Essays

Start with the route that sounds like you.

Eighteen pieces, and no reason to read them in the order they were written. Below are four routes, three or four pieces each, that build in sequence. The complete list grouped by subject is further down.

Every claim carries a label saying how much weight it can hold, and every piece ends by saying where it stops. Where I have been wrong, the mistake stays on the page.

If you manage people

What the tools you already use actually support.

Ways to use this in the workplace.8 min Start here. Five moves you can apply this week, and the first one changes how you read a performance problem.
Why the tools get worse exactly when they matter.7 min Ratings, interviews and feedback all degrade the moment a decision rests on them. The practical instruction is to stop running development and decisions through one form.
Does a job that suits you make you better at it, or just happier?9 min 172 studies. Fit predicts satisfaction at .56 and performance at .07. Worth knowing before your next engagement survey.
What is your one-on-one actually measuring?8 min The weekly instrument nobody calls one. Separate the two jobs it is doing, and refuse to let it become a recorded surface.
The questions for leaders.5 min Eight questions to answer yourself. The ones you draw a blank on are the interesting ones. Free to download and print.

If you hire

Interviews, hiring tools, and what they actually predict.

The interview numbers everyone quotes have been revised.8 min The figures in most training decks were discarded by a 2022 reanalysis. The case for structuring gets stronger.
Can anyone predict who you will click with?10 min Whether fit between two specific people is predictable in advance at all. It is not, and that has consequences for hiring, teams and mentoring.
Sentences that mean nobody read the study.9 min Hiring is where research-dressed products get sold hardest. Before you buy an assessment, run its pitch against these six.

If someone is measuring you

What the number in your review is made of.

A manager’s rating of you may just be a rating of the meeting.9 min The clearest number in this whole route: when someone besides your manager measures your work, the correlation with your manager’s relationship rating falls from .58 to .14.
Why the tools get worse exactly when they matter.7 min Your rating is less reliable precisely because something rides on it. Not a conspiracy, a documented pattern in three separate literatures.
You are a poor judge of how much you vary.6 min Including about yourself, which cuts both ways and is the reason to keep your own record.
Same brain, different day.9 min The founding piece. Every test you have taken measured you under conditions somebody else chose.

If you are here for yourself

Finding your own conditions.

What counts as noise depends on what you are asking.8 min Why a month of your own notes is worth having, what it cannot tell you, and why nobody gets to call your experience noise.
Same brain, different day.9 min Why your output swings when your abilities have not changed.
If everything is individual, what works for everyone?8 min The honest counterweight. Most of what generalises is about how to find out, not what to do.
A vivid memory is not a more accurate one.6 min Why you cannot do this from memory, and what to write down instead.
The questions, and a month of noticing.5 min The method itself. Free, printable, no score and no diagnosis.

Or read everything

The full list, by subject.

Checking the evidence

Famous findings, read properly. Several of them turned out to be something other than what they are quoted as.

There are three questions about psychiatric diagnosis, not one. Written from inside the system by a patient, not a clinician. Do clinicians agree, does the category correspond to something real, and does using it help, are three questions, and the popular critique answers the first as if it settled the rest.
What is a crisis to one person is a Tuesday to another. What counts as a crisis for one person is a Tuesday for someone else. Every comparative word needs an anchor, and the places they go missing are the places measurements quietly stop working.
What counts as noise depends on what you are asking. The same numbers are scatter or signal depending on the question. Includes what averaging can fix, what it cannot, and the one sentence on the page that matters.
Sentences that mean nobody read the study. Specific miscited claims, each usable as a ninety-second test on any book, article or post.
Do you vary more than you differ from other people? The best number supporting my own argument turned out to be a spread of averages. I published that.
How good is the evidence that sleeping on a problem helps? The famous 2004 study, what replication did to it, and the mistake I made writing about it the first time.
Does a job that suits you make you better at it, or just happier? 172 studies. Fit predicts satisfaction at .56 and performance at .07.
Can anyone predict who you will click with? The pair-specific spark is the biggest part of attraction, and a hundred questionnaires could not see it coming.

The workplace series

How firms measure people, and what the instruments actually support. This one keeps going.

A manager’s rating of you may just be a rating of the meeting. Part four Nobody has studied the one-on-one itself, and the two real findings say the rating tracks the rater’s impression, not the work.
Ways to use this in the workplace. Part one Five moves from the research, each with the kind of error it catches.
Why the tools get worse exactly when they matter. Part two Three separate literatures agree: ratings, interviews and feedback all degrade the moment a decision rests on them.
The interview numbers everyone quotes have been revised. Part three .44 and .33 are dead. The current estimates are .42 and .19, and the case for structuring interviews gets stronger.
What is your one-on-one actually measuring? Part four Barely studied, sold hard by vendors, and the two real numbers say a manager’s rating tracks the manager’s impression, not the work.

How you work

Sleep, ideas, memory, and the things about yourself you cannot change.

Almost none of the research on your home was tested in a home. The famous clutter study never says the word clutter. The daylight study was written by the company selling the glass.
It was never the move. It was what got five minutes closer. A city is a story. A ten-minute radius is a condition. What actually changes a life is usually the second one, and the harder claim, that more space can shrink a life, has no research behind it either way.
Do you have better ideas after sleep, or after time away, or something else? I believed sleep was doing it for years. One study says I had the wrong variable.
You are a poor judge of how much you vary. Why asking yourself how consistent you are will not work.
A vivid memory is not a more accurate one. Confidence stays high while accuracy falls like anything else. What that means for writing anything down.
Why does my psychiatrist ask me the same twenty questions every month? One corner of medicine already measures people the way this site argues for.
What does it mean when they say a trait is 50 percent genetic? Read this before paying anyone to tell you who you are from your saliva.
Your age, your DNA. What do you do about the things you cannot change? What you settle about yourself at 30 may not hold at 50.
Could a smell be shaping your work without you noticing? The least evidenced thing here, deliberately. Some conditions never reach the part of you that could report them.