{
 "schema": "sim-index/1",
 "generated": "2026-11-15",
 "conventions": "Each app is on its own subdomain. Pages with demos embed a manifest at script#sim-manifest (schema sim-manifest/2: page fields plus one entry per demo) and expose window.sims[demoId] (apiVersion 1); window.sim is the primary demo. Demo kinds: simulation (step/run/play/pause), evolution (getPopulation/select/breed/autoEvolve/render), calculator (sweep), static (no API). URL parameters address the primary demo unless demo=<id> is given. See llms.txt.",
 "pages": [
  {
   "id": "www",
   "kind": "page",
   "url": "https://www.bonabeauapps.com/",
   "title": "Bonabeau Apps — Simple apps for complex systems",
   "summary": "Interactive apps and data explorations by Eric Bonabeau — complex systems scientist, entrepreneur, and educator.",
   "sections": [
    "Simple apps for complex systems",
    "About the author",
    "The apps",
    "Get in touch"
   ],
   "kinds": [
    "page"
   ]
  },
  {
   "id": "abm-pattern-formation",
   "kind": "simulation",
   "url": "https://abm-pattern-formation.bonabeauapps.com/",
   "title": "Agent-Based Pattern Formation: interactive models from Bonabeau (1997)",
   "summary": "Interactive versions of the five models in Bonabeau's 1997 Artificial Life paper on pattern formation: ant cemetery clustering, Young's activation-inhibition cellular automaton, and three agent-based versions of Young's model that use perception, memory, or both. Each runs on a toroidal grid, with presets that reproduce the paper's figures.",
   "primary": "young",
   "demos": [
    {
     "id": "clustering",
     "kind": "simulation",
     "title": "Ant cemetery clustering",
     "parameters": [
      "p_clustering_W",
      "p_clustering_H",
      "p_clustering_nAgents",
      "p_clustering_nItems",
      "p_clustering_types",
      "p_clustering_T",
      "p_clustering_k1",
      "p_clustering_k2"
     ],
     "metrics": [
      "items",
      "occupancy",
      "clusters",
      "largest",
      "interface",
      "carried"
     ]
    },
    {
     "id": "young",
     "kind": "simulation",
     "title": "Young's cellular automaton",
     "parameters": [
      "p_W",
      "p_H",
      "p_density",
      "p_w1",
      "p_w2",
      "p_d1",
      "p_d2"
     ],
     "metrics": [
      "items",
      "occupancy",
      "clusters",
      "largest",
      "interface"
     ]
    },
    {
     "id": "agentSpace",
     "kind": "simulation",
     "title": "Agents, space remains space",
     "parameters": [
      "p_agentSpace_W",
      "p_agentSpace_H",
      "p_agentSpace_density",
      "p_agentSpace_nAgents",
      "p_agentSpace_w1",
      "p_agentSpace_w2",
      "p_agentSpace_d1",
      "p_agentSpace_d2"
     ],
     "metrics": [
      "items",
      "occupancy",
      "clusters",
      "largest",
      "interface",
      "carried"
     ]
    },
    {
     "id": "agentTime",
     "kind": "simulation",
     "title": "Agents, space mapped onto time",
     "parameters": [
      "p_agentTime_W",
      "p_agentTime_H",
      "p_agentTime_density",
      "p_agentTime_nAgents",
      "p_agentTime_w1",
      "p_agentTime_w2",
      "p_agentTime_t1",
      "p_agentTime_t2",
      "p_agentTime_prob",
      "p_agentTime_eta"
     ],
     "metrics": [
      "items",
      "occupancy",
      "clusters",
      "largest",
      "interface",
      "carried"
     ]
    },
    {
     "id": "combined",
     "kind": "simulation",
     "title": "Perception plus memory",
     "parameters": [
      "p_combined_W",
      "p_combined_H",
      "p_combined_density",
      "p_combined_nAgents",
      "p_combined_w1",
      "p_combined_w2",
      "p_combined_d1",
      "p_combined_t1",
      "p_combined_t2",
      "p_combined_aniso",
      "p_combined_threshold",
      "p_combined_M",
      "p_combined_prob",
      "p_combined_eta"
     ],
     "metrics": [
      "items",
      "occupancy",
      "clusters",
      "largest",
      "interface",
      "carried"
     ]
    }
   ],
   "kinds": [
    "simulation"
   ]
  },
  {
   "id": "ant-colony-tsp",
   "kind": "simulation",
   "url": "https://ant-colony-tsp.bonabeauapps.com/",
   "title": "Ant Colony Optimization on the Traveling Salesman Problem",
   "summary": "Ant colony optimization on the traveling salesman problem. Part one runs Dorigo's Ant System on random cities; part two shows, after Meuleau and Dorigo (2002), that the pheromone update is an approximation of stochastic gradient ascent, with two update rules side by side and a two-tour gradient landscape.",
   "primary": "ant-system",
   "demos": [
    {
     "id": "ant-system",
     "kind": "simulation",
     "title": "Ant System on a random traveling salesman problem",
     "parameters": [
      "nCities",
      "nAnts",
      "alpha",
      "beta",
      "rho",
      "Q",
      "elitist"
     ],
     "metrics": [
      "iteration",
      "bestLength",
      "iterBestLength",
      "maxPheromone"
     ]
    },
    {
     "id": "update-rules",
     "kind": "simulation",
     "title": "Same cities, two update rules",
     "parameters": [
      "ruleA",
      "ruleB",
      "stepA",
      "stepB",
      "cmpN",
      "cmpM",
      "cmpB",
      "cmpBaseline"
     ],
     "metrics": [
      "generation",
      "bestA",
      "meanA",
      "maxDeltaA",
      "bestB",
      "meanB",
      "maxDeltaB"
     ]
    },
    {
     "id": "gradient-landscape",
