Callum

the world is magic if you pay attention

How to capture attention: Fifteen years ago when we were building RethinkDB it was on top of Hacker News all the time. To this day ppl remember this and maybe once a month someone asks me for help with launching their project. It worked like this. I would think of an HN headline first, then we'd build the next product version to do whatever the headline promised. We'd submit to HN and it would either rise to the top or it wouldn't. Then we'd do it again. This loop taught me a few things which I use to this day: 1. It RL'ed me to generate good headlines. It's a skill you can develop through trial and error. I can maybe distill it to some rules of thumb, but a lot of it is, does it sound right? Is the rhythm right? Does the headline “hit the spot”? Not easy to explain. 2. I became headline-pilled. The world runs on headlines. People will grant you maybe 15 tokens of attention. Unless they feel there is some irreducible essence captured in these 15 tokens, they’ll move onto something else. This is true of your boss, your investors, your colleagues, your employees, your customers, social media audiences, even old acquaintances you randomly bump into. “So… what’s been going on in your life?” You only have 15 tokens, i.e. a headline, to answer this question. 3. Unless you already think in headlines, you can't build something and come up with a good headline post-factum. It just doesn't work out well. If you do that, either the headline will be true but boring, or it will be enticing but not credible. You've got to come up with a headline first and build towards it. 4. Working this way prevents feature creep. You build the minimum feature set necessary to make the headline credible. If smtg doesn't serve the headline, it's out. Once you have enough to make the headline credible, you ship. And headlines are by definition short, so your development cycle becomes very nimble and efficient. 5. It makes time-boxing easy. "Headline goes out by Sep 15th (or whatever), we gotta build enough to credibly share it with the world. Ready, set, go." This preempts all the debate about sprints, estimates, methodologies, prioritization, and so on. You just loop: headline -> build -> ship, headline -> build -> ship. It makes the development process really smooth. In the age of Codex when building is cheap, this becomes even _more_ relevant. People are DDoS'ed with information, and it will only get worse. So a headline is all you get. You can make much more ambitious headlines come true now, but you still only get 15 tokens before ppl scroll past you. Btw, once you master this idea a whole world opens up. You can play with violating these rules in a hundred ways and still capture attention. But not easy to do that until you've mastered thinking in headlines first.

Related: the tokenomics of coordination lesswrong.com/posts/4ZvJab25…

The idea is simply to find the smallest set of assumptions you need to get from people reaching for stuff to an actual system of signs, and prove each step in Lean, in the spirit of Gans' quote: "Mysteries need not be multiplied beyond necessity."

Ok, for the dozen people following me for NERD stuff, here's my opinion on why aiming for antifragility is bad: brianlui.dog/antifragilebad.…

Here's a few idiomatic prompts I keep reaching for: callum.website/prompt-framin…