Kairos — Upside Economics and the Robertas Conversation (2026-08-12)

Upside economics, and the one conversation

Captured 2026-08-12 from Lee's session question: "If their gap is, say 3M yearly, and we do manage to cover that — do we capture a percentage of it or not? Or do we only get the percentage from the profits?" Nothing here is sent or agreed.

Where the commercial position actually stands

Today the answer is neither. Live terms are €10k/month against ~100 combined hours, fiat, and nothing else. If Kairos generated €3M of annual revenue tomorrow it would capture €0 of it. There is no upside instrument in force.

The architecture already specifies the right shape. Kairos — Mammoth Protocol Architecture (2026-07-24) carries an instrument matrix: growth inside an existing product → time-limited share of incremental gross profit or cash received. Incremental, measured against a pre-registered baseline. Under that instrument MN's ~€2M net loss never touches Kairos compensation — the payment basis is the delta the intervention produced, not whether the company crossed zero. Lee's worry (our €3M gets absorbed filling a pre-existing hole) is real and the protocol already designed against it. It also holds that base fees continue regardless and are never traded for speculative upside.

What is missing is the instance, and the clock is the problem. The protocol's own heading is "Open economics"; Re-contract clause 10 says the upside agreement comes before the hunt. The hunt is running now, with a board date in the first week of September. Leverage moves one way only: negotiating a share after showing MN which opportunity is the winner is negotiating face-up. Today there are candidates and no winner — maximum leverage, closing.

Measurement rights are the enforcement mechanism. The protocol's own line: "Without measurement rights, any percentage is decorative." Every warehouse query currently runs through Šaras's principal, informally, with no post-engagement continuity. A share of incremental gross profit that cannot be measured after the engagement ends is unenforceable. This reclassifies the Lee-owned BigQuery grant from plumbing to commercial protection.

The Robertas gap question — decouple the two asks

Ro deflected when Lee asked what the gap was, having himself framed the goal as "close the gap." Lee's line at the time was correct and went unanswered.

Under a delta-based instrument, Kairos does not need MN's P&L to price its own upside. So drop the coupling and say so:

I need the gap to do the job you gave me — break-even is the mandate. I don't need it to price myself; I'll price off the measured delta of whatever we ship.

That removes the only reading under which the ask sounds self-interested, and makes the mandate ask purely operational.

Do not read intent into the deflection. Kairos — Frameworks Applied (HRO + Dunford) already documents Ro's "floor is break even, but let's not stress, we're about abundance" posture. The deflection fits that pattern rather than a commercial maneuver, and the structural consequence is identical either way — so protect the position without resolving his motive.

The Motive — and it points at us first

(Lee raised Lencioni's The Motive, 2026-08-12.) The reward-versus-responsibility split, and the responsibility reward-centered leaders most reliably avoid: the uncomfortable conversation. Tempting to aim at Ro. Two soft data points do not diagnose a person.

The version that costs something — look at what the session was. Agent autonomy, model tiers, a 21-item ask packet, a readback protocol, a canon correction. All legitimate. And of the four Tier-1 items in Kairos — MN Ask Packet (2026-08-12) — the mandate-changing ones — three are Robertas, in a room, and little below them matters until they land. Everything else produced today is downstream of one conversation nobody has scheduled.

Lane 3 was named as where the 10x lives; the whole session ran in lane 1. This is the same gradient identified in Kairos — Autonomous Agent Lanes (2026-08-12) — attention flows to where the feedback is — operating on the unit's own strategy session, not just on its research. Claude helped build the machinery, so this is a mirror flag as much as an observation.

Three things that make the conversation easier than it was in July

1. The decoupling move above — removes the self-interested reading.
2. A concrete, non-accusatory artifact. Not "show me your finances" but: "We carried your cost base at €2M for four weeks because I wrote down 'burn' when you said 'loss.' It's €4M. I don't want to be wrong like that in front of your board in September." Reporting our own error and asking for help preventing the next one. Service framing, and true.
3. A deadline that isn't ours. The board meets in the first week of September. That converts an intimate question into an operational one: what do you want me to put in front of your board, and against what baseline? Ro can decline to discuss the treasury with a consultant; it is much harder to decline to specify the baseline for a document he asked for.

Honest limit: none of this predicts that he answers. What has changed is that the cost of not asking is now measurable — four weeks of work on a halved denominator, and a September deliverable pointed at an unnamed perimeter.