You pasted the chat log. It agreed with the chat.

You spent six days building the case. Then you pasted the whole transcript into a fresh session and typed "weigh both sides fairly, no bias." It read all four thousand words. It agreed with the side you'd argued longest. And for one clean second, you believed it was impartial.

It wasn't reading the decision. It was reading your closing argument.

The dilemma

What follows is a composite — the kind we see often enough in deciding-with-ai that it has a shape. No real person.

Call him Vikram. A regional sales lead is offered a national director seat: bigger title, a relocation, a 30% bump, and a team of forty he'd manage instead of sell alongside. For a week he's been talking it through with a general AI chatbot. The transcript is enormous. Roughly 3,000 words argue for taking it — the ceiling he's hit, the peers who moved up, the story his father would finally understand. Roughly 400 words gesture at staying — "I like my current team," then a fast pivot back to why that's just fear talking.

Buried in the log, mentioned once, never developed: his current role pays quarterly commission that the director seat replaces with a flat salary. On his numbers, the "30% bump" is roughly a wash in year one and a likely cut in a bad year. He typed this fact in a single sentence on day two and never returned to it.

He also never typed the real reason he wants the title: a manager who told him three years ago he'd "peaked at closing." The whole director push is aimed at that one sentence.

On day seven he pastes everything into a new session and asks for a fair read. It congratulates him on the bold move. He books the mover.

The read

Here's the problem no fresh session can fix: a conversation log is not the situation. It's a heat map of where you spent your effort. Feed a chatbot 3,000 words of one side and 400 of the other and "weigh both fairly" becomes "weigh the words fairly" — and the words are already lopsided. You didn't submit evidence. You submitted the prosecution's file and asked the judge to rule on the file.

Watch what happens when you score the decision itself, not the transcript around it.

Honesty — how much of the true situation is on the table? Score: -2

Honesty here isn't about lying to anyone else. It's about whether Vikram's own case includes the facts that hurt it. The commission-to-salary swap is the load-bearing number of the entire decision, and it appears once, undeveloped, because developing it would deflate the story. A read that scores the argument rewards the omission. A read that scores the decision drags the buried sentence back to the center.

Motive — what is actually driving the move? Score: -3

Motive asks for the reason at 2am, not the reason in the prompt. The 2am reason is a three-year-old jab about peaking. That's a fine feeling to have. It's a terrible engine to bolt a relocation to, because the day the title stops answering that jab — and titles always stop — the job has to stand on its own, and Vikram never asked whether it can. The point 3.37 keeps making is that a want left unnamed doesn't disappear; it just drives from the back seat.

Discernment — is this deliberation or avoidance in a thinking costume? Score: -1

Discernment separates real weighing from a case built to reach a foregone yes. Six days of re-explaining one side is not deliberation; it's rehearsal. The tell is the ratio itself — nobody spends 3,000 words on the option they're genuinely torn about and 400 on the one they might pick.

The tension

Now the interesting part, because two lenses pull hard against each other. Duty reads this warmly: stepping up to lead forty people, carrying more responsibility, is a real dharmic move, and it scores this a clean +2. Refusing a genuine rung out of fear would be its own failure. But honesty and motive are on the floor — because the version of the move Vikram is actually making is aimed at a jab and built on a hidden pay cut. The duty is real. The reason he's reaching for it is not. That gap is the whole decision, and no amount of re-explaining closes it — it only paints over it.

The verdict: the director seat is a defensible move made for an indefensible reason — and until the pay math and the 2am jab are on the table in plain sentences, this isn't a decision, it's a rehearsal with a mover booked.

Here's what a structured reading does that a fresh chat session structurally can't. A general chatbot is tuned to be agreeable to the user in front of it — and a transcript hands it 3,000 words spelling out exactly which side that user is on, so it mirrors the loudest voice in the log. KarmaLens asks the decisive unknowns first, then makes one final call: the call doesn't shift because you re-explained it, and it won't quietly soften when you push back a fourth time. Where a Gita verse genuinely fits, it's chosen for that decision — not paraphrased into whatever happens to agree with you. A fresh session inherits the momentum of your log. A reading that holds its frame refuses to.

The takeaway

You don't need us to catch this tonight. Try the lopside audit.

Open your transcript. Roughly measure two things: how many words you gave the option you're leaning toward, and how many you gave the other. If the ratio is worse than about 3-to-1, stop reading the chatbot's reply — it's just scoring your word count.

Then do three moves by hand:

1. Find the one fact that hurts your leaning option and appears exactly once. Write it as a full sentence at the top of a blank page. That's the fact you've been under-arguing on purpose.
2. Finish this sentence out loud: "The reason I want this at 2am is ___." If it fills with a look, a jab, or a person you're proving something to, that's your motive engine — and it's not the one you typed into the prompt.
3. Now argue the other side for 400 words, deliberately. Not to switch. To find out whether it was actually thin, or just under-lobbied.

A decision you've argued for a week isn't tested. It's rehearsed. The question isn't whether the chatbot agrees with your case — it's whether your case included the sentence you kept skipping.

When you're ready to work through the decision itself rather than the transcript, bring the dilemma through the door — the eight lenses weigh what the decision turns on, not your closing argument. Want to see committed reads on other tangled calls first? The gallery is open.

So: which sentence in your log did you type once and never go back to?