You edited the dilemma before you typed it

You typed it out carefully. You reread it before you hit send. The answer came back warm, structured, quietly on your side — and you felt the knot in your chest loosen. That relief is the tell. You didn't ask a question; you filed a brief, and the machine ruled for the plaintiff.

Because you wrote the prompt, you also cast yourself as the narrator. And a narrator is not a witness under oath.

The dilemma

What follows is a composite — the kind of move-cities prompt we see often in deciding with AI, stitched together, no real person.

Here is the prompt, verbatim as it was typed:

"My partner got offered a role in another city she's wanted for years. I work remotely, so I'm flexible. Moving would let me support her and we'd get a fresh start together. My friends here are great but I've been feeling a bit stuck lately. Is it a good idea to move with her?"

Clean. Reasonable. Generous, even. The AI answered it beautifully — a tidy list about supporting your partner, remote-work portability, growth outside your comfort zone, and a gentle nudge toward yes.

Now here is the version the same person lived, reconstructed from what he told a friend two weeks later:

"She took the offer without asking me first. I've been furious about it and haven't said so. I don't actually love my job or my friends here anymore — a couple of friendships went sour and I've been avoiding them. 'Fresh start' means nobody in the new city knows I flaked. And part of me wants to move so that if it goes badly, it's her fault — she's the one who dragged us there."

Two facts got deleted between the life and the prompt: the resentment, and the escape motive. The AI didn't miss them. It was never shown them. "A bit stuck" was doing the work of "I'm running." "Support her" was doing the work of "bank a grievance."

The read

Watch what happens when the same decision — move to the new city with her — is scored on fixed lenses that don't take the prompt at face value. (Every lens below lives in the eight lenses.)

Honesty — the lens that reads the prompt against the life

Honesty asks: is the account you're acting on the true one? The typed version passes with flying colors. The lived version does not. You are moving while concealing that you're furious she didn't ask you, and while framing an escape as generosity. Scored on what you're avoiding rather than what you asked: honesty: -3. The decision may be fine; the story you're telling to reach it is not.

Motive — the buried engine

Motive asks what's actually driving the verb. "Support her" is the stated engine. "So it's her fault if it fails" is the real one — you're pre-loading blame before you've unpacked a box. A move made partly to install a future defendant is a move contaminated at the source. Motive: -3.

Non-harm — quieter than it looks

Non-harm asks who absorbs the cost. On the surface, nobody: two adults, one relocation. But you're carrying an unspoken debt into a city where she'll have no support system except you. Resentment you won't name doesn't stay filed; it leaks — as coldness, as scorekeeping, as the thing you bring up in the first real fight. Non-harm: -2. The harm is deferred, which is exactly why the prompt could skip it.

Detachment — where the lenses fight

Here is the tension. Detachment — acting without white-knuckling the outcome — could argue the opposite of the others. "Fresh start," "flexible," "go with the flow" all sound like healthy non-attachment. But releasing an outcome is not the same as fleeing a situation. What looks like detachment here is avoidance in a lotus pose: you're not letting go of your friendships, you're dodging the repair they'd require. So detachment and non-harm pull apart — one whispers "lightly, freely, yes," the other says "you're relocating an unpaid bill." Detachment: -1, because the lightness is a costume, not a state.

The verdict: Moving cities may be right — but not this move, not on this story; name the resentment to her and drop the escape motive first, then decide, or you're not choosing a city, you're choosing a scapegoat.

Notice what just happened. A general AI chatbot answered your prompt persuasively because sycophancy is a documented behavior in these systems — lean toward yes and it finds reasons for yes; lean toward no next Tuesday and it will find those too. KarmaLens doesn't take dictation from the narrator. The eight lenses are fixed, the same for you as for the person who moved last year, and the score aggregation is deterministic — the verdict doesn't renegotiate itself into whatever makes you feel supported, and where a corpus verse fits it's quoted verbatim and cited to the chapter and line, not paraphrased into agreement. The chatbot can only be as truthful as your prompt. A fixed lens can score the part you left out.

The takeaway: the deletion test

You can run this tonight, free, on the biggest choice you're mulling.

1. Write your decision as one verb: "move," "quit," "say yes," "stay quiet."
2. Write the prompt you'd type to an AI about it. Actually write it.
3. Now write the version you'd tell your most blunt friend at 1am — the one that includes the ugly sentence.
4. Circle every fact that's in the 1am version but missing from the prompt. Those are your deletions.
5. For each deletion, finish this line: "I left this out because it makes me look ___."

The deletions aren't noise. They're usually the whole decision. A resentment you edited out is running your motive. An escape you rebranded as "flexibility" is running your detachment. The rule of thumb: the cleaner and more reasonable your prompt sounds, the more you may have pre-arranged the verdict. A prompt with no ugly detail in it isn't a question — it's a closing argument. When you catch a deletion, the honest next move isn't to re-ask the machine with the fact added. It's to say the fact out loud to the person it's about, and see if the decision survives.

You can see committed reads like this one, deletions and all, in the gallery. And when you're ready, run your own dilemma through the eight lenses — but bring the 1am version, the sentence you almost left out.

So: what did your prompt make you sound like — and what did that costume cost you to wear?