable in raw content creation is simply the quantity of information. Other things equal, you should always opt to register your thoughts in the most information-rich content type possible. This is for the simple reason that more information-rich media are easily converted into less information-rich media, but not vice-versa. Consider one idea, in the form of one sentence. Speaking this sentence into a video camera results in a content item with several millions of bits of information. Typing this sentence in a plain text file results in a content item with only a few thousand bits of information. A video clip is easily converted to an audio clip, and transcribed into a text file. But an audio clip is not easily converted to video, and a text file is not easily converted to audio or video. Thus, there is an obvious hierarchy of content types: To whatever degree possible, record thoughts and ideas in the form of video, then implement systems that distribute video into audio and text.
Of course, some types of work are easier to produce with text only, such as research articles.1 Some personality types are highly averse to facing a camera, etc. These are all perfectly fine considerations, and all intellectuals should adapt my method accordingly! I only say that video is better than audio, which is better than text given that all other things are equal. Not to mention that information-richness also seems correlated with audience “engagement.”
The second key variable related to raw content creation is the latency between idea conception and content creation, or simply the time that elapses between conception and fruition in a medium. Latency is crucial for two reasons. Every minute of latency, in my experience, increases the probability of fruition never occurring. “If I don’t write it down immediately, there’s a good chance I’ll never write it down.” Second, conception brings its own motivation and is associated with intrinsically satisfying psychological states, which make the labor of creation much, much easier than it will be the next morning after the rush of “Eureka!” has worn off.
Ideally, upon the intuition of any good thought, a neural implant would immediately and automatically convert it to a “deep fake” video of me speaking it, an mp3 file of the audio, and a text file. Until this technology is readily available, all we can do is cultivate workflows that reduce latency as much as possible.
There are many possible buckets, depending on how, why, and when you have and record thoughts and develop ideas. Generally, my primary buckets are as follows.
- One primary writing app. Ideally it’s available on all of your devices and syncs easily. All text goes into this app. I use Ulysses but there are many others. I have not done an exhaustive review although I have tried Drafts, Typora, and plain text files in folders. Ulysses seems the best balance between simplicity and functionality, at least for my type of writing.
- The SD card on my nicer camera collects video and photos over time.
- My iPhone camera roll is a bucket that collects video and photos over time.
- The Voice Recorder app on my iPhone collects voice recordings.
- Livestreamed videos archived on Youtube are themselves a kind of bucket, for my particular case.
- One directory in my Dropbox is a bucket for the outputs of reading/note-taking automations (see Routines).
- Devonthink Pro is my catch-all Master Bucket for everything that doesn’t fit into any of the above
Every one of my systems involves a routine (a standardized, repeatable, multi-step process) that takes items from these buckets and results in some set of scheduled pushes to my sites.
A site is any place where one can publish content recurringly. There are many possible sites today. Any one account on any social media platform is a site. A blog is a site. Other examples include Youtube and email newsletters. Not sites:
books (electronic or print), courses (online or in real life), as well as talks and one-off performances. These are better understood as goals and should be the culmination of a particular operation, project, or sprint.
The primary sites of my own machine are:
I often play with other social media platforms but I cannot claim to have systematically worked them into my machine quite yet. There are few platforms I do systematically syndicate to, but the results they produce are insignificant enough to skip them for now (e.g., Tumblr and LinkedIn).
External inputs, or just inputs for short, refer to all media one consumes as fodder for the machine. The primary inputs for my machine are:
- Web pages (blogs, news, full-text of books, or links to videos, podcasts, etc.). To whatever degree possible, I run all web pages through Feedly as my one-stop for reading, highlighting, and commenting on internet content.
- Digital books
- Paper books
I don’t really use video/audio content as inputs; but rather I use a text placeholder, usually in the form of a link to the video/audio.
Routines represent the real engineering that define one’s intellectual systems.
The primary consideration in constructing routines is to make them as standardized and automation friendly as possible. Ideally, web services such as IFTTT or Zapier can truly automate whole routines. More often, you can automate parts of a routine but you must do manually other parts of a routine. Even still, for the parts one must do manually, one still wants those parts to be standardized to make it easy to manually execute (and also to be prepared as the tasks become automatable).
One of my most important workhorse routines is what, in my documentation, I call the Podflow. Here’s how it works.
Whenever I conduct a live video podcast with someone via the Streamyard service, it livestreams and auto-archives on Youtube. As I explained above, Youtube serves as a bucket for these videos. The Podflow routine processes these archived videos as follows. All of this is written on an index card, which I refer to when I execute the Podflow.
