The Traverse

The sky was clear. It had been clear for three days. The surface was firm under the sledges, and the dogs were pulling well. He could see for twenty miles in any direction, maybe more. No weather coming. No reason to stop.

They had covered their miles by early afternoon. He called the halt at the same place he always called it.

One of the men looked at the sky, then back at the leader. He said nothing. The others began making camp. The tents went up in the order they always went up. The dogs were fed and staked out. The stove was lit.

The man who had looked at the sky did the work alongside the others. He understood the distances. He understood what a day like this was worth. On a day like this, you could go thirty miles. You could go forty. The window wouldn’t last.

The leader had stopped anyway.

Twenty miles

That was Roald Amundsen, November 1911, leading the first expedition to the South Pole.

Robert Falcon Scott left for the pole around the same time. He had comparable resources, a capable team, and the same destination. The two men were running parallel races across the same continent, in the same season, toward the same fixed point on a map.

Amundsen reached the South Pole on December 14, 1911. He returned safely. Scott reached it on January 17, 1912, five weeks behind, and died on the return journey with the rest of his polar party.

The difference in outcome was not capability. Both men were serious explorers. Both worked hard. The difference was how each structured his movement.

Amundsen set a daily target and held it. His team marched between 15 and 20 miles per day, no more and no less. On good days, when conditions invited more, he stopped at the upper bound. On bad days, when every instinct said to shelter and wait, he still moved. Jim Collins writes that Amundsen “throttled back his well-tuned team to travel between 15 and 20 miles per day” regardless of what the weather offered. The consistency was the strategy.

Stopping on a good day is the harder discipline. The team had energy. The surface was firm. The weather would hold, or it seemed like it would. Calling the halt anyway felt like waste. But that restraint was not waste. It was the act of building reserves for days that hadn’t arrived yet. The logic is only visible in retrospect: what you hold back on the good days is what keeps you moving on the bad ones.

Scott’s team pushed hard when conditions allowed and sheltered when they didn’t. They moved in surges, burning deep on clear days and losing ground on difficult ones. The cumulative rhythm never steadied. By the time the deficit showed, it was structural, not recoverable with one strong push. They fell behind in ways that weren’t visible until the gap was irreversible.

The failure was not a failure of effort. It was a failure of design. Scott built his plan around the conditions he hoped for. Amundsen prepared for the conditions he feared.

The Stockdale Paradox names something similar. Admiral James Stockdale spent seven years as a prisoner of war in Vietnam, and the prisoners who broke were not the pessimists. They were the optimists who had staked their hope on a specific date, Christmas or Easter or the next exchange, and collapsed when those dates passed. That is a different thing from belief in a forecast: unwavering about the outcome, clear-eyed about current conditions.

You don’t need to know how this ends to know how to move through it. You need practices that hold regardless of conditions.

What does any of this have to do with your team, your work, and the specific uncertainty you are sitting inside right now?

The manager in the middle

This is the decision being asked of you right now. Not about polar expeditions. About your team, your roadmap, the quarter you’re already behind on, and the pressure coming from two directions at once.

From above: a business watching competitors move faster. Leadership that has seen the earnings reports, the analyst upgrades, the case studies. Meta’s “Year of Efficiency” cut roughly 21,000 roles in 2023 and was followed by a significant stock recovery and margin expansion. The data is real. The pressure derived from it is real. You are being asked to explain why you aren’t moving at that pace.

From below: a team that already knows it is stretched. Engineers doing the work of more engineers, shipping faster with fewer people, and watching the gap between what’s asked and what’s sustainable widen every sprint. They are not complaining about it loudly. That silence is not the same as capacity.

The data cuts both ways. Klarna replaced roughly 700 customer service workers with AI, saw customer satisfaction on complex interactions deteriorate, and began rehiring. What looked like a clean efficiency gain turned out to be a slower-moving cost that didn’t show up in the initial numbers.

You are not failing to lead. You are standing at the structural center of a genuine tension. It has a name.

The right data, the wrong frame

The business is not misreading the numbers. Meta cut 21,000 roles and the stock recovered. Klarna replaced 700 workers and costs dropped. Those outcomes are documented. Dismissing them is not a credibility move; it is a credibility loss. The data is real.

The problem is the frame used to interpret it.

Most organizations are running expedition logic on a problem that requires traverse logic.

An expedition ends. You identify a destination, plan the route, push hard toward the summit, recover at base camp, and go home. The push is bounded. The hardship is temporary. You can spend reserves because you know when resupply comes.

A traverse does not end. There is no summit. There is no base camp coming. The terrain keeps changing for as long as the terrain exists, and the only option is to keep moving through it. Spending reserves now is spending reserves you will need later, on conditions you cannot yet name.

The AI transition is a traverse. It has no finish line. There is no quarter where you arrive, declare efficiency, and recover. The landscape will shift again before the current shift has settled.

Extracting maximum output for three quarters looks identical to sustainable growth until the lag catches up. The debt accumulates in team capacity, institutional knowledge, and morale. It is invisible until it is acute. By then, the frame that caused it has already been rewarded.

The frame is wrong. The intention behind it is not. The job is to introduce the right one.

Two coherent responses, both wrong

Two responses to this pressure show up everywhere, and both of them make sense on their own terms.

