It’s become a meme that every AI essay trying to Do SomethingTM in the world must launch with its own custom-built microsite, usually with a serif font, minimal design, high-minded title, and off-white background. Situational Awareness, AI 2027, Superintelligence Strategy, Gradual Disempowerment, The Intelligence Curse. Their designs each suggest that an ordinary Substack post would be too unserious—what really defines an idea’s quality is striking the perfect balance of “trying really hard to seem like I’m not trying hard,” like the “no-makeup makeup” of thinkboi blogs.

The more I look at these new sites, the more familiar they seem. The old-style serifs and thin rectangular boxes on AI 2027 reminded me of OG rat forum LessWrong. Several abundance-adjacent nonprofits all use the UK-based design studio And-Now. The website for automate-all-jobs startup Mechanize is a whole-cloth ripoff (or homage) of Ilya Sutskever’s notoriously hermetic SSI, and both are LARPing the IDGAF aesthetic tradition pioneered by Warren Buffett’s Berkshire Hathaway, which hasn’t redesigned its “WEB page” of purple hyperlinks since 1999. And while shades of orange aren’t particularly popular for decorating, well, most anything else, the palette has been a fixture of tenured-professor-sites and YC-adjacent hacker blogs since the mid-aughts.

These Einsteins would probably prefer us to think that they’re all substance and no style. The less CSS the better; not a gradient in sight. But their design choices are no accident—you can see an organization’s idealized intellectual forefathers in their muted colors and bookish fonts. It’s signaling all the way down.

The Cup of Coffee Theory of AI

Macrodoses #7