Sep 24, 2026
I started posting in the current format on Jul 26. In less than two months account went from ~500 to 19k followers. Some notes on what I set out to do (no idea if it will keep working, but so far worked out pretty well):
- my plan was to completely ignore likes and maximize bookmark count on posts. That implies writing things people find useful and will want to go back to at some point (even though realistically they probably won't)
- also I decided to ignore follower count completely and optimize reach for "high value people". By high value I don't mean high follower count, or famous, or important. Rather it's competent people doing (or hoping to do) interesting, difficult, or valuable things. So for example phd students studying difficult fields, early startup founders trying to build magical products, engineers working at hypergrowth companies trying to keep the ship from exploding, managers who need to build high output teams/orgs very quickly, young ambitious ppl trying to figure out what to do, etc.
- for example, if a tweet gets three bookmarks and seems to have reached just one of these people (from looking at bios from likes) I consider it a success
- this is hard btw, as X's entire design is organized to push you in the "sugar water" direction-- having short, surprising, or controversial takes that will get you high like count. I tried to make browser extensions to hide likes, but could never get them to stick. So I force myself to follow above goals through discipline
- topic inspiration mostly comes from things I screwed up in the past that I would have done very differently today. This is a surprisingly good source for two reasons. First, it's virtually inexhaustible. Second, if these things were unobvious to me they're likely unobvious to others, so utility is baked in automatically
- there are topics I'm just interested in (e.g. what makes some ppl/orgs great, projects/experiments I'm working on, etc.) If I think they might be interesting to others, I post about those too
- I vibe coded https://t.co/dhtCMg32Jb to cache and index these posts (stole this idea from @melissa). X acts as a distribution platform, and the website acts as a place that aggregates the posts into a browsable body of knowledge
- I write each post so it can be added to this body in order to eventually form some coherent whole (though this is a very soft requirement). Also, since posting on X has much lower friction than writing a blog post, this solves the problem of a blog that never gets updated
- the format is meant to minimize writing and reading friction. Good essays are hard to write and can take weeks. They also take a considerable commitment from readers. Bullet points are easier to write and read. For better or worse social media reduced attention spans, so I decided to embrace the medium and give ppl bite-sized chunks
- my hard rule for AI is to only use it as a thesaurus. Claude is amazing at suggesting synonyms! Hunting for just the right word is hard so I'll often paste a sentence into Claude and tell it some word isn't quite right for reasons xyz and ask for alternatives
- bullet points mostly come as a stream of consciousness. When I get stuck I leave it alone, then come back to add more. Eventually no new ideas come to mind. I do a few passes for grammar, rhythm, flow, removing useless words and so on. I deliberately shoot for very informal messy tone to get with the times. (It's sort of like writing code in the style of surrounding codebase, back when people used to do such things)
- I deliberately end posts without a satisfying conclusion so they feel cut off. Montaigne did that in his essays and I always found it charming, so wanted to replicate that here. (Also, conclusions often feel satisfying but low signal, and I try to keep low signal stuff out)