Conclusions of my DIY Inkhaven

Epistemic status: describing my own experiment results.

AI use: review, research (stats), fact-checking, grammar & light phrasing.

I applied to the next Inkhaven cohort and was rejected. This was surprising and very unfortunate. While writing my application and answering why they should accept me, I felt like I'd one-shotted myself into writing. My goals were:

• Run an experiment to verify my hypothesis that I couldn't write well. I also implicitly assumed that I didn't have anything useful to say or to add, so no one would read what I write anyway.
• I wanted to get better at thinking, and to me writing is a better form of thinking.
• I wanted to use language more eloquently even outside of writing.
• Writing is meditative and deep work, so I wanted to become more productive by writing more.

I was sad to be rejected, and I still don't know why. Fortunately, I have my agency, so I just decided to do my own Inkhaven instead. Obviously, I wouldn't stay at Lighthaven with a bunch of other people or receive guidance from people like Scott Alexander and Aella. Also, all of this is happening as part of my normal life, and I have other stuff to do as well.

In reality, I started writing even before I submitted my application, because I needed samples of my writing and I wanted to check whether I actually enjoyed doing it. After submitting my application, it felt cool to just write until the start of the cohort. I could compare about 30 days of writing on my own with a month at the cohort with guidance. Also, I would have more writing to compare before and after. After the rejection, there was a temptation to just give up, but I decided to continue.

Including this post on October 31, my totals are:

• 21 posts on LessWrong: about 19,000 words and 109,000 characters.
• Around a dozen posts on my private blog: roughly 6,000 words and 30,000 characters.

The earlier 20 LessWrong posts have 740 karma across 441 votes and 95 comments as of October 31. These word and character counts exclude footnotes; characters include spaces. One additional LessWrong submission was automatically rejected after an LLM detector flagged it.

I started by writing on LessWrong about stuff I found interesting. I would just get some random idea that I found interesting enough to explore. I did try to optimize for more karma.

Writing 500 words for LW, assuming you want other people to like it, felt quite hard. I could obviously bumble on for 500 words pretty easily and just publish that. I still had to relax my requirements a bit - for example, I never checked whether someone had written on the topic before. I knew that if I did, I would feel like I had nothing to add, so I just wrote, ignoring the fact that this could be a duplicate.

My first hit was "A fascinating case of an LLM prompt-injecting its own user". I saw this a long time ago and was extremely impressed and fascinated. Originally I assumed that everyone had seen it too. Apparently, something didn't have to be new to me to be useful to someone else. But since I had to write something, I decided to write about this. This was fairly early in the process (day 5, October 6) and definitely helped me feel less like "everyone else is so cool, you have nothing to add or say, and no one will read whatever you write."

My post with the most karma is "Quantity vs quality in pottery is a lie", which as of now has 176 karma. I literally read that anecdote in a book. I feel like, since I was looking for ideas to write about, I got curious about their experimental setup. To my surprise, it wasn't a scientific study, but an anecdote about a photography class that had been rewritten as pottery. My conclusion from this was that people at LessWrong like becoming less wrong. But this is a hard niche to reliably produce content in.

Two other highly upvoted posts are "Gluten-free water" and "I can't believe it's not BUTTER!".

In both cases, I noticed something peculiar in my life, dug deeper into it, and found more examples of similar things.

My "worst" post was "When bureaucracy gives you something it shouldn't". To be honest, I was writing it late in the day in a hurry just to get something out. Originally, it was downvoted below my personal strong upvote, but with time it got to +1 above my own strong upvote. I have no clue why.

I didn't engage in the comments much, since it didn't count toward my original proxy goal of publishing 500 words a day.

At some point, I started running out of steam, and overall it felt cumbersome. Writing for LW is quite hard. I applied high standards, and it took time. Eventually I switched to my private blog, where I would cover more personal topics and the posts didn't have to be as rigorous.

This made it easier and more fun. I mostly wrote my opinions or stories from my life, and it was easy and pleasant to write. This didn't feel like work and was enjoyable. I suspect these topics are just closer to my heart than the topics I would be writing about on LW (but I now catch myself realizing that I actually have no clue whether that's true - perhaps my personal reflections would be enjoyed on LW too). I feel like I will likely continue writing on my private blog (e.g., I now often catch myself thinking that this situation or feeling is interesting and I want to write about it to preserve it or to signal to others) and on LW very occasionally when there is something really burning inside me that I want to write about.

In terms of my original goals

Run an experiment to verify my hypothesis that I couldn't write well.

"Well" is unclear. I've got 176 karma on a post, and my posts have around 700 karma between them (so that would be a $7 discount for the next LessOnline, if they keep the karma discount). So some people definitely liked what I wrote. My original prediction was that all my posts would be around 0 karma and basically no one would interact with them. This was totally wrong - there were a bunch of comments.

I also implicitly assumed that I didn't have anything useful to say or to add, so no one would read what I write anyway.

This is definitely false. There was even one commenter who explicitly told me how much they enjoyed my writing (by the way, thank you, August Morley, for taking the time to write that comment).

I wanted to get better at thinking, and to me writing is a better form of thinking.

Partially achieved, subjectively. I did notice that sometimes I would start writing and be extremely unclear about the direction or outcome, but in the process converge on something. Overall, writing as thinking feels useful, but I haven't properly tried it for personal topics that I don't want to publish. After my DIY Inkhaven, I can imagine doing that.

I wanted to use language more eloquently even outside of writing.

I didn't feel any impact here. I did use AI quite a lot for grammar. Overall, not worrying much about grammar or sentence structure helped me greatly with writing. I.e., I could write in a flow of consciousness. If I try to polish the phrasing too much, I get stuck.

Writing is meditative and deep work, so I wanted to become more productive by writing more.

Subjectively, this seems to have worked. I've started doing much more deep, productive work unrelated to writing. I felt like, with AI, I had stopped having flow at work. Writing gave me flow back, and that somehow spilled into other areas.

It felt impossible to predict what people would like. I also got the impression that posting early in the US workday seemed to work better (catching myself thinking that this is an idea for a post - analyze whether the time of posting affects a post's performance).

I am extremely curious to try applying this approach of N days of something to other areas of my life (perhaps drawing, music or something business-oriented, but it is very hard to come up with good proxy rules like 500 words a day). Also, I wouldn't do 30 days from the get-go. I would start with 2-7 days to test whether I like doing the thing. Even with writing, after 20 days it became cumbersome and hard.

Overall, I am happy that I've done it, and it feels net positive to me subjectively.