Module 01 · signal fire · Cooperation & Trust ·
The Stag Hunt Never Ended
Why a 300-year-old parable about hunting parties becomes the operating manual for a world full of delegated agents.
Hamburg, March 2036. Your procurement agent has spent the night negotiating with fourteen supplier agents, and it opens the morning briefing with a confession: it walked away from the best price. The cheapest counterparty’s agent, it explains, showed a pattern of defecting on delivery windows whenever spot prices moved — hunting hare, reliably, whenever a stag was inconvenient.
You approve the walk-away. Not because you checked the maths — because three years of working together taught the two of you which promises this counterparty keeps. The agent found the pattern. Deciding what your firm’s word is worth remained, stubbornly, your job.
In the harbour arbitration hall, ten thousand agents assemble as a single shimmering parliament, settling a decade of trade disputes in eleven minutes of pure negotiation — and then pause, all at once, to wait for nine human signatures.
The mechanism
Rousseau’s parable is three sentences long: hunters can feast if all of them hold the line on a stag; any one of them can bail out for a sure, small hare. Mutual gain requires mutual reliance, and the safe defection is always available. It is the structure underneath supplier relationships, open-source maintenance, standards bodies, and every team you have ever been on.
Delegation does not shrink that game; it multiplies the players. Every agent you send into the world is a new principal–agent relationship on your side and a new counterparty on someone else’s. The number of trust decisions grows roughly with the square of the participants — while the thing that makes cooperation stable, a credible and accountable commitment, still traces back to an entity that can be sued, shamed, or bankrupted. Under the baseline’s assumption B4, that entity is not the agent.
Machines can compute the equilibria faster than we can name them. What they cannot do is be the reputation at stake.
So the skill that appreciates is the old one: judging whose commitments are credible, making your own credibly binding, and knowing when to walk away from a good price offered by an unreliable partner. Axelrod’s tournament results — be nice, retaliate, forgive, be legible — were derived for humans and hold for any strategic actor; what changes by 2036 is only how often you delegate the playing while retaining the promising.
⏵ Toy 01 · The stag hunt, with agents
counterparty agent: temperament hidden — you learn it from play
their moves: —
your moves: —
- Write down which of your working relationships run on reputation rather than contract — that list is what you will be lending to your agents.
- Play one iterated game seriously this year — a negotiation, an open-source contribution rhythm, a standing trade with a colleague — and notice your own defection temptations.
- When you delegate to an AI today, decide in advance which commitments it may make in your name. Practice drawing that boundary before it carries money.