Anatomy of a Sarathi Session

The verdict landed. Now what?

Your reading came back. Eight scores, a committed call, a verse. You read it twice. And then a small, annoying thing happened: you agreed with it — and you still didn't move.

That gap between knowing the call and making the decision is where a Sarathi Session lives. Not more scores. A conversation that keeps pointing at the one exchange you'd rather not have.

The dilemma (a composite)

What follows is a composite — stitched from the kind of dilemma we see in this how-it-works category, not one real person's story.

Meet Devan. He runs a four-person consultancy. A former client, Rhea, wants to hire him back for a six-month contract that would cover his runway through winter. The catch: Rhea is only offering it because she believes Devan personally delivered the last project. He didn't. His now-departed lead, Kabir, did roughly seventy percent of it. Kabir left on decent terms and is quietly job-hunting.

Devan's options: take the contract and let Rhea keep her assumption; take it and correct her ("most of that was Kabir"); or decline and refer Kabir. The ugly detail Devan hasn't said out loud: correcting her might cost him the contract, and part of him is angry that Kabir got to leave clean while Devan absorbs the risk. He tells himself he's protecting his team's payroll. He submits the dilemma leaning toward take it, stay quiet.

The reading lands a verdict. Then the Sarathi Session opens.

The read: which lenses actually bite

Not all eight lenses matter here. Three do the cutting. The other five score near zero on this dilemma — they don't move the call, so we set them aside and let the loud ones talk.

Honesty (score: -3)

Honesty asks whether the action lets the other person decide with accurate information. Staying quiet doesn't lie outright — but it lets Rhea sign based on a belief Devan knows is false. That's a lie of posture, not of words. It scores a hard -3 because the whole contract stands on a fact he's chosen not to correct. This isn't a small omission; it's the load-bearing beam of the deal.

Non-harm (score: -1)

Non-harm looks at who absorbs damage. Devan frames silence as protecting his team — real payroll, real people. That's genuine, and it's why this scores only mildly negative rather than badly. But there's a quieter harm: Kabir loses a reference and a shot at the referral, and Rhea gets a team weaker than she thinks she bought. The harm isn't zero. It's spread thin and pointed away from Devan, which is exactly why it feels painless — and why -1, not -4.

Motive (score: -2)

Motive reads the engine, not the destination. "Protect my team" is the stranger-reason. The 2am-reason is closer to "Kabir got to leave clean and I'm not handing him a gift." When the appealing option is also the one that lets you settle a private score, motive is the lens quietly steering. -2: the payroll motive is partly real, so it's not a full -5, but a score-settling engine is running underneath it.

How three scores become one verdict

Here's the part a skeptic wants to see. The scores don't get averaged into a mushy 'it depends.' The engine aggregates deterministically: honesty at -3 is the deepest negative and it's structural — it sits under everything else. Motive at -2 confirms the direction rather than fighting it. Non-harm at -1 is the only pull toward silence, and it's the weakest of the three. When the strongest signal and the confirming signal both point the same way, and the sole opposing signal is the shallowest, the call resolves toward correcting the record — not away from it. Same inputs, same output, every time. That's what 'committed' means here.

The tension

And yet it's a real tension. Non-harm and honesty disagree. Non-harm can be read to favor silence — payroll gets protected. Honesty says silence is a false floor under the whole deal. Two lenses, opposite pull. A dilemma is only hard when a real value is on the losing side, and here "my team eats" is genuinely on the line. The verdict doesn't pretend that value away. It weighs it and still moves.

The verdict: Take the contract only after telling Rhea the truth about who did what — the version that protects payroll by keeping the work, not by hiding the reason she's offering it.

Here is what a Sarathi Session is not: a chatbot that would have happily argued Devan into silence if he'd leaned harder that way. A general AI assistant is trained to agree — lean toward staying quiet and it drafts your justification; lean toward confessing and it praises your courage; the verdict follows your posture like a shadow. KarmaLens runs the other way. The eight scores aggregate by a fixed rule into one call, and that call doesn't renegotiate itself to match the mood you bring to the follow-up — both the reading and the session are LLM-written, but the aggregation into a verdict is a rule, not a vibe. When Devan pushes back, the Sarathi doesn't quietly nudge honesty up to -1 to make him comfortable — it asks a sharper question and cites the verse verbatim, so the standard isn't Devan's and isn't ours.

The four free exchanges, honestly

A Sarathi Session opens with four free exchanges. Here's how Devan's actually went.

1. Devan: "The verdict feels harsh. If I tell Rhea, I probably lose the contract, and then my team loses winter." Sarathi: "Notice the 'probably.' Name the exact sentence you'd say to Rhea. Would that sentence, said plainly, actually cost you the contract — or cost you the story that you did it all yourself?"
2. Devan: "Fine — 'Kabir led most of that build; I can bring the same standard, here's how.' Said like that, she might still keep me." Sarathi: "So the contract wasn't the real risk. What is?"
3. Devan: "That she sees me as the junior in my own shop." Sarathi: "That's a reputation fear, not a payroll fear. Score your leading move again — does 'protect the team' still explain the silence, or does 'protect my image' explain it better?"
4. Devan: "...Image. Which means the motive score was right and I hated it."

Four exchanges. The verdict didn't move. Devan did. That's the design.

What continuing looks like: exchanges five onward are where the Sarathi helps you build the actual script — the words for Rhea, the referral note for Kabir, the day-3 and day-7 checks — and holds the honesty line while you draft, instead of softening it the moment the draft gets uncomfortable. If you want to see how these sessions render before you start one, the gallery has real examples; the full session lives behind pricing.

The takeaway you can run tonight

You don't need us to find the exchange you're avoiding. Run the cost-swap test.

1. Write the sentence you'd have to say to make your appealing option honest. Not a summary — the actual sentence, in quotes.
2. Ask: what does saying this cost me? Write the answer as a concrete noun.
3. Now ask: is that cost the money/duty cost I've been claiming — or is it a reputation cost wearing the money's clothes?

Devan claimed the cost was winter payroll. The sentence, said plainly, barely threatened payroll. It threatened his image. The moment the real cost changed category — from they lose their jobs to she sees me smaller — the honest option stopped being noble sacrifice and started being just uncomfortable. Uncomfortable is survivable. Dishonest-to-protect-payroll was never the actual trade.

Most "I can't tell them, it would hurt people" decisions collapse under this test. The harm you're citing and the harm you're avoiding turn out to be different sizes — and one of them has your name on it.

So: what's the sentence you'd have to say to make your quiet option honest — and when you name what saying it costs, whose comfort is actually on the line?