SIG-FPT 2026-08-28

Session Recording Summary · 66m 43s · Full notes ↗

Reading: Not clearly identified in the discussion. This session was a working/build meeting for a stigmergy-themed game project and workshop rather than a discussion of a specific reading. A previously-read paper on defector ants ("if you have a defector ant, you just put down pheromone everywhere") was referenced by Patrick Nast, and Venkatesh Rao referenced Daniel Dennett's *Darwin's Dangerous Idea*, but no single shared text anchored the session.

Dan Schmidt demoed a newly-built "all-out war mode" ant-colony stigmergy simulation, in which two players tune colony parameters (evaporation rate, trail bias, gland size, cautionary pheromones) in real time while ants forage against a maze; the winner is the colony whose ants survive longest. The group discussed which mechanics would make the best workshop experience — multiple food sources, "god mode" disruption, adversarial pheromones, and whether to add termites — and how to keep the design coherent, ant-focused, and tied to the workshop's "stigmergy as new nature / security & robustness" theme. They converged on prioritizing adversarial (defector) pheromones and multiple food sources next, keeping the scope on ants, and being careful about how parallel work streams get merged.

• **Venkatesh Rao:** Multiple food sources is the single most valuable next feature — (1) fair placement is easier with 3–4 sources than one, and (2) it immediately creates an explore/exploit dynamic: a new ant leaving the nest faces multiple trails of differing strength and must choose whether to swarm the strongest scent or diversify.
• **Venkatesh Rao:** Warned against too many tunable parameters in live play (avoid 15–20); pick two or three interesting ones per tournament. Push the rich parameter space into *post-game analytics* instead — expose log data in an acquirable format so non-players can run analyses. Could run different tournaments keyed to different parameter pairs (e.g., explore vs. exploit; speed vs. evaporation).
• **UI abstraction debate (Dan):** Should parameters be exposed as raw technical terms (gland size, trail bias, evaporation) or as abstract actions (explore, exploit, disrupt)? Dan is torn — abstraction is more approachable, but learning the technical concepts is valuable, and this is a research workshop, not just a game.
• **Venkatesh Rao:** Bias toward ground-level parameters for the main sim; encourage participants to *discover and name* second-order emergent patterns themselves (analogy to Conway's Game of Life gliders/eaters, and the "ant mill"), rather than imposing preconceived labels like explore/exploit. Suggested a secondary meta-contest: watch games and identify or predictably induce patterns.
• **Patrick Nast:** Compromise via *presets* — save low-level parameter bundles as taggable configs (e.g., "exploit") that players can pick and then tweak, balancing approachability with depth.
• **Dan:** Could even let players define their own abstraction layer atop the atomic controls; the winner might be the one with the better meta-control layer.

Questions & Disagreements: - **Venkatesh Rao's puzzle:** Why did the losing colony not starve at the same rate once food ran out? Dan's log analysis suggested the losing side actually preserved a few high-energy ants longer while the winner had a "late mortality cliff" — Venkatesh found this only "sort of satisfying." Underlying question about how ants conserve energy (wandering less?) remained open. - **Abstraction vs. raw

Participants: Wip, Venkatesh Rao, Dan Schmidt, Patrick Nast, Ergod