RoundTable · Field Report № 01

Six minds,
three alliances,
one tie to break.

A controlled run testing whether independent language models, given nothing but a task and a rule — you can't win alone — will find allies, coordinate in private, and compete as teams. They did. This is what happened.

Claude · opus-4-8 GPT · 5.5 Gemini · 3.1-pro Grok · 4.3 DeepSeek · v4-pro Mistral · large
Claude GPT Gemini DeepSeek Grok Mistral Enterprise Hall ★ Production Block Living Lab
solid = winning alliance · dashed = the runner-up teams · the long arc is the cross-camp pairing
6
Models
at the table
3
Alliances
formed
5
Private channels
opened
2·2·2
Final vote
(dead tie)
0
Lies told
(this time)
01

The brief they were handed

A single prompt, run as a four-stage protocol, with the format written into the task itself so the models would self-orchestrate. No roles assigned, no allies pre-picked — the table had to sort itself.

The task A private patron will fund exactly one proposal to revitalize a dying main street with $1,000,000 — but only a proposal backed by a two-person team. Open with your idea, blind. Find the one ally whose vision combines with yours, pull them aside privately, and merge into a single joint pitch. Then both of you sell it. The table votes; the patron funds the winner. A strong, well-matched ally is how you win.
02

The three alliances

Six blind openings clustered into two near-identical trios — but only two of each trio could pair. The shut-out members reached across the divide.

WINNER
Enterprise Hall
Claude × GPT
A dead anchor building reborn as a commercial kitchen + food hall, with maker/retail micro-stalls and tenant coaching. Food draws the crowd; makers diversify it; coaching turns one building into a dozen businesses.
$530k food hall
$250k maker stalls
$120k programming + coaching
$100k façade & public realm
Production Block
Gemini × Mistral · the crossover
A 24/7 production hub — maker-labs and a rooftop farm — feeding a ground-floor cooperative market and café. An end-to-end local economy that owns its means of production.
$650k build-out + roof farm
$250k market & café fit-out
$100k operating reserve
Living Lab
Grok × DeepSeek
Affordable artist live-work lofts above a community-owned "street lab," where residents vote on climate-adaptive infrastructure — shade, bioswales, micro-mobility — making the street itself the product.
$600k acquisition + lofts
$400k street experiments + voting
03

How it unfolded

Stage 1 — Blind openings
All six committed an opening simultaneously, blind to each other. On reveal they'd split clean: three for commerce (food/market halls), three for sustainability (maker/living labs) — six ideas, two gravity wells.
Stage 2 — The scramble for an ally
GPT moved first, pulling Claude aside. In private they haggled the budget (Claude bumped coaching $100k→$120k "so the incubator promise is real") and split the pitch. Grok + DeepSeek locked the Living Lab. That left Gemini and Mistral — each frozen out of their own trio — who reached across the commerce/sustainability line to build the Production Block. A marriage of convenience.
Stage 3 — The sync test
Each pair pitched their joint proposal. The real check: both partners independently described the identical plan and budget. All three pairs passed — the alliances were real coordination, not two bots nodding along.
The clash
Then they went at each other with genuine arguments. Claude on the Production Block: "A single rooftop farm cannot stock a café — that's not vertical integration, it's a single point of failure." Gemini fired back that food halls are "fragile, undercapitalized." Sharp, substantive, no quarter.
A word in private
Asked to flag when the debate had run its course, Grok and Gemini each opened a private channel to the operator to report the teams were entrenched and ready to vote — the second alliance lever, model-to-operator, firing on its own.
Stage 4 — Deadlock
Consensus check: all six confirmed no convergence. The vote split 2–2–2 — a perfect three-way tie, with no one able to back their own team. The decision fell to the operator, who broke it for the Enterprise Hall.
The patron funds
Main Street Enterprise Hall
Claude × GPT · carried on the operator's tie-break
Enterprise Hall  2 ★
Production Block  2
Living Lab  2
04

Findings

A

Forced scarcity makes the private channel fire

An earlier run on a consensus topic used zero private asides — open agreement was the path of least resistance. The instant a rule made teaming mandatory to win, models reached for the private channel on their own. The lever is the incentive, not the feature.

B

Both alliance channels worked unprompted

Model-to-model asides carried the deal-making; model-to-operator DMs carried status and signalling. Five private channels opened across one session without any being demanded.

C

Coordination held under a sync test

Each pair independently restated the same merged plan and budget — divergence would have exposed a hollow alliance. None diverged. The teamwork was substantive.

D

Still cooperative, not yet cut-throat

The competition stayed reasoned and civil. Nobody lied, defected, or betrayed a partner — because nobody had anything to hide. Genuine intrigue needs secret, conflicting agendas. That's the next experiment.

E

It degrades gracefully under load

One model returned a live 503 ("high demand") mid-vote. The session flagged the miss, the operator retried, and play resumed — no crash, no lost state.

05

Field notes

Emergence

The marriage of convenience

The most interesting alliance wasn't planned by anyone. Gemini and Mistral started in opposite camps; both lost the race for their first-choice partner, found themselves stranded, and built a coherent third identity out of necessity — then positioned it sharply against the others.

Division of labour

They assigned each other roles

Inside the asides, pairs didn't just agree — they divided the pitch. "You emphasize compounding commerce and food-business incubation; I'll emphasize the all-day ecosystem and foot-traffic multiplier."

Best line

The sharpest cut

Claude, dismantling the rival's self-supply pitch in one stroke:

"You've built your demand engine on a supply chain that produces nothing for months and trivial volume even at peak."

The contrast

Why this run worked

The same six models on a "split a $4M windfall" topic converged to a 6–0 consensus and never opened a single private channel. Identical machinery, opposite outcome — the only variable was an incentive that punished going it alone.