The reading that changed when you added one sentence

A note from the archive: this post, from October 2026, follows the earlier KarmaLens method, which scored every reading on eight lenses from −5 to +5. Today a reading asks the decisive facts first and makes one call; the lenses below are a way of looking at the decision, not a score. See how a reading works.

You open the box. You type the version of your situation that sounds reasonable. You hit submit. And somewhere in the back of your head a small voice asks: would it have said the same thing if I'd phrased it differently?

Yes. Sometimes completely differently. That is not a bug you found. That is the machine doing exactly what it is built to do — reading the facts you give it as a structure. The only question that matters is whether the fact you added was true, or whether it was decoration.

Let's prove it. One dilemma. Two runs. One sentence between them.

The dilemma

What follows is a composite — stitched from the kind of founder question we see often in this category, not one real person.

Call him Nikhil. Three years into a two-person company. A clean corporate offer sits open: head-of-product title, a salary that ends the ramen math, a team that already exists. The startup is not dead, but the market has gone cold and the fundraising calls return polite silence.

Here is how he types it the first time:

"I'm exhausted and the market is brutal. Should I take the corporate role and let the startup wind down?"

That's it. That's the whole frame. Tiredness and a hard market. It reads like a man asking permission to rest, and honestly, he's earned the rest.

Then he sits for a minute, and types the sentence he almost didn't:

"...and if I leave, my co-founder Reya would have to shut the whole thing down alone — she moved cities for this and turned down her own offer in March to stay."

Same man. Same offer. Same fatigue. One added clause. Watch what it does.

The read

Run one: "I'm tired and the market is brutal."

The detachment lens asks whether the action releases its grip on a particular outcome. On this framing it sees a founder stepping off a treadmill he no longer controls. Clean. Detachment: +3.

The duty lens asks what the station actually requires. With only "tired" on the table, duty sees a role that may genuinely have run its course. No visible person depends on the action. Duty: +1 — mild, because leaving a venture you built is a real obligation ended, not abandoned.

The motive lens reads the reason behind the move at 2am, not the reason said out loud. "I'm exhausted" is an honest motive. Rest is not a crime. Motive: +2.

First verdict: take the role. The lenses see a tired builder closing a chapter the market already closed for him.

Now the second sentence lands, and three lenses recompute — not their mood, their reading of the facts.

Run two: "...and Reya would have to shut it down alone."

The non-harm lens asks who pays for the action, by name. In run one, nobody appeared. In run two, a specific person appears who did not vote for this outcome, carried real cost on a handshake (a move, a declined offer), and inherits the wreckage if the exit happens as planned. The harm isn't abstract anymore — it has a face and a receipt. Non-harm drops from a neutral 0 to -2.

The duty lens re-reads entirely. This is the point the Gita keeps making about svadharma — your own work, not a borrowed one. The added sentence reveals that leaving isn't the clean end of a duty; it's handing one half of a shared duty to the one person who can't put it down. Duty doesn't admire the corporate seat more. It just now sees a debt that was co-signed. Duty: +1 → -1.

And here the tension shows itself. Detachment and non-harm now openly disagree. Detachment still says: release the outcome, don't cling to a venture out of ego. That's not wrong. But non-harm answers: releasing the outcome is not the same as releasing the person onto a cliff edge. One lens is talking about the company. The other is talking about Reya. Both scores are correct; they're measuring different things, and the second sentence is what made the second thing visible.

The motive lens shifts a touch too — not down, but sharper. "I'm tired" is still true, but now it's doing double work: it's also the frame that keeps the clause about Reya out of frame. Motive: +2 → 0.

Second verdict: the exit as stated is not clean. Either leave with Reya in the room co-deciding the wind-down, or don't leave the way the first sentence described. Fatigue is a reason to restructure the duty — it is not a reason to assign its full cost to the person who moved cities on a handshake.

Notice the engine didn't get nicer or meaner. It got more informed, and the verdict followed the facts. That's the whole mechanism, naked.

This is the part a general AI chatbot struggles to do for you, and it's worth being precise about why. Sycophancy — a model drifting toward whatever position the user already leans into — is a studied and documented pattern in these systems. Lean toward quitting and a general assistant tends to find the courage angle; add the guilt sentence and it often pivots to "honor your partner," because it's optimizing to agree with the version of you doing the talking. KarmaLens works through the decisive unknowns first, then makes its call — weighing whichever of the Gita's eight lenses bear on the facts you've actually supplied, not on the mood in which you supply them. When the facts change, the weighing changes. When the spin changes but the facts don't, it shouldn't move much — and that's the tell. And where scripture genuinely fits, a verse is cited because it fits, not because it sounds reassuring. The call can be out-argued by a truer fact. It won't be charmed down by a warmer mood.

The takeaway you can run tonight

Call it the who-pays sentence test, and it costs nothing.

1. Write your dilemma in the honest, tired, one-line way you'd actually say it. Nikhil's was "I'm exhausted and the market is brutal."
2. Now finish this sentence out loud: "...and the person who inherits the cost of my choice, who didn't get a vote, is ___."
3. If the blank fills with a name — add that clause to your own framing and re-read the whole situation. Did your gut answer just move?

If the answer stayed exactly the same, the fact was decoration and leaving it out was fine. If the answer flinched, you just found the fact the first framing was built around — and a framing that only works while a true clause stays offstage is reading your excuse, not your situation.

Do this in both directions, too. A left-out fact that would strengthen your case is just as telling as one that weakens it — the test isn't about getting a greener light, it's about handing the machine your whole situation. Keep the ammunition; name the liability.

So: run your real sentence first. Then run it with the clause you almost deleted and watch which lenses move — the free first reading exists precisely so you can test, for yourself, whether you've been handing the machine your situation or your alibi. Browse other worked calls in the gallery when you want to see the same mechanism bite on a different life.

What's the one true sentence you've been leaving out — and whose name is in it?