in this post on how I learned to program, I originally had the dates wrong on when the competition programmers from my high school made money (it was a couple years after I thought it was). In that case, and many others, I don't think that the date being wrong changes anything significant about the post.
Although I'm publishing the original version of this in 2021, I expect this list to grow over time. I hope that I've become more careful and that the list will grow more slowly in the future than it has in the past, but that remains to be seen. I view it as a good sign that a large fraction of the list is from my first three months of blogging, in 2013, but that's no reason to get complacent!
I've added a classification below that's how I think of the errors, but that classification is also arbitrary and the categories aren't even mutually exclusive. If I ever collect enough of these that it's difficult to hold them all in my head at once, I might create a tag system and use that to classify them instead, but I hope to not accumulate so many major errors that I feel like I need a tag system for readers to easily peruse them.
- Insufficient thought
- 2013: Using random algorithms to decrease the probability that good stories get "unlucky" on HN: this idea was tried and didn't work well as well as putting humans in the loop who decide which stories should be rescued from oblivion.
- Since this was a proposal and not a claim, this technically wasn't an error since I didn't claim that this would definitely work, but my feeling is that I should've also considered solutions that put humans in the loop. I didn't because Digg famously got a lot of backlash for having humans influence their front page but, in retrospect, we can see that it's possible to do so in a way that doesn't generate backlash that effective kills the site and I think this could've been predicted with enough thought
- 2013: The institution knowledge and culture that create excellence can take a long time to build up: At this time, I hadn't worked in software and that thought that this wasn't as difficult for software because so many software companies are successful with new/young teams. But, in retrospect, the difference isn't that those companies don't produce bad (unreliable, buggy, slow, etc.) software, it's that product/market fit and network effects are important enough that it frequently doesn't matter that software is bad
- 2015: In this post on how people don't read citations, I found it mysterious that type system advocates would cite non-existent strong evidence, which seems unlike the other examples, where people pass on a clever, contrarian, result without ever having read it. The thing I thought was mysterious was that, unlike the other examples, there isn't an incorrect piece of evidence being passed around; the assertion that there is evidence is disconnected from any evidence, even misinterpreted evidence. In retrospect, I was being naive in thinking that there was a link to evidence that people wouldn't just fabricate the idea that there is evidence supporting their belief and then pass that around.
- Insufficient verification of information
- 2016: Building a search engine isn't trivial: although I think the overall point is true, one of the pieces of evidence I relied on came out of using numbers that someone who worked on a search engine told me about. But when I measured actual numbers, I found that the numbers I was told were off by multiple orders of magnitude
- Pointlessly difficult to understand explanation
- Errors in retrospect
- 2015: Blog monetization: I grossly underestimated how much I could make on Patreon by looking at how much Casey Muratori, Eric Raymond, and eevee were making on Patreon at the time. I thought that all three of them would out-earn me based for a variety of reasons and that was incorrect. A major reason that was incorrect was that boring, long-form, writing monetizes much better than I exepected, which means that I monetarily undervalued that compared to what other tech folks are doing.
- A couple weeks ago, I added a link to Patreon at the top of posts (instead of just having one hidding at the bottom) and mentioned having a Patreon on Twitter. Since then, my earnings have increased by about as much as Eric Raymond makes in total and the amount seems to be increasing at a decent rate, which is a result I wouldn't have expected before the rise of substack. But anyone who realized how well individual writers can monetize their writing could've created substack and no one did until Chris Best, Hamish McKenzie, and Jairaj Sethi created substack, so I'd say that this one was somewhat non-obvious. Also, it's unclear if the monetization is going to scale up or will plateau; if it plateaus, then my guess would only be off by a small constant factor.
Thanks to Anja Boskovic and Ville Sundberg for comments/corrections/discussion.