- Manually trim the video’s beginning and end in YT editor
- Download audio track with Clipgrab app
- Upload audio to Auphonic (automated audio editing)
- Select my custom template for adding podcast metadata in Auphonic (adds intro, outro, and other things)
- Have Auphonic distribute the final track to:
- Libsyn as a podcast draft
- Patron-only Google Drive folder
- Manually post edited podcast to Patron audio feed
- Share Patron-only audio post to Buffer (pretty much the only patron “marketing” I do)
- Schedule Libsyn draft to publish one week later; Libsyn includes automatic syndication to multiple pod directories.
- When podcast publishes, Libsyn automatically syndicates a new post with embedded audio player to my blog
- A Zapier zap automatically pushes the blog post link to my two Twitter accounts via Buffer, my Discord server, my Facebook, LinkedIn, and Tumblr.
- A Zapier zap automatically adds an html link item, with formatting, into a Dropbox-hosted text file that accumulates items for my weekly newsletter.
This routine converts a raw content item (a video) from a bucket (my Youtube archive) into almost 20 different sites. It relies on much automation, but requires some labor. I know how long every part takes, on average, and I execute the Podflow on a regular schedule (once a week). I’ve converged on this particular flow after a lot of trial and error with different processes. Manual audio editing is simply too time-consuming, so currently I refuse to do it (though I tried many times). All the syndications options have to be figured out iteratively, as you check how things look when they auto-post. So start small and improve routines between project cycles.
My Podflow could be improved in many ways, which I just haven’t done yet. For instance, ideally, I would download every video podcast and import it to the Descript app, which is an app that can be used to produce a transcript, podcast, and multiple video highlight clips for distribution on social. It’s great, but it requires more labor than I can give, and which I can’t yet afford to hire. Though I can tell you how I might try! Highlights from the transcripts could be collected into a book for sale, which would help to pay for an assistant who does all of the manual editing required by this more sophisticated process. I would have even more free time to think and create, and publish even more content, at the same time. But if you’re bootstrapping your own production system, you can only build up over time. Right now, I just can’t afford this better system. What matters is that I have a system, I am capable of doing it on schedule (more or less!), and it produces the best possible results I’ve been able to engineer. After staying consistent with this system, for 6 more months, perhaps I will have the resources (not just capital but followers, etc.) to justify designing and implementing more sophisticated process.
I have many other routines. It would require a whole book to document in detail all of them. The Podflow serves as one specific example. My other primary routines are:
- The Podcast Guest Scheduling routine (systematizes an Airtable, emails, Calendly, Streamyard event creation, promo tweets, and preparation tasks)
- The Blog Post routine (Ulysses, Airtable, Convertkit email sent to subscribers who signed up to receive all blog posts)
- The weekly newsletter routine
Temporal Dimension: Operations > projects > sprints
Temporally, an intellectual career is decomposed into operations. I use the word “operation” as militaries use it, which means you can assign them exciting names such as Operation Rolling Thunder. One operation cycle is composed of a machine, plus at least one measurable goal. Implicitly or explicitly — explicitly is better — an operation always involves a hypothesis or set of hypotheses. A hypothesis is simply a rationale for how and why the machine is expected to achieve the goal. At the end of each operation, the machine either meets the goal (hypothesis confirmed) or fails to meet the goal (hypothesis rejected).
The shorter the operation cycle, the more rapidly the intellectual will learn what is working, with respect to both truth-discovery and financial sustainability. Between every operation cycle is an opportunity to update one’s machine. Do not update machines in the middle of an operation, although you may decide to dedicate a whole operation to the construction or overhauling of a machine.
Though my own project cycles are 6-weeks-long, I suspect this length of time should be adjusted proportionate to trait Conscientiousness. Impatient and distractible intellectuals (the wild artist archetype) should decrease the length of project cycles until they can reliably follow through on them. Obsessively meticulous intellectuals (the scholarly archetype) may wish to try longer project cycles later, but only after settling on a well-functioning machine achieving adequate results. Long project cycles are very risky, because if you’re doing something sub-optimally, you might not have time to course-correct.
Projects and sprints
Operation cycles are further decomposed into projects, and then sprints. Again the parameters should probably be adjusted according to personality, but I try to plan 2-week projects (3 per operation) and 5-day sprints (2 per project, for a total of 6 sprints per operation).
Justin Murphy for IndieThinkers.org
Version 1. February 18, 2020
1 Actually, research articles are highly processed aggregations of hundreds of thoughts, ideas, and insights. In my approach, one research article would likely be a whole operation, the process of which would generate potentially dozens of videos, podcasts, and blog posts. Composing the actual research article would likely be a short sprint, primarily involving the synthesis of already published content items.