Skin in the game

My engineering manager likes to say that he can give advice and warn us before a big mistake, but he can't tell us what to do. He doesn't live with the consequences.

He's one of those ultra experienced engineers who will even go into management, if that's what it takes. Full of great advice borne of pain in production.

But it wasn't until I read Nassim Taleb's Skin in The Game that I understood the wisdom behind my manager's words. A person who doesn't feel the consequences cannot tell you what to do.

This is important. You have to maintain this code therefore you know best.

Don't tell me what's right, show me what's working

Investors (maybe just Taleb) have a saying: Don't tell me what you think, tell me what's in your portfolio.

You may think lots of things. You may have lots of ideas about what tech is good or bad. But only what you do matters. What are your revealed preferences?

A revealed preference is what you do in spite of what you say. The candy you eat when you don't want another bite. Because it's just so damn good. The mother-in-law's baking you don't eat despite saying how great she is at baking and how the burnt crust adds an innovative new dimension.

Taleb argues that survival is the only criteria.

You can theorize and argue for perfection all you want – did your code survive? That's it. The one criteria.

If the code works, it stays. If it's bad, someone will remove it. The longer it survives, the longer it's likely to keep surviving. Lindy effect.

Good advice comes from production

The best advice starts with "Here's what I've seen work in the past". Taleb even says that's the only advice. Anything less than that is crap.

He closes Skin in The Game with a wonderful quote:

When the beard is black, listen to the reasoning but drop the conclusion. When the beard is grey, listen to the reasoning and consider the conclusion. When the beard is white, skip the reasoning and listen to the conclusion.

Don't believe everything devrels and professional theorizers tell you about your code [name|]. Ask them what's in production.

Cheers,
~Swizec