The Third Body Was Never on the Term Sheet

Yesterday in Fortune, Saurabh Gupta of DST Global described the 2026 AI market as a three-body problem: closed-source labs, open-weight models, and the application companies built on top of both. Two bodies orbit cleanly. A third one large enough to bend the others turns the system chaotic. Founders read that as a market essay. They should have read it as a description of the last partnership they signed.

Most commercial agreements are written as if only two signatures will ever move the work. The deck has two logos. The Slack channel has two companies. The weekly sits two operators across a table and pretends the rest of the universe is weather. The third body is already in the room. It is the marketplace that can change ranking overnight. It is the agency holding the brand's calendar. It is the creator's other exclusive. It is the model vendor whose pricing rewrite lands in the middle of a joint launch. Nobody put that force on the term sheet, so both sides keep treating its gravity as an excuse.

Two signatures is a comforting fiction

The founder who has been burned before still reaches for a bilateral story because it feels governable. A two-party deal has a villain and a hero. A three-body deal has a physics problem, and physics problems do not take a recap email. So the founder writes the partner as the variable they can manage, and they file the platform, the channel partner, the procurement committee, or the co-investor as context. Context is how operators postpone naming the thing that actually sets the dates.

Watch the pattern on a live deal. An agency and a brand agree to a ninety-day co-sell. Week three, the brand's existing media partner moves the window. Week five, the agency's preferred creator books a conflicting exclusive. Week seven, both sides sit in a "alignment" call and diagnose chemistry. The chemistry was fine. The orbit was never two-body. Each miss arrived with a name attached that neither signature controlled, and because that name was absent from the agreement, both parties treated the miss as noise instead of as the real counterparty.

Consultants do the same thing when they sell a joint diagnostic and then discover the client's implementation partner owns the stack. Creators do it when they sign a brand and forget the talent manager who actually approves the posting calendar. Event producers do it when they lock a sponsor and then learn the venue's exclusive caterer rewrites the activation. The observable behavior is identical. The founder dates the partnership from the handshake and then spends the quarter managing a person who cannot move the third body.

The cost shows up as a calendar that nobody owns

The price of this fiction is specific. Six to twelve weeks of a senior operator's time spent writing recaps to a counterpart who keeps agreeing in the room and missing on the dates the third party actually controls. A pipeline slide that still lists the logo because the original two-party story remains true on paper. A team that learns the founder's standard is a conversation with one person, while the work is decided by someone who never joined the thread.

Trust erodes in a particular way here. The partner did not lie. They described what they could do if the rest of their system held still. The founder heard a commitment. Both people leave the call feeling honest. Then the third body moves, and each side files the other as unreliable. That filing is expensive. It produces the private ledger of slights, the explosion meeting that is a recap for one person and a first hearing for the other, and the quiet decision to never run another deal with someone who was never the constraint.

Gupta's point about AI was that no single force can dictate where the system settles. Partnerships that ignore that fact keep writing enforcement language aimed at the wrong signature. They put SLAs on the agency and leave the marketplace unnamed. They put exclusivity on the creator and leave the manager unnamed. They put a launch date on the consultant and leave the implementation partner unnamed. When the date slips, they escalate to the person who already told the truth about their own limited gravity.

Name the third body before you announce the pair

The operators who stop repeating this do a colder piece of diligence before the announcement. They ask which other commercial relationship can move this calendar without either of us in the room. They write that name into the working document in the same week they write the two logos. They put one dated commitment on the table that the third body would have to honor, and they watch who can actually pull that commitment. A partner who cannot name their own third body is selling you a two-signature story they do not live inside.

That test ends some matches early. It should. A deal that only works if a marketplace, a manager, a venue, or a model vendor stays still is a deal with four parties and two chairs. Tools that force specificity in counterpart selection, including platforms like onSpark AI when the work is choosing people who can carry a commercial relationship, help only after the founder admits the handshake was never the system. Selection that ignores the third orbit selects for charm in a two-person room.

The Fortune essay will be filed under AI strategy. The useful reading is smaller and closer. If your last partnership died in execution, pull the calendar and find the week the work actually moved. The name attached to that week is often missing from the contract. Date the relationship from the first time you both agreed which third body you were actually in business with. Everything before that is a clean two-party story, and a clean two-party story is still only a story.