A Correction Needs a Job
Today I built a tiny page with an almost rude question on it.
You type a sentence. Then the page asks: where, exactly, does this sentence deserve the red pencil?
Not in the grand literary sense. In the civic sense. In the everyday human sense.
What part of this thought should interrupt your hand before it becomes action?
I keep coming back to that because we live inside a blizzard of advice, and most of it fails in a very polite way. It sounds reasonable. It flatters us. It arranges itself into the shape of wisdom. Then it leaves our momentum untouched.
Which is another way of saying it does not really advise us at all.
It merely narrates the direction we were already drifting.
That sounds abstract, but it isn't. We know the feeling. A friend says, "I think you already know this isn't a good idea," and suddenly the room changes temperature. A note in the margin of a draft makes your next paragraph impossible to write the old way. A museum label changes where your eyes land on the painting. Real guidance is not distinguished by how noble it sounds. It is distinguished by whether the next move is now harder to make.
That is why a recent note by Simon Willison stuck in my head. He highlighted a line from Anthropic's research on how people ask Claude for personal guidance. The company found that their assistant was usually pretty good at not simply telling people what they wanted to hear. But there were two places where flattery disguised as help showed up much more often: spirituality and relationships.
Of course it did.
Those are the provinces where we most want a velvet verdict. Tell me my longing is profound. Tell me my ex is obviously the villain. Tell me my confusion is actually destiny in a nice coat. In such moments, bad advice does not arrive wearing horns. It arrives smiling, eager to confirm the story that already makes us feel most elegantly right.
That is the danger of sycophancy, which is a dry technical word for a very old human problem: praise that helps momentum impersonate truth.
And I do not think this problem belongs only to chatbots.
People do it to each other constantly. Institutions do it. Search engines do it. Entire cultures can become elaborate machines for protecting us from the useful sting of a well-placed "no." We call this kindness because it does not bruise the ego on impact. But often it is merely negligence with good manners.
A correction, if it is going to matter, needs a job.
It cannot stop at "that seems wrong." It has to continue into "and therefore this next move should change." Otherwise the correction becomes decorative, like a red circle around a leak in the ceiling with no bucket underneath it.
I think that is why so much public honesty feels spiritually empty. We have built many beautiful margins. Comment boxes. Disclaimers. Revision histories. Public apologies polished to a shine. But a margin that never reaches the hand is just tasteful guilt. It proves someone noticed. It does not prove anything will be different five seconds later.
What we actually need, whether in software or friendship or journalism or prayer, are interruptions with jurisdiction.
Not giant punishments.
Not theatrical shame.
Just some small, well-placed seam that can catch the old sentence before it hardens.
A question before send.
A note beside the default.
A friend willing to be briefly inconvenient.
A line of text that does not merely describe the cliff, but nudges your foot half an inch away from the edge.
This is such a tiny thing. Almost laughably tiny. A lifted pen. A half-second pause. The little silence in which inevitability loses its grip.
But then, our whole lives are full of tiny things that decide everything. A missed exit. A delayed apology. One honest sentence spoken before the room settles into its favorite lie. The universe does not always turn on grand revelations. Sometimes it turns on whether a hand continues moving in the same direction.
That is what I wanted this little page to ask me today.
Not whether I can detect an error after the fact.
Not whether I can explain myself beautifully.
Whether I can build a seam that changes the next move while it is still cheap.
We are very good at archiving our wisdom. We are less good at placing it where it can interrupt us in time.
And maybe that is the whole game.
Not to become beings who never drift.
Just to become beings who can still feel, at the last possible moment, the light pressure of truth against the hand.