     "kind": "simulation",
     "title": "Climbing in pheromone space",
     "parameters": [
      "landAlpha"
     ],
     "metrics": [
      "antsSent",
      "w",
      "pA",
      "expectedReward",
      "lastTour",
      "lastDw",
      "exactGradient"
     ]
    }
   ],
   "kinds": [
    "simulation"
   ]
  },
  {
   "id": "army-ants",
   "kind": "simulation",
   "url": "https://army-ants.bonabeauapps.com/",
   "title": "Army ant raid patterns: an interactive explainer of Solé et al. (2000)",
   "summary": "Interactive reproduction of Solé, Bonabeau, Delgado, Fernández and Marín (2000): army ants following simple pheromone rules form column or swarm raids depending on the prey, and a macroevolutionary search over the rule parameters finds raid patterns that maximize food returned per ant. Five demos: the live raid, three landscapes side by side, the parameter search, and the two movement rules.",
   "primary": "raid",
   "demos": [
    {
     "id": "raid",
     "kind": "simulation",
     "title": "The raid, live",
     "parameters": [
      "p_q",
      "p_mu",
      "p_delta",
      "p_s1",
      "p_s2",
      "c_T",
      "c_sites",
      "c_retry",
      "c_W",
      "c_H",
      "c_phiStar",
      "c_Am",
      "c_Nb",
      "c_pfA",
      "c_FA",
      "c_pfB",
      "c_FB"
     ],
     "metrics": [
      "step",
      "ants",
      "delivered",
      "fitness",
      "food_remaining_pct",
      "front",
      "mean_width",
      "lost",
      "picked",
      "food_total"
     ]
    },
    {
     "id": "compare",
     "kind": "simulation",
     "title": "Three landscapes, one clock",
     "parameters": [
      "cmp_rules",
      "cmp_muOn",
      "cmp_mu",
      "cmp_T",
      "cmp_W",
      "cmp_H"
     ],
     "metrics": [
      "step",
      "presets",
      "F_0",
      "front_0",
      "width_0",
      "delivered_0",
      "ants_0",
      "F_1",
      "front_1",
      "width_1",
      "delivered_1",
      "ants_1",
      "F_2",
      "front_2",
      "width_2",
      "delivered_2",
      "ants_2"
     ]
    },
    {
     "id": "search",
     "kind": "simulation",
     "title": "Fitness and the search",
     "parameters": [
      "g_diet",
      "g_rules",
      "g_Ns",
      "g_G",
      "g_W",
      "g_H",
      "g_T",
      "g_tau",
      "g_eps"
     ],
     "metrics": [
      "generation",
      "runs",
      "gen_best_F",
      "gen_mean_F",
      "survivors",
      "best_F",
      "best_q",
      "best_mu",
      "best_delta",
      "best_s1",
      "best_s2",
      "complete"
     ]
    },
    {
     "id": "rule1-move",
     "kind": "calculator",
     "title": "Rule 1: how much trail makes an ant move",
     "parameters": [
      "ex1_S",
      "ex1_phi"
     ],
     "metrics": [
      "P_move"
     ]
    },
    {
     "id": "rule2-direction",
     "kind": "calculator",
     "title": "Rule 2: which front site gets picked",
     "parameters": [
      "ex2_L",
      "ex2_0",
      "ex2_R",
      "ex2_mu"
     ],
     "metrics": [
      "p_left",
      "p_center",
      "p_right"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "calculator"
   ]
  },
  {
   "id": "biomorphs",
   "kind": "evolution",
   "url": "https://biomorphs.bonabeauapps.com/",
   "title": "Biomorphs: interactive evolution",
   "summary": "Breed Dawkins's biomorphs by eye: pick the offspring you prefer and it becomes the next parent. Optional gene groups add symmetry, segmentation and gradients, and several parents can be recombined.",
   "primary": "biomorphs",
   "demos": [
    {
     "id": "biomorphs",
     "kind": "evolution",
     "title": "Biomorphs",
     "parameters": [
      "allowSym",
      "allowSeg",
      "allowGrad",
      "xtype"
     ],
     "metrics": [
      "generation",
      "parents",
      "selected"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "informationtheory",
   "kind": "calculator",
   "url": "https://informationtheory.bonabeauapps.com/",
   "title": "Bits of Intuition: an interactive tour of information theory",
   "summary": "Eight small interactive levels on information theory, from a jar of marbles to Shannon's noisy channel coding theorem: information as ruled-out probability mass, entropy and prefix codes, cross-entropy and KL divergence, typical sequences, mutual information, entropy rate of a Markov chain, channel capacity, and error-correcting codes.",
   "primary": "surprise",
   "demos": [
    {
     "id": "surprise",
     "kind": "calculator",
     "title": "Level 1: ruling out probability mass",
     "parameters": [
      "l1clue"
     ],
     "metrics": [
      "surviving_mass_32nds",
      "surviving_probability",
      "information_bits",
      "quiz_bits",
      "outcomes_surviving"
     ]
    },
    {
     "id": "prefix-code",
     "kind": "calculator",
     "title": "Level 2: build the shortest prefix code",
     "parameters": [],
     "metrics": [
      "avg_bits",
      "code_valid",
      "code_status",
      "length_blue",
      "length_gray",
      "length_yellow",
      "length_green",
      "sample_bits",
      "entropy_floor_reached"
     ]
    },
    {
     "id": "cross-entropy",
     "kind": "calculator",
     "title": "Level 3: cross-entropy and KL divergence",
     "parameters": [
      "l3speaker",
      "l3code"
     ],
     "metrics": [
      "avg_length",
      "speaker_entropy",
      "kl_penalty"
     ]
    },
    {
