Published: February 18, 2026
This article is intentionally mediocre. So is this entire blog in some way (see How you know this is not AI), although I have to laud the quality of the articles my co-author Leif has put out since we started this project over winter break. I intentionally write we (and I had to look up how to use italic in HTML), because in the two months since we made the definite plan of starting, I have not published a single piece. Every so often, I would share some idea with Leif and he would suggest I write a blog post about it. Most times, I did immediately start writing until I got stuck on some sentence or idea and buried the post in my drafts. In some way, this is just a manifestation of the perfectionism I have always had. In another way, I think this is also a byproduct of LLMs. With LLMs, the cost of generating coherent, grammatically-correct, and arguably well-informed text converges to zero. Mentally, this raises the bar for what I consider publishable, both from an internal and external perspective.
Internal, because I expect my writing to be sufficiently coherent and grammatically correct. A non-native English writer, I used to merely look up the spelling and meaning of words and grammatical constructions to make sure I was using them correctly and that was sufficient. Now, I catch myself benchmarking every sentence against what an LLM could have writtern faster and with fewer dangling clauses. I ask: Is this better than what a machine could have produced in 3 seconds?
External, because it is now possible to scrutinize at scale. A horde of agents is just waiting to take this article apart. The imagined audience used to be a small group of readers who stumbled upon this niche blog. Now the audience is the infinite reviewer. Of course, on the internet, everything was always discoverable in principle. But in practice, at any given moment, only a handful of people would encounter most writing. And over time, it would be buried beneath newer words, drifting out of attention because attention was finite. Now, text may go unread, but never unprocessed. It is indexed, embedded, compared. The crowd has grown.
But writing was never supposed to be about ultimate correctness. It should be about thinking in public. Getting an e-mail from a reader about a factual mistake in your article. The serendipity of the thought and writing process. Like when I looked up the word "arguably" because I was not sure if I was using it correctly earlier and I found this cool blog.
If I wait until I can outperform the statistical ghost on every paragraph, I will never write again. The asymptote moves too quickly. So perhaps the only viable rebellion is to lower the bar publicly. To allow posts that are 70% formed. To publish thoughts that might be wrong.
So this article is intentionally mediocre.
← Back to main.