Thirty violin students at the Music Academy of West Berlin kept a diary for one week. Ten had been named by their professors as possible international soloists, ten were good players from the same department, and ten came from the music education program.
The psychologists K. Anders Ericsson, Ralf Krampe and Clemens Tesch-Römer expected the stronger players to practice more, as their theory predicted, and the diaries agreed. The two top groups practiced alone about 3.5 hours a day, weekends included. The music education students practiced about 1.3.
But the diaries also showed how. The practice sessions averaged 80 minutes in every group. The top students didn’t practice longer at a stretch; they practiced more often, with a peak between 10 a.m. and 2 p.m. And they slept more: 8.6 hours a day, against 7.8. Counting daytime sleep, they napped 2.8 hours a week. The music education students napped 0.9.
The hours and the dose
Ericsson and his colleagues read the diaries as a limit. Deliberate practice, they wrote, “is an effortful activity that can be sustained only for a limited time each day during extended periods without leading to exhaustion.” The training studies they reviewed found essentially no benefit from practicing more than four hours a day.
The paper became famous for its hours anyway. By 18, the best violinists had accumulated an average of 7,410 hours of practice. Malcolm Gladwell’s 2008 book Outliers later made a “10,000-hour rule” famous. The naps didn’t travel nearly as far.
When you can’t add hours, what you can still manage is the energy you bring to them.
The consultants’ version
Tony Schwartz and Catherine McCarthy turned a version of that sentence into a title. In “Manage Your Energy, Not Your Time,” in Harvard Business Review in October 2007, they wrote: “The core problem with working longer hours is that time is a finite resource. Energy is a different story.” Energy, they argued, comes from body, emotions, mind and spirit, and each can be expanded and renewed through rituals: behaviors practiced at set times until they become automatic.
Their opening example, an Ernst & Young partner working 12- to 14-hour days, set an earlier bedtime, gave up drinking and took walks. He still put in long hours, they wrote, but renewed himself along the way. The approach builds, they said, on Schwartz’s former partner Jim Loehr’s work with athletes; the two had published The Power of Full Engagement in 2003.
The four sources are a framework, not a finding. Its showcase result came from a pilot at Wachovia Bank run by the authors’ own firm, the Energy Project. By their account, participants outperformed a control group on financial measures such as the value of loans they generated.
What the research supports
Independent research is more modest. A 2022 meta-analysis pooled 22 study samples, about 2,300 people, testing breaks of ten minutes or less. The breaks raised vigor and lowered fatigue by small but statistically significant amounts. Their effect on overall performance was smaller still and not statistically significant; it showed up only for less demanding tasks, and longer breaks helped more.
The fuel-tank picture also recalls ego depletion, the claim that self-control uses up a limited supply. In a preregistered test across 36 labs and 3,531 people published in 2021, the effect came out small and not statistically significant. And the violinists’ sleep figures come from one diary week. When Brooke Macnamara and Megha Maitra repeated the study in 2019, its core claim about accumulated practice did not replicate in full. In the replication’s three-group analysis, practice explained 26 percent of performance differences; the original paper, which combined the two stronger groups, reported 48 percent. The two figures come from different groupings.
So energy as a resource you can measure and grow on schedule is a metaphor. The weaker version holds up better: hard concentration comes in limited doses, and rest changes how you feel.
The strategy paper written at night
Imagine a product manager whose mornings fill with status meetings. The strategy paper that will set her team’s next year gets written after dinner, two tired hours at a time. It is on her calendar every week, so on paper she has made room for it.
The violinists point to a different question: which hours. Moving the meetings to the afternoon and giving the paper two 80-minute blocks before lunch costs no time at all. Whether the paper improves, she can judge from drafts written in both kinds of hours.
The colleague who doesn’t nap
Here the conclusion flips. Nothing in an AI agent’s working day corresponds to the violinists’ afternoon nap. For the agent, time is the thing to budget again: how long it runs, and how many copies run at once.
The energy that runs out is the human’s. In HBR in February 2026, Aruna Ranganathan and Xingqi Maggie Ye of Berkeley’s Haas School of Business reported initial findings from a study of 200 employees at a U.S. technology company, run from April to December 2025. As the blogger Simon Willison summarized and quoted them, workers ran “multiple agents in parallel” and revived long-deferred tasks, and the reality was “a continual switching of attention, frequent checking of AI outputs, and a growing number of open tasks.” Willison added his own experience: after an hour or two of this, “my mental energy for the day feels almost entirely depleted.”
So schedule the agent like a machine, and yourself like a violinist. Count your open agent threads at the end of each day. Review in fixed sessions instead of whenever a run finishes. Then test your own fatigue: the next morning, reread one output you approved late the day before and ask whether you’d approve it again. Do it for a week. If late approvals keep failing the morning read, that’s a lead worth acting on.
In Berlin, the top students practiced most before lunch and slept more in the afternoon. The agents in the next window can run straight through the night, and in the morning someone still has to read what they made.