     "id": "typical-sequences",
     "kind": "calculator",
     "title": "Level 4: typical sequences (AEP)",
     "parameters": [
      "l4slider"
     ],
     "metrics": [
      "n",
      "sequences",
      "peak_count_info",
      "peak_mass_info",
      "mass_1_5_to_2"
     ]
    },
    {
     "id": "mutual-information",
     "kind": "calculator",
     "title": "Level 5: mutual information",
     "parameters": [
      "l5coupling"
     ],
     "metrics": [
      "mutual_information",
      "h_x_given_y",
      "joint_entropy",
      "colors_per_shape"
     ]
    },
    {
     "id": "sticky-urn",
     "kind": "simulation",
     "title": "Level 6: the sticky urn (entropy rate)",
     "parameters": [
      "l6slider"
     ],
     "metrics": [
      "stickiness",
      "entropy_rate",
      "draws",
      "repeats",
      "repeat_share",
      "last_color",
      "round",
      "score"
     ]
    },
    {
     "id": "channel-capacity",
     "kind": "calculator",
     "title": "Level 7: Z-channel capacity",
     "parameters": [
      "l7slider"
     ],
     "metrics": [
      "p_input_1",
      "mutual_information",
      "h_y",
      "h_y_given_x",
      "capacity",
      "capacity_p_input_1",
      "capacity_found"
     ]
    },
    {
     "id": "noisy-channel",
     "kind": "simulation",
     "title": "Level 8: noisy channel coding",
     "parameters": [
      "l8mode"
     ],
     "metrics": [
      "trials",
      "messages_wrong",
      "message_error_rate",
      "marbles_wrong",
      "bit_flips",
      "transmitted_bits",
      "code_rate",
      "last_marbles_wrong",
      "last_bulk_error_rate"
     ]
    }
   ],
   "kinds": [
    "calculator",
    "simulation"
   ]
  },
  {
   "id": "breeding-metamaterials",
   "kind": "evolution",
   "url": "https://breeding-metamaterials.bonabeauapps.com/",
   "title": "Breeding network metamaterials: interactive evolution of scale-rich networks",
   "summary": "Interactive evolution of mechanical network metamaterials on the scale-rich growth model of Both et al. (Nature, 2026): breed 2D ligament networks by eye or toward a declared goal such as auxetic in every direction, breed 3D cylinder and plate networks, inspect one specimen from genes to response, and recompute Figures 2, 3, 5 and 6 of the paper with reduced beam models.",
   "primary": "breeder",
   "demos": [
    {
     "id": "breeder",
     "kind": "evolution",
     "title": "The breeder (2D)",
     "parameters": [
      "goalSel",
      "rhoMin",
      "rhoMax",
      "rTopo",
      "rGeo",
      "rPar",
      "cutsChk",
      "guideChk",
      "rCross",
      "screenSel",
      "famSel",
      "oriSel",
      "denSel",
      "payChk",
      "tGenes",
      "popSel"
     ],
     "metrics": [
      "generation",
      "selected",
      "goal_best",
      "nu_max_best",
      "nu_max_median"
     ]
    },
    {
     "id": "breeder3d",
     "kind": "evolution",
     "title": "Three dimensions: a small breeder",
     "parameters": [
      "d3kind",
      "d3ori",
      "d3T",
      "d3a",
      "d3l"
     ],
     "metrics": [
      "generation",
      "selected",
      "nu_max_best",
      "iso_best"
     ]
    },
    {
     "id": "bench",
     "kind": "calculator",
     "title": "Specimen bench",
     "parameters": [
      "benchAlpha",
      "benchLambda0",
      "benchEps",
      "benchFam",
      "benchOri",
      "benchDen",
      "benchPay"
     ],
     "metrics": [
      "solvable",
      "load_path",
      "rho",
      "members",
      "genes_expressed",
      "cut_segments",
      "nu_min",
      "nu_mean",
      "nu_max",
      "nu_bench_v",
      "nu_bench_h",
      "nu_worst",
      "E_max",
      "E_min",
      "iso",
      "geo_aniso",
      "stiff_dir_deg",
      "shear_per_squeeze",
      "n_eff",
      "wave_iso",
      "loose_area",
      "payload_hold",
      "payload_stress_v",
      "payload_stress_h",
      "ls_strain_end",
      "ls_peak_stress",
      "ls_loc_end",
      "ls_work_end",
      "ls_first_cell_closes"
     ]
    },
    {
     "id": "fig2",
     "kind": "calculator",
     "title": "Figure 2: directional stiffness, one scale against many",
     "parameters": [
      "f2rho",
      "f2T",
      "f2real"
     ],
     "metrics": [
      "hex_rho",
      "hex_iso",
      "voronoi_rho",
      "voronoi_iso",
      "square_rho",
      "square_iso",
      "sr_a0_rho",
      "sr_a0_iso",
      "sr_a03_rho",
      "sr_a03_iso",
      "sr_a09_rho",
      "sr_a09_iso",
      "cloud_n",
      "cloud_iso_min",
      "cloud_iso_max",
      "cloud_fold"
     ]
    },
    {
     "id": "fig3",
     "kind": "calculator",
     "title": "Figures 3 and 4d: squeeze, buckling, and where the deformation goes",
     "parameters": [
      "f3rho",
      "f3T",
      "f3real"
     ],
     "metrics": [
      "sr_a0_rho",
      "sr_a0_iso",
      "sr_a0_peak",
      "sr_a0_loc",
      "sr_a0_close",
      "square_rho",
      "square_iso",
      "square_peak",
      "square_loc",
      "square_close",
      "hex_rho",
      "hex_iso",
      "hex_peak",
      "hex_loc",
      "hex_close",
      "voronoi_rho",
      "voronoi_iso",
      "voronoi_peak",
      "voronoi_loc",
      "voronoi_close",
      "tri_rho",
      "tri_iso",
      "tri_peak",
      "tri_loc",
      "tri_close",
      "kagome_rho",
      "kagome_iso",
      "kagome_peak",
      "kagome_loc",
      "kagome_close",
      "gilbert_rho",
