Show Notes
Every hour anyone has put into getting good at something — all of them, added up, across every field studied properly — explains about twelve percent of how good people end up. Five percent in the studies that measured practice properly. This one turns on the sailor's word the show is named after: the one thing it gets right about that, and the two things it gets badly wrong.
More from fetch
"Don't treat yourself as the main character" is half right. Talking to yourself as "you", by your own name, cuts shame before a stressful speech and makes people reason about their own problems as wisely as they do about a friend's. But "don't be so sensitive" is distraction, and distraction rebounds a day or a week later. Hana argues the other side: people who see their life as a hero's journey find more meaning, and in therapy, a growing sense of agency comes before people get better. Gus brings the mirror, the video-call self-view tile and the feeling that a room is watching. It turns out both sides want the same thing. Be the narrator of your story, not its audience.
Does AI give you time back? On single tasks, often yes: in good studies, writing, customer support and summarising get 25 to 40 percent faster. Across a working week, people say they save a few percent, and in Denmark fewer than one in ten took the saved time as time off. The rest went straight back into work. Hana says salaried workers can't go home early, so the hour goes to the employer. Gus brings a Berkeley study where people filled the time themselves, often without being asked. The household machines of the last century did give time back, and it went to the kids. The question isn't whether AI saves time. It's who gets the hour.
How often does meditation go wrong? Under one in a hundred, or two in three, depending on whether anyone asks. Two large US surveys find about one in ten meditators say a bad effect touched their daily life, and a few in a hundred say it lasted more than a day. Hana says discomfort isn't harm, and brings the only study with a control group: a short session was no more distressing than browsing the internet. Gus says the risk grows with practice, and that something with roughly the side-effect rate of talking therapy mostly comes without anyone whose job is to ask. They agree on the asking. They don't agree on the phone.
Do you need to know what you want before you act? About 22,000 people who couldn't make up their minds let a coin decide whether to make a big change, and six months later the ones told to change were happier. Hana says that proves our caution is miscalibrated: we overestimate how likely a no is and how many people are watching. Gus says the coin only worked because those people had already thought themselves to a standstill. Along the way: a fig tree, a choice-overload meta-analysis, and Jeff Bezos's one-way and two-way doors. They agree on the door test. They don't agree on how many doors are two-way.
Blocking all internet on your phone for two weeks improved mental health, wellbeing and attention in a well-built 2025 trial. But only 119 of 467 volunteers kept the block on, and the ones who didn't improved too. Gus defends the study, then takes apart its most quoted line, the comparison with antidepressants. Hana argues that a plan most people can't follow isn't a prescription, then finds better news in the data: you may not need to do all of it. Or some of the benefit was never the block.