Nestor G Pestelos Jr · Writing

Positioning Doesn't Escape Prediction, It Demotes It

Published August 14, 2026

TL;DR

Positioning over predicting is right, but incomplete. It swaps a guess about outcomes for a guess about timing. Calibrated forecasting is the real exception; timing judgment is the real cost.

In 1980, Tom West's team at Data General was building a 32-bit minicomputer called Eagle. Digital Equipment Corporation had already shipped its VAX, and the trade press was calling it a breakthrough. West's engineers read about DEC's win with nothing of their own to show. By the numbers, they were late.

West didn't think so. His read wasn't about the machine. It was about the customers: whether their dissatisfaction still outweighed their switching costs, or they'd already committed past that point. Tracy Kidder, who spent a year embedded with the team, captured that read in The Soul of a New Machine: "You did not have to be the first company to produce the new kind of machine; sometimes, in fact, it was better not to be the first. But you had to produce yours before the new market really opened up and customers had made other marriages." West bet Eagle could ship before the marriages happened. He was right. Being first didn't matter. Guessing the deadline did.

That story is usually filed under "positioning beats predicting": build the capability, ship into the open window, don't try to call the market's winner in advance. I went through my notes gathering everything I had that supports that idea, then spent a session trying to break it. West's story is where the crack shows: he was still guessing, just about something narrower, when the window would shut rather than which machine would win. Positioning changed what he was guessing about.

The strongest version of "stop predicting, start positioning" comes from a tweet by Shane Parrish: predictions fail often and fail badly, so invest in response capability instead. It's the right instinct pointed at the wrong verb. W. Ross Ashby gave the formal version decades earlier in An Introduction to Cybernetics (1956): a control system needs at least as much variety as the environment it's trying to manage, or it can't manage it. Build enough response range and any single wrong prediction stops being able to sink you.

"It is far easier to figure out if something is fragile than to predict the occurrence of an event that may harm it," Nassim Taleb writes in Antifragile (via Farnam Street's synthesis), sharpening the same point from the other direction. Stop asking what will hurt you. Ask what's already exposed. That's a real substitution, but it still requires a judgment call about what would break if the hit landed, a different guess from what the hit will be.

Career strategy runs the same pattern. Buying broad, transferable skills early is "basically buying yourself optionality," argues Abhishek Singh, not because breadth is safe, but because strong fundamentals are what make a new domain fast to learn. Trading runs it too: the discipline isn't refusing to have a view, it's replacing "the price will go up" with "if the price does X, I will do Y." Still a bet, just a conditional one. Organizations do it at scale. Trend reports are usually wrong, and BusinessWorld's Reynaldo Lugtu argues that's not the point: reading them seriously builds the reflex to move fast once something real happens. My extension of his point: that reflex only works if you've already guessed, early and often, which kind of change would matter.

Two problems survived that an earlier piece of mine hadn't already covered.

First: calibrated forecasting isn't fragile. State a confidence level alongside a prediction, get scored on it over time, and you get a trackable, improvable skill, not a coin flip. That's the finding behind decades of Philip Tetlock's tournament data in Superforecasting, confirmed in 2020 when Atanasov and colleagues found the accurate forecasters updated often and in small steps. The low-skill pattern was defending an initial judgment, or making rare, large revisions instead. "Predictions fail often and badly" describes the uncalibrated, unaccountable kind, not this one.

Second: positioning has a boundary, not a blank check, and the West story has a sharper version of its own limit. When a firm's own exploring is what generates its future options, rather than the market resolving uncertainty on its own, the flexibility gained by staying open erodes the clean walk-away option that made "staying flexible" valuable in the first place. That's Ron Adner and Daniel Levinthal's 2004 finding in the Academy of Management Review. Optionality isn't free. Past a point, chasing more of it costs you the very flexibility you were chasing.

"Stop predicting" was never quite right. "Predict less, and predict something narrower" is what survives.

That narrower target connects back to a different note in my own vault, one I almost filed as a contradiction: Mustafa's summary of Cecily Sommers's Think Like a Futurist, that the future is driven by slow underlying forces (demographics, capital, institutions), not the fast trends sitting on top of them, and that the real skill is tracking those forces rather than betting on any one trend. I don't think that contradicts positioning. I think it's the same move at a longer time horizon: you're not predicting an outcome, you're building the attentional habit to notice when several slow forces have already narrowed the field to one option. It's West's timing bet, run at the scale of a decade instead of a product cycle.

What changes, in practice:

Parrish's line, stop trying to guess what happens next and become who can handle it either way, is still the right instinct, just missing a clause. You never stop guessing. You just get more disciplined about which guess you're making.