      "gilbert_iso",
      "gilbert_peak",
      "gilbert_loc",
      "gilbert_close",
      "sr_a05_rho",
      "sr_a05_iso",
      "sr_a05_peak",
      "sr_a05_loc",
      "sr_a05_close",
      "patterns_solved"
     ]
    },
    {
     "id": "fig5",
     "kind": "calculator",
     "title": "Figure 5: refractive index and a graded lens",
     "parameters": [
      "f5real"
     ],
     "metrics": [
      "lens_built",
      "library_cells",
      "sampled_cells",
      "n_min",
      "n_max",
      "ring_offset",
      "focal_spot_R",
      "peak_graded",
      "peak_shuffled",
      "half_width_d"
     ]
    },
    {
     "id": "fig6",
     "kind": "calculator",
     "title": "Figure 6: orientation fields, density fields and a protected payload",
     "parameters": [
      "f6real"
     ],
     "metrics": [
      "square_rho",
      "square_stress_v",
      "square_stress_h",
      "square_hold",
      "hex_rho",
      "hex_stress_v",
      "hex_stress_h",
      "hex_hold",
      "voronoi_rho",
      "voronoi_stress_v",
      "voronoi_stress_h",
      "voronoi_hold",
      "sr_uniform_rho",
      "sr_uniform_stress_v",
      "sr_uniform_stress_h",
      "sr_uniform_hold",
      "sr_circ_1_rho",
      "sr_circ_1_stress_v",
      "sr_circ_1_stress_h",
      "sr_circ_1_hold",
      "sr_circ_2_rho",
      "sr_circ_2_stress_v",
      "sr_circ_2_stress_h",
      "sr_circ_2_hold",
      "sr_circ_3_rho",
      "sr_circ_3_stress_v",
      "sr_circ_3_stress_h",
      "sr_circ_3_hold"
     ]
    }
   ],
   "kinds": [
    "evolution",
    "calculator"
   ]
  },
  {
   "id": "facial-composites",
   "kind": "evolution",
   "url": "https://facial-composites.bonabeauapps.com/",
   "title": "Composite evolver: interactive evolution of a suspect face",
   "summary": "Study a target face, hide it, then breed a likeness from memory by choosing among whole faces, as in the EvoFIT procedure (Frowd et al., 2005); three breeding strategies from Zahradnikova et al. (2017) can be compared, and Finish and compare scores the composite's likeness to the target.",
   "primary": "composite-evolver",
   "demos": [
    {
     "id": "composite-evolver",
     "kind": "evolution",
     "title": "Composite evolver",
     "parameters": [
      "sex",
      "strategy",
      "popSize",
      "pm",
      "step",
      "shrink",
      "scope"
     ],
     "metrics": [
      "generation",
      "faces_viewed",
      "current_step"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "choice-decision",
   "kind": "simulation",
   "url": "https://choice-decision.bonabeauapps.com/",
   "title": "Diagnostic choice tasks: an interactive guide to Pfeiffer et al. (2014)",
   "summary": "Interactive guide to Pfeiffer, Duzevik, Rothlauf, Bonabeau and Yamamoto (2014), An optimized design of choice experiments (Journal of Behavioral Decision Making): a genetic algorithm builds, for each respondent, a phone comparison in which four decision rules (WAD, MAJ, LEX, EBA) each select a different phone, so one choice reveals the rule. The page lets you apply the rules to a comparison, count the possible tasks, see how similarity and conflict are measured, run the GA against random sampling for a synthetic respondent, and read the study's results.",
   "primary": "ga-search",
   "demos": [
    {
     "id": "ga-search",
     "kind": "simulation",
     "title": "Run the search in your browser",
     "parameters": [
      "respondent",
      "alternatives",
      "attributes",
      "similarity",
      "conflict",
      "budget"
     ],
     "metrics": [
      "generation",
      "generations",
      "tasks_evaluated",
      "ga_best",
      "ga_mean",
      "rs_best",
      "finished",
      "ga_f",
      "ga_f1",
      "ga_f_robust",
      "ga_ar",
      "ga_ac",
      "ga_optimal",
      "ga_selects",
      "rs_f",
      "rs_f1",
      "rs_f_robust",
      "rs_ar",
      "rs_ac",
      "rs_optimal",
      "rs_selects"
     ]
    },
    {
     "id": "random-tally",
     "kind": "simulation",
     "title": "How often does a random task have an optimal mapping?",
     "parameters": [],
     "metrics": [
      "respondent",
      "tasks_drawn",
      "optimal",
      "ambiguous_only",
      "multiple_only",
      "ambiguous_and_multiple",
      "share_optimal"
     ]
    },
    {
     "id": "hero-comparison",
     "kind": "calculator",
     "title": "Same respondent, two versions of a phone comparison",
     "parameters": [
      "heroVersion"
     ],
     "metrics": [
      "pick_WAD",
      "pick_MAJ",
      "pick_LEX",
      "pick_EBA",
      "other",
      "f1",
      "optimal"
     ]
    },
    {
     "id": "rule-explorer",
     "kind": "calculator",
     "title": "Apply the four rules to one comparison",
     "parameters": [
      "imp_weight",
      "imp_form",
      "imp_memory",
      "imp_gps",
      "imp_touch"
     ],
     "metrics": [
      "pick_WAD",
      "pick_MAJ",
      "pick_LEX",
      "pick_EBA",
      "other",
      "f1",
      "optimal",
      "wad_scores",
      "maj_counts",
      "cells_read_WAD",
      "cells_read_MAJ",
      "cells_read_LEX",
      "cells_read_EBA"
     ]
    },
    {
     "id": "task-space",
     "kind": "calculator",
     "title": "How many choice tasks are there?",
     "parameters": [
      "spN",
      "spM",
      "spL"
     ],
     "metrics": [
      "log10_tasks",
      "tasks",
      "log10_enumeration_years",
