Read this on Substack.

Today, I saw one of the creators (Twitter influencers? Growth engineers?) that I actually choose to follow post a video on how she uses AI to make “high quality educational content at scale.” Holy oxymoron.

As someone who has been trying (in vain, I might add) to grow a following for a while, I am no stranger to this sentiment. The key to gaining traction on basically every platform — X, TikTok, LinkedIn — is to post everyday, multiple times a day. As someone who also enjoys making things — essays, lessons, videos — I can say with absolute certainty that the key to making these things high quality is most certainly not to do them at scale.

I’ve seen an explosion of this sentiment of scalability across everything. Tokenmaxxing and agent loops are the new vogue — every startup dreams of hiring engineers that will run up a bill on Anthropic that is higher than their yearly salary. It exists in the faster and faster cycles of fashion, in endurance sports, in the explosion of networking events and the number of people at said networking eventsI find 100+-person coworking events rather antithetical to the oft-stated purpose of making meaningful new connections, but I do think they are cute and I would totally throw one. At that point, it’s really just a computer party. all over San Francisco. And, of course, it exists in the world of online content creation. As our attention spans have gotten smaller, the demand for bite-sized pieces of content that pique our interest but don’t force us to stay engaged for too long has exploded. My empirical review of the platforms I use confirm this; even on Substack, where the whole point is supposedly to enable individuals to have a platform to produce long-form content, what really does numbers are the notes. I’ve seen pictures of tweets go more viral than any essay written by (Nobel Prize winning economist) Paul Krugman.

The scalability of content is a difficult thing to escape, because engagement truly is king. The social platforms we use are not inherently malicious, they are just trying to drive numbers. A five second video, a ten word tweet is way more digestible, is seen more widely by way more people. And as it is liked more, the algorithms that govern content placement on these platforms recognize that package of bytes is getting a lot of traction. It is thus shoved to the top of your unsuspecting feed because all the priors point to the fact that you’ll see it and after you see it, you won’t click out of the platform. God forbid you saw something that made you put down your phone and think.

On the software engineering front, I think the way that AI has supercharged my ability to create the things I want is incredible. I am empowered by the frictionless workflow of dreaming up something to creating the digital manifestation of that thing within hours. I have found, however, that this workflow teeters dangerously on the cusp of creating absolute garbage. My experience with agents is that a single ill-crafted fix in a program leads to an endless loop of worse fixes piled on top, which in turn results in a haphazard, 8,000 line diffyou could probably blame this on me and my lack of agent-loop prowess, but I did in fact experience this once on something that could be fixed by reading one line of documentation.

On the content front, though, I think this bite-size, multiple-times-a-day, repeat-the-thing-that-goes-viral-thoughtlessly approach is pure tragedy. LinkedIn is a pit of utter despair and I find Twitter and Substack only marginally better. Ultimately, though, I’ve experienced this most acutely in my desire to grow a following on TikTokwhich has arisen because I realized that influencers, even the micro ones, get a lot of free clothes. I love well-made fashion-centered video content and well-made video content in general. My favorite YouTuber of all time is bestdressed, who I think changed both my 14-year-old-self’s brain chemistry as well as the entire fashion YouTuber’s recipe for making great videos. But having studied this algorithm, having watched the videos that actually go viral, it’s become quite clear to me that if your goal is to reach a large audience, quantity of quality is how you win. There is very little benefit in captions or artful editing. Take your phone out, talk into it three or four times a day, and hit post. Watch what goes viral and do it again and again and again. It’s not about making something that you find artistic or enjoyable, the point is make content that is consumable, that draws a user in for all of the 10000 milliseconds in your 10 second video. This formula is not rocket science. It is also, at least for me, not that rewardingbut I am nevertheless trying to conform because I really want free clothes.

The result of this algorithm on the content front is that even creators who I think make genuinely really excellent content must supplement their high-effort posts with a bunch of low-effort fodder. A single long form video must be clipped 900 different ways and tested from every angle for virality. Great journalists promote their articles by breaking it into little chunks of quotes in a thread, many writers of great essays do the same. I would certainly argue that there is some benefit into making content more digestible, especially technical content such as academic research (I love the Hazy Research blog posts!). But for the most part, to me, it serves as a great dilution of a piece of work that was crafted with intention. It rather sours the entire experience of reading or watching the original thing.

My only contribution in this essayrant is to say that while I see merit in following the formula, I don’t think we should cheapen the value of high-quality content by deluding ourselves into thinking it can be produced at scale. I have spent a lot of time around a lot of really smart people, I have had the incredible privilege of seeing people who are truly exceptional students, researchers, engineers, and artists at work. What produces great work is nothing short of obsession. It is coming back to the same thing over and over and over, fixating on every line of code until it makes sense, rewriting every word in an essay until it flows like water, reworking every seam in a dress until each stitch is invisible.

Amidst this mass production of consumption of thoughtless garbage, I find it helpful to remember that every piece of content that has ever influenced me, every relationship that I have built that mattered, and every thing that I have made that I am proud of took time. There is beauty, I think, in creating things that cannot be scaled. There is beauty in some things taking time.