Antifragile Ideas

In this tech talk, John Carmack makes the following argument: a good idea is an idea is that thrives under scrutiny; so to have good ideas, have many ideas and test them quickly. “Antifragile” is a word coined by Nassim Taleb to refer to objects that thrive under stress, disorder, or randomness. Carmack calls this “antifragile idea generation”. Good ideas do not just survive scrutiny, which is what Taleb calls “robust”; good ideas are better after scrutiny.

I like this theory of idea generation for a few reasons. First, this theory aligns with what I’ve heard from a few famous scientists. In his talk You and Your Research, Richard Hamming said:

Most great scientists know many important problems. They have something between ten and twenty important problems for which they are looking for an attack. And when they see a new idea come up, one hears them say, “Well that bears on this problem.” They drop all the other things and get after it.

According Gian–Carlo Rota, this is the Feynman method:

Richard Feynman was fond of giving the following advice on how to be a genius. You have to keep a dozen of your favorite problems constantly present in your mind, although by and large they will lay in a dormant state. Every time you hear or read a new trick or a new result, test it against each of your twelve problems to see whether it helps. Every once in a while there will be a hit, and people will say, “How did he do it? He must be a genius!”

In my own experience, the most productive researchers I’ve met typically have lots of ideas while remaining uncommitted to any of them. In Carmack’s theory, these researchers are not necessarily having better ideas on average; rather, they are increasing the odds that they find a good idea. This reminds me of something Leonard Baum, of Baum–Welch fame, once said (Zuckerman, 2019):

Bad ideas is good, good ideas is terrific, no ideas is terrible.

As a caveat, note the distinction between problems and ideas. The problem is the goal, while the idea is the line of attack. So you can productively work on unimportant problems by applying Carmack’s theory. My interpretation is that Hamming and Feynman encourage applying Carmack’s theory against a set of pet important problems. Define “important” and “unimportant” however you’d like.

One of Taleb’s main takeaways from the concept of antifragility is to look for asymmetrical situations in which your the downside risk from random events is low or capped, while your potential upside is high. In this framing, the Feynman method is antifragile. You slowly chip away at important problems rather than spending 40 hours a week on them. This minimizes your downside risk. If you have a bad idea, who cares? You only spent a few minutes thinking through the implications. But if you get lucky, if the random thought does in fact solve the important problem, you have a big potential upside: you’ve solve a hard problem and are now labeled a genius.

This theory also aligns with how other creative people talk about idea generation. In college, I studied architecture, and our professors would often ask for three building designs by a particular deadline. This was hard because it was easy to overcommit to your first design. “Have three designs by Wednesday,” forced us to avoid this pitfall. This also feels related to the common writing advice of “kill your darlings”. Put differently, a major failure mode of ideation is getting stuck on one idea. Carmack gave the example of people who email him with a “big idea” that they haven’t actually tested or thought through fully. Carmack’s advice to such people might be: don’t overcommit; test your idea as quickly as possible; test alternative lines of attack.

As a corollary, we can estimate how good an idea might be by how testable it is. Carmack gave an example of an idea he had for a new way to weld rockets. Since he no longer builds rockets, he can’t test this idea, and this means the idea is worse than other ideas in this sense.

I suspect this theory—that “good” just means antifragile so focus on testable throughput—has other uses. For example, in this Planet Money episode, Lisa Chow’s strategy for dating was to follow a few rules: her first message was always just the winking face emoji; the date needed to happen as quickly as possible after first contact; and dates couldn’t be scheduled for weekends. Why? She was lowering the transaction cost of each date and avoiding the disappointment of overcomitting without enough information. Maybe we could call this “antifragile dating”, where the goal is to not go on good dates but to eliminate bad dates as quickly as possible.

Finally, I like this theory because it converts the abstract notion of “having ideas” into a process. I like process because it changes your focus from goals, which are often long-term and open-ended, to daily habits. Whenever I’m stuck in my research, my new process is to write down as many testable ideas as I can. And the process is freeing because ideas do not have to be good, since “good” is a property of antifragility that can only arise out of testing.

As a concluding caveat, I typically dislike advice on productivity habits, especially when the advice-giver has only recently discovered the habit themselves. There are too many false summits in this area. However, I’m publishing these thoughts as an exercise: writing this post was another test of Carmack’s theory of antifragile idea generation, and I like the idea the more that I test it.