Dear Magicians,

In 2010, I helped choose the site for a telescope in the Chilean Atacama Desert. The decision matrix had seventeen variables. Atmospheric opacity. Precipitable water vapor. Logistics costs. Stability of the host country. We debated for months.

You know what we didn’t model? The fact that the access road would wash out every three years during El Niño, stranding equipment worth more than my department’s annual budget on the wrong side of a mudslide.

That telescope is no longer there. But the road still washes out from time to time. Looking back, we spent more on road repair than we spent on parts of the telescope. And every time I think of the costs we paid over the life of the instrument for snowplowing, I think about the meeting in 2010 where someone said, “The road seems fine…it’s the driest desert on Earth!” and everyone nodded because we were tired and the more fun argument we wanted to get on to was about detector sensitivity. We did end up solving the road maintenance issues but those first few years sure were rough. But not for the cosmological reasons I would’ve naively thought were most important.

Path dependence is the term economists use for this. It sounds clinical. It isn’t.

Path dependence is the reason Japan runs two separate electrical grids — one at 50 hertz, one at 60 — because in 1895, Tokyo imported German generators and Osaka imported American ones.

Nobody standardized. By the time anyone noticed, both systems were too embedded to rip out. During the 2011 Tōhoku earthquake, power from western Japan couldn’t be rerouted east at scale. A purchasing decision made by people who’d been dead for decades was still shaping disaster response.

I see this everywhere now. In physics, we’re still publishing in journals designed for the postal system. The formatting requirements — those absurd margin specifications, the insistence on .eps figure files — exist because someone in 1960 optimized for a phototypesetting machine that hasn’t existed since the Reagan administration. Nobody remembers why. Nobody can change it, because the tenure committees still count publications in those journals, and the tenure committees are staffed by people who got tenure by publishing in those journals.

It’s generators all the way down.

The uncomfortable version of this insight is personal. The major decisions in your career, like the advisor you chose, the subfield you wandered into during your second year, the city you moved to for a postdoc because your partner had a job there…none of those were data driven and optimized. They were contingent. And now they’re load-bearing. You’ve built twenty years of infrastructure on top of choices you made when you were twenty-six and running on caffeine and the vague sense that you should probably say yes to the offer that came first.

I’m not saying those choices were wrong. I’m saying they were arbitrary in a way that we retroactively narrate as intentional. The access road seemed fine. The generators worked. The journal was prestigious. And by the time the cracks showed, the switching costs were astronomical. I’m reminded of another economics term, this time from coding — technical debt.

The antidote isn’t better planning. You can’t model the El Niño you haven’t met yet. The antidote is building in enough slack that when the road washes out — and it will — you can reroute without plowing your entire career to the other side of the mountain.

That’s what I wish someone had told me back then. Not “choose better.” Just “leave room to choose again”…or as some wise economist once said, “always take the option that gives you the most options”.

Until next time, have a debt-free, M.A.G.I.C. Week,

Brian