The first is the all-in maximalist. Move fast. Adopt aggressively. Bet on the wave and optimize for the speed the wave rewards. This is not reckless. The data supports it. Early movers have posted real gains. The engineers who got fluent with AI tooling early are shipping faster. The teams that restructured around AI-assisted workflows cut costs while holding output. If the trend holds, hesitation is the expensive choice. The maximalist reads the room correctly and acts accordingly.

The second is the frozen skeptic. Hold back. Let the terrain settle before committing. Protect the team from the noise of tools that will look different in six months. This is also not irrational. The tooling landscape is changing fast enough that last year’s stack recommendation is already dated. Over-committing to today’s infrastructure has a real cost. Waiting for the picture to clarify before making irreversible decisions is a defensible form of discipline.

Both responses fail for the same reason. They anchor certainty to a prediction.

The maximalist’s plan depends on AI capability maturing on a specific timeline. If the timeline slips, if the gains plateau, if the regulatory landscape shifts, the whole structure rests on an assumption that hasn’t been tested yet. The skeptic’s plan depends on the terrain settling enough to evaluate clearly. If the pace of change doesn’t slow down, the waiting position becomes permanent. Clarity never arrives. The bet on eventual stability is still a bet.

This is Scott’s mistake and its mirror. Scott’s optimism was not irrational. It was load-bearing. His plan required the conditions he expected. When the conditions didn’t cooperate, there was no fallback, because the plan hadn’t been built for anything else. The frozen skeptic makes the same structural error in the opposite direction. Waiting for better conditions is itself a forecast. The conditions may not cooperate.

The failure mode is not making the wrong bet. It is anchoring to a bet at all, instead of building practices that survive any outcome.

What stays stable

There is a third position. It is not the maximalist’s conviction that the wave rewards the fastest movers, and not the skeptic’s bet that clarity is coming if you wait long enough. It holds genuine belief that better tools expand human capability over time. It also confronts, every day, that this sprint is too aggressive, this output needs review, this person is running on fumes. Strategically hopeful. Tactically realistic. Both at once, not in alternation.

The tools will keep changing. What doesn’t change is the obligation underneath them.

What the team owes users doesn’t shift because the tooling did. What good work looks like, how you measure honestly, how you protect trust: none of that is determined by what’s in your stack. Values are the anchor. They are not the roadmap.

Some things deserve to be stated plainly.

“This changes everything” is almost always wrong. So is “nothing will change.” Hold the harder position anyway.

Hype is a tax on understanding. You pay it in attention and decisions. Every sprint organized around an untested forecast, every headcount decision made to match a competitor’s press release, every tool adopted to signal agility rather than solve a problem: these are payments on a debt that hype created. Evaluate what’s in front of you, not what the moment is telling you to believe.

Build with your users, not for an abstract vision of progress. The users tell you what’s holding and what isn’t. The abstract vision doesn’t.

You don’t need certainty about where AI lands to know what to do tomorrow. The practices that hold in any weather are the ones worth building now. Build those.

What this actually requires

Pushback upward. When the business is running expedition logic, name it. Not as resistance, but as the accurate frame. The argument is not “slow down.” It is “this is a traverse, and traverses have different rules.” That takes political courage: willingness to be wrong about the timeline while being right about the principle. The argument is specific: the current plan assumes expedition conditions. Pace, review cycles, headcount decisions, and onboarding cannot all be optimized at once. Compressing all four simultaneously means something fails quietly, and the manager is usually the last to see it formally and the first to see it actually. Most organizations will not welcome this framing on first contact. Make it anyway.

Listening to the actual weather. Not the status update. The real answer when you ask. Sustainable cadence requires knowing true conditions, and true conditions require an environment where people can tell the truth. If the team reports fine when they are not, the manager doesn’t have the information needed to pace correctly. Creating the conditions for honest answers is part of the job, not a soft benefit of good culture. That condition is whether a team member can say in a 1:1 that they are drowning without it touching their performance review or their standing on the team. No process substitutes for it. BCG’s research found that AI-intensive work is producing cognitive overload symptoms at measurable rates: difficulty focusing, slower decision-making, degraded output quality. The risk is highest among high performers, the people least likely to volunteer that they are struggling. You have to ask well, and you have to hear what comes back.

Stopping at 20 miles. On good quarters, when the team has momentum and delivery is clean and the business wants more, the discipline is still the discipline. This is the harder enforcement. Over-extending when conditions are favorable is exactly what leaves the team depleted when they turn. Amundsen understood this that morning on the ice. In practice it looks like turning down scope on a sprint that already went well, pushing back on a feature request when the team just shipped something large, or protecting a recovery week after a hard push. None of those decisions feel urgent in the moment. All of them matter when conditions turn.

A team doing the work of ten with AI compression may look like a productivity win. It may also be a team one quarter away from breaking. The difference is invisible from the outside. The manager is the only person positioned to see it.

The traverse ahead

You don’t need certainty about where AI lands. You need clarity about what you stand for while it’s landing.

The teams that arrive aren’t the ones that predicted correctly. They’re the ones whose practices held. The prediction and the outcome diverged, and the practices held anyway. That is the difference.

The long arc bends toward more capability for more people. That belief is worth holding. It doesn’t tell you what to ship this week. It doesn’t tell you how to pace this quarter, or how to protect the person beside you from burning out before the terrain gets easier.

That’s the work. It doesn’t end. Build for it.

The destination is real. The path is uncertain. The practices are yours to set.