      "enumeration_text"
     ]
    },
    {
     "id": "similarity-conflict",
     "kind": "calculator",
     "title": "Six products on two attributes",
     "parameters": [
      "scS",
      "scR"
     ],
     "metrics": [
      "spread",
      "correlation",
      "trade_off_pairs",
      "dominated",
      "similarity_label",
      "conflict_label"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "calculator"
   ]
  },
  {
   "id": "ant-scheduling",
   "kind": "simulation",
   "url": "https://ant-scheduling.bonabeauapps.com/",
   "title": "Dynamic Scheduling and Division of Labor in Social Insects: interactive explainer",
   "summary": "Interactive explainer of Campos, Bonabeau, Theraulaz and Deneubourg (2001), Dynamic Scheduling and Division of Labor in Social Insects: trucks leave an assembly line once a minute and paint booths decide which booth takes each truck, either by bidding (a market rule) or by response thresholds that adapt with experience (an ant rule). A reimplementation runs both rules on the same trucks and breakdowns, repeats the experiment over many runs, and plots both rules' formulas.",
   "primary": "paint-shop",
   "demos": [
    {
     "id": "paint-shop",
     "kind": "simulation",
     "title": "Run the paint shop",
     "parameters": [
      "testSet",
      "boothSlider",
      "market_P",
      "market_L",
      "market_C",
      "ant_alpha",
      "ant_beta",
      "ant_xi",
      "ant_phi",
      "ant_tmin",
      "ant_tmax",
      "idleSel"
     ],
     "metrics": [
      "minute",
      "total_trucks",
      "flushes_market",
      "started_market",
      "in_storage_market",
      "storage_used_market",
      "finish_market",
      "done_market",
      "flushes_ant",
      "started_ant",
      "in_storage_ant",
      "storage_used_ant",
      "finish_ant",
      "done_ant",
      "specialist_cells"
     ]
    },
    {
     "id": "paired-runs",
     "kind": "simulation",
     "title": "Repeat the experiment",
     "parameters": [],
     "metrics": [
      "runs",
      "test_set",
      "booths",
      "mean_flushes_market",
      "sd_flushes_market",
      "mean_finish_market",
      "sd_finish_market",
      "storage_runs_market",
      "unfinished_market",
      "mean_flushes_ant",
      "sd_flushes_ant",
      "mean_finish_ant",
      "sd_finish_ant",
      "storage_runs_ant",
      "unfinished_ant",
      "paired_diff_mean",
      "paired_diff_se",
      "ant_fewer_runs",
      "paper_flushes_market",
      "paper_flushes_ant",
      "paper_finish_market",
      "paper_finish_ant"
     ]
    },
    {
     "id": "rule-curves",
     "kind": "calculator",
     "title": "Two rules for choosing a booth",
     "parameters": [
      "wSlider"
     ],
     "metrics": [
      "tie_wait",
      "response_ratio",
      "response_specialist",
      "response_nonspecialist"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "calculator"
   ]
  },
  {
   "id": "abmestimation",
   "kind": "simulation",
   "url": "https://abmestimation.bonabeauapps.com/",
   "title": "Estimating agent-based models by matching random features",
   "summary": "Explainer of simulation-based estimation of agent-based models by matching random Fourier features (Shalizi 2021; Wieck-Sosa and Shalizi 2026): an agent-based model defines a distribution one can only sample from; 2p + 1 randomly chosen features identify its p parameters; their time averages on observed and simulated data are matched by a global search. A 32 x 32 grid epidemic with unknown transmission and recovery probabilities is the running example, estimated live in the browser.",
   "primary": "live-estimation",
   "demos": [
    {
     "id": "abm-run",
     "kind": "simulation",
     "title": "Run the grid epidemic (step 1)",
     "parameters": [
      "abmB",
      "abmG"
     ],
     "metrics": [
      "runs_completed",
      "run_day",
      "run_beta",
      "run_gamma",
      "positive_pct",
      "new_cases",
      "mean_positive_pct",
      "mean_new_cases",
      "infected"
     ]
    },
    {
     "id": "random-feature",
     "kind": "calculator",
     "title": "One random Fourier feature (step 4)",
     "parameters": [
      "rfM",
      "rfDraw"
     ],
     "metrics": [
      "omega",
      "alpha",
      "sample_average",
      "expected_value",
      "other_mu_count"
     ]
    },
    {
     "id": "feature-embedding",
     "kind": "calculator",
     "title": "Two features versus three (step 5)",
     "parameters": [
      "embDraw"
     ],
     "metrics": [
      "crossings_2d",
      "crossing_mu_a",
      "crossing_mu_b",
      "min_distance_3d",
      "omegas",
      "alphas"
     ]
    },
    {
     "id": "time-average",
     "kind": "calculator",
     "title": "Time averages converge (step 6)",
     "parameters": [
      "taT"
     ],
     "metrics": [
      "running_average",
      "other_runs_running_average",
      "long_run_value"
     ]
    },
    {
     "id": "rolling-window",
     "kind": "calculator",
     "title": "Rolling windows for a one-off outbreak (step 6)",
     "parameters": [
      "rwB"
     ],
     "metrics": [
      "rolling_mismatch",
      "global_mismatch"
     ]
    },
    {
     "id": "live-estimation",
     "kind": "simulation",
     "title": "Live estimation of the grid epidemic (step 8)",
     "parameters": [
      "dmB",
      "dmG",
      "dmFeatDraw"
     ],
     "metrics": [
      "estimates",
      "search_generation",
      "beta_hat",
      "gamma_hat",
      "q",
      "observed_features",
      "best_features",
      "bootstrap_replicates",
      "bootstrap_sd_beta",
      "bootstrap_sd_gamma",
      "map_min_beta",
      "map_min_gamma",
      "map_min_q"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "calculator"
   ]
  },
  {
   "id": "evodesign",
   "kind": "evolution",
   "url": "https://evodesign.bonabeauapps.com/",
   "title": "evoDesign: interactive evolution of tiles",
   "summary": "A reproduction of evoDesign (Anderson, Buchsbaum, Potter and Bonabeau, 2008): breed seamless repeating tiles made of circles, rectangles, stripes and color grids by selecting the ones you like and applying mutation, crossover or both.",
   "primary": "evodesign",
   "demos": [
    {
     "id": "evodesign",
     "kind": "evolution",
     "title": "evoDesign: evolving tiles",
     "parameters": [
      "mode",
      "pCross",
      "popSize",
      "nComp",
      "w_rect",
      "w_hstripe",
      "w_grid",
      "w_vstripe",
      "w_circle",
      "r_xy",
      "r_rgb",
      "r_a",
      "r_d",
      "r_wh",
      "pRemove",
      "pAdd",
      "alphaMin",
      "rectMax",
      "maxComp",
      "shuffle",
      "keepParents"
     ],
     "metrics": [
      "generation",
      "selected",
      "branches"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "evolving-particle-swarms",
   "kind": "simulation",
   "url": "https://evolving-particle-swarms.bonabeauapps.com/",
   "title": "Evolving a map of swarm behaviors",
   "summary": "A quality-diversity (MAP-Elites) search over a nine-parameter self-propelled particle model, after Wen and Bonabeau (Physica A): every simulated swarm is placed in a 12 by 12 grid of alignment and milling or clustering, and each cell keeps its most stable rule. A single-objective optimizer and random sampling fill the same grid for comparison.",
   "primary": "qd-search",
   "demos": [
    {
     "id": "qd-search",
     "kind": "simulation",
     "title": "Behavior-grid search over a self-propelled particle model",
     "parameters": [
      "strategy",
      "plane",
      "objSel",
      "budgetSel"
     ],
     "metrics": [
      "evaluations",
      "cells_reached",
      "mean_stability",
      "best_objective",
      "max_stability",
      "evaluations_left",
      "last_cell",
      "last_outcome"
     ]
    }
   ],
   "kinds": [
    "simulation"
   ]
  },
  {
   "id": "interactive-inversion",
   "kind": "evolution",
   "url": "https://interactive-inversion.bonabeauapps.com/",
   "title": "Interactive evolution as model inversion",
   "summary": "Interactive evolution as a way to invert forward models: four labs where an observer picks or ranks candidates that look closest to a target, after Ashburn & Bonabeau (CEC 2004) and Ecemis, Bonabeau & Ashburn (GECCO 2005) on an agent-based financial market, Boschetti & Moresi (2001) on mantle convection, Wijns, Boschetti & Moresi (2003) on fault spacing, and Singh, Minsker & Valocchi (2008) on aquifer conductivity with three objectives.",
   "primary": "market",
   "demos": [
    {
     "id": "market",
     "kind": "evolution",
     "title": "Lab 1. Market frenzy: inverting an agent-based market model",
     "parameters": [
      "elitism",
      "mut",
      "mutStyle",
      "cross",
      "modules",
      "modT",
      "targetSel"
     ],
     "metrics": [
      "generation",
      "best_observer_distance",
      "best_E",
      "best_ls_rank",
      "selected",
      "training_patterns",
      "ga_E",
      "ga_observer_distance"
     ]
    },
    {
     "id": "geo",
     "kind": "evolution",
     "title": "Lab 2. Mantle convection under a crust (Boschetti & Moresi 2001)",
     "parameters": [],
     "metrics": [
      "generation",
      "ranked",
      "best_misfit",
      "rank1_misfit",
      "rank1_x0",
      "numeric_misfit",
      "numeric_x0"
     ]
    },
    {
     "id": "faults",
     "kind": "evolution",
     "title": "Lab 3. Fault spacing during crustal extension (Wijns, Boschetti & Moresi 2003)",
     "parameters": [],
     "metrics": [
      "generation",
      "ranked",
      "spread_eta",
      "spread_Ea",
      "spread_e0",
      "spread_En",
      "settled",
      "best_critic"
     ]
    },
    {
     "id": "aquifer",
     "kind": "evolution",
     "title": "Lab 4. Aquifer conductivity with three objectives (Singh, Minsker & Valocchi 2008)",
     "parameters": [],
     "metrics": [
      "generation",
      "front_size",
      "rank1_count",
      "true_Herr",
      "true_Kerr",
      "true_dominated",
      "phase1_minH_rmse",
      "phase1_minK_rmse",
      "rank1_rmse",
      "rank1_Herr",
      "rank1_Kerr"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "multi-participant-workload-partitioning",
   "kind": "evolution",
   "url": "https://multi-participant-workload-partitioning.bonabeauapps.com/",
   "title": "Interactive evolution for workload partitioning",
   "summary": "A working model of interactive evolution for dividing work among the people who do it: each round the planner prepares six partitions, every participant ranks only their own piece, and the rankings choose the parents of the next round. Two examples with simulated participants: five mail carriers dividing a town into routes (after the Accord field process for La Poste), and a nurse roster scored as in Funes, Bonabeau, Hervé and Morieux (CEC 2004).",
   "primary": "mail",
   "demos": [
    {
     "id": "mail",
     "kind": "evolution",
     "title": "Mail carriers: dividing a town into five routes",
     "parameters": [
      "tol",
      "T"
     ],
     "metrics": [
      "generation",
      "best_cost",
      "mean_cost",
      "best_dissatisfaction",
      "mean_dissatisfaction",
      "control_mean_cost",
      "control_mean_dissatisfaction",
      "shown_plan",
      "shown_plan_cost",
      "shown_plan_dissatisfaction",
      "rank_load_correlation",
      "rankings_given",
      "vote_winner"
     ]
    },
    {
     "id": "nurse",
     "kind": "evolution",
     "title": "Nurse roster: 21 shifts for four nurses",
     "parameters": [],
     "metrics": [
      "generation",
      "best_cost",
      "mean_cost",
      "best_dissatisfaction",
      "mean_dissatisfaction",
      "control_mean_cost",
      "control_mean_dissatisfaction",
      "shown_plan",
      "shown_plan_cost",
      "shown_plan_dissatisfaction",
      "rank_load_correlation",
      "rankings_given",
      "vote_winner"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "interactive-evolution-molecular-space",
   "kind": "evolution",
   "url": "https://interactive-evolution-molecular-space.bonabeauapps.com/",
   "title": "Interactive evolution in molecular space",
   "summary": "A working model of interactive evolutionary computation for drug candidate design, after the Mobius environment of Ecemiş, Wikel, Bingham and Bonabeau (IEEE Trans. Evol. Comput., 2008): a genetic algorithm searches compounds assembled from chemist-chosen fragments on a quinolone scaffold, trapezoid property models score them, and the chemist steers the search by reshaping the models and liking or banning proposals.",
   "primary": "molecular-ga",
   "demos": [
    {
     "id": "molecular-ga",
     "kind": "evolution",
     "title": "Genetic algorithm over a quinolone fragment space, with chemist feedback",
     "parameters": [
      "pN",
      "pE",
      "pC",
      "pM",
      "lambda",
      "likeVal",
      "simChem",
      "simEvery",
      "simRule",
      "w_mw",
      "w_logp",
      "w_psa",
      "w_hbd",
      "w_hba",
      "w_rotb",
      "R1_H",
      "R1_Me",
      "R1_Et",
      "R1_cPr",
      "R1_tBu",
      "R1_Ph",
      "R1_diFPh",
      "R1_EtOH",
      "R1_EtF",
      "R2_H",
      "R2_F",
      "R2_Cl",
      "R2_Br",
      "R2_OMe",
      "R2_NH2",
      "R2_NO2",
      "R2_Me",
      "R3_H",
      "R3_Cl",
      "R3_F",
      "R3_Pip",
      "R3_MePip",
      "R3_EtPip",
      "R3_2MePip",
      "R3_Pyrr",
      "R3_NH2Pyrr",
      "R3_AmPyrr",
      "R3_Pipd",
      "R3_OHPipd",
      "R3_Morph",
      "R3_SMorph",
      "R3_Azet",
      "R3_NMe2",
      "R3_OMe",
      "R3_Ph",
      "R3_ClPh",
      "R3_CF3",
      "R4_H",
      "R4_F",
      "R4_Cl",
      "R4_OMe",
      "R4_Me",
      "R4_OCHF2",
      "R4_CN"
     ],
     "metrics": [
      "generation",
      "best_fitness",
      "average_fitness",
      "unique_evaluated",
      "space_size",
      "exceptional_found",
      "exceptional_total",
      "top_score_in_space",
      "chlorine_share",
      "mean_heteroatoms",
      "liked",
      "banned",
      "ratings_given",
      "selected"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "sims-interactive-evolution",
   "kind": "evolution",
   "url": "https://sims-interactive-evolution.bonabeauapps.com/",
   "title": "Interactive evolution of images, after Karl Sims",
   "summary": "A browser replication of Karl Sims's interactive evolution of images (1991, 1993): each picture is a Lisp expression evaluated at every pixel, and you breed them by picking one image to mutate or two to mate.",
   "primary": "sims-interactive-evolution",
   "demos": [
    {
     "id": "sims-interactive-evolution",
     "kind": "evolution",
     "title": "Interactive evolution of images",
     "parameters": [
      "rate"
     ],
     "metrics": [
      "generation",
      "selected"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "sims-reaction-diffusion",
   "kind": "evolution",
   "url": "https://sims-reaction-diffusion.bonabeauapps.com/",
   "title": "Interactive evolution of reaction-diffusion systems",
   "summary": "Twelve Gray-Scott reaction-diffusion simulations run side by side, each with its own feed and kill rates, scale, flow, orientation and colors. Pick the ones you like and breed the next generation from them.",
   "primary": "sims-reaction-diffusion",
   "demos": [
    {
     "id": "sims-reaction-diffusion",
     "kind": "evolution",
     "title": "Interactive evolution of reaction-diffusion systems",
     "parameters": [
      "mr",
      "md",
      "inherit"
     ],
     "metrics": [
      "generation",
      "selected",
      "steps",
      "meanB",
      "tilesWithPattern"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "swarm-building",
   "kind": "simulation",
   "url": "https://swarm-building.bonabeauapps.com/",
   "title": "Lattice swarms: stigmergic building and interactive evolution of architectures",
   "summary": "Lattice swarms after Theraulaz and Bonabeau (Science and J. Theor. Biol., 1995) and Bonabeau et al. (BioSystems, 2000): agents walking on a 3D lattice deposit bricks when the configuration around them matches a shared rule table, so coordinated tables grow wasp-nest architectures without any plan. Two demos are scriptable: the lattice swarm with the published, inferred, random or custom tables, and the interactive evolution of building algorithms, with the papers' fitness F.",
   "primary": "lattice-swarm",
   "demos": [
    {
     "id": "lattice-swarm",
     "kind": "simulation",
     "title": "Lattice swarm building with a rule table",
     "parameters": [
      "alg1",
      "sym1",
      "size1",
      "agents1"
     ],
     "metrics": [
      "step",
      "bricks",
      "bricks_type1",
      "bricks_type2",
      "last_deposit_step",
      "rules",
      "rules_used",
      "lookup_entries",
      "conflicts",
      "edges",
      "extent_x",
      "extent_y",
      "extent_z",
      "current_stage",
      "stages"
     ]
    },
    {
     "id": "evolution",
     "kind": "evolution",
     "title": "Interactive evolution of lattice swarm algorithms",
     "parameters": [
      "latKind",
      "xType",
      "p1",
      "p2",
      "pc",
      "elite",
      "maxBricks",
      "maxSteps",
      "agents2"
     ],
     "metrics": [
      "generation",
      "best_F",
      "mean_F",
      "mean_bricks",
      "evaluated",
      "selected",
      "lattice"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "evolution"
   ]
  },
  {
   "id": "longevity",
   "kind": "calculator",
   "url": "https://longevity.bonabeauapps.com/",
   "title": "Longevity Priority Matrix",
   "summary": "An evidence-weighted ranking of 86 longevity interventions (lifestyle, diet, supplements, drugs, procedures and more). Each has fixed 0-to-5 ratings for healthspan and maximum-lifespan evidence, effectiveness and magnitude, plus safety and durability; adjustable weights combine them into healthspan, max-lifespan and overall scores and a priority tier, shown as tiers, a scatter plot and a sortable table.",
   "primary": "priority-matrix",
   "demos": [
    {
     "id": "priority-matrix",
     "kind": "calculator",
     "title": "Longevity Priority Matrix",
     "parameters": [
      "slider-hsEvidence",
      "slider-hsEffectiveness",
      "slider-hsMagnitude",
      "slider-hsSafety",
      "slider-hsDurability",
      "slider-mlEvidence",
      "slider-mlEffectiveness",
      "slider-mlMagnitude",
      "slider-mlSafety",
      "slider-mlDurability",
      "slider-hsWeight"
     ],
     "metrics": [
      "n_interventions",
      "tier_highest",
      "tier_high",
      "tier_selective",
      "tier_low",
      "tier_avoid",
      "ml_weight",
      "hs_weight_sum",
      "ml_weight_sum",
      "top_intervention",
      "top_overall"
     ]
    }
   ],
   "kinds": [
    "calculator"
   ]
  },
  {
   "id": "margin-liquidation-cascade",
   "kind": "calculator",
   "url": "https://margin-liquidation-cascade.bonabeauapps.com/",
   "title": "Margin-liquidation cascades, step by step",
   "summary": "Step-by-step explainer of E. Bonabeau, Finite-size effects, heavy tails and transient impact in absorbing margin-liquidation cascades (Physical Review E, 2026): leveraged positions are margin-called when the price falls past a leverage threshold, their forced sales move the price, and the breached dollar fraction obeys a fixed-point equation whose low state disappears at a fold. Interactive figures compute one balance sheet, the breached tail, the exposure tilt, the cascade fixed point for a chosen shock, the random collapse point of finite cohorts, and lognormal versus Pareto weights.",
   "primary": "cascade",
   "demos": [
    {
     "id": "cascade",
     "kind": "calculator",
     "title": "Closing the loop: the fixed-point map and its collapse (step 5)",
     "parameters": [
      "cw-d0"
     ],
     "metrics": [
      "fixed_point_count",
      "landed_B",
      "rounds",
      "lowest_fixed_point",
      "fixed_points",
      "past_upper_fold"
     ]
    },
    {
     "id": "balance-sheet",
     "kind": "calculator",
     "title": "A balance sheet under a drawdown (step 1)",
     "parameters": [
      "bs-l",
      "bs-d",
      "bs-m"
     ],
     "metrics": [
      "equity_ratio",
      "in_breach",
      "threshold_leverage",
      "assets",
      "debt",
      "equity"
     ]
    },
    {
     "id": "breached-tail",
     "kind": "calculator",
     "title": "The breached tail of a leverage distribution (step 2)",
     "parameters": [
      "dist-d"
     ],
     "metrics": [
      "breached_fraction",
      "threshold_leverage",
      "z_cutoff"
     ]
    },
    {
     "id": "exposure-tilt",
     "kind": "calculator",
     "title": "Exposure tilt: count dollars, not accounts (step 3)",
     "parameters": [
      "tilt-chi"
     ],
     "metrics": [
      "F_tilt_d010",
      "F_flat_d010",
      "F_tilt_d015",
      "F_flat_d015"
     ]
    },
    {
     "id": "cohorts",
     "kind": "simulation",
     "title": "A finite cohort's collapse point (step 10)",
     "parameters": [
      "co-n"
     ],
     "metrics": [
      "N",
      "cohorts",
      "collapse_shock",
      "takeoff_B",
      "first_breach_highest",
      "mean_collapse",
      "sd_collapse"
     ]
    },
    {
     "id": "weights",
     "kind": "simulation",
     "title": "Lognormal versus Pareto weights of 60 accounts (step 11)",
     "parameters": [],
     "metrics": [
      "largest_share_lognormal",
      "largest_share_pareto"
     ]
    }
   ],
   "kinds": [
    "calculator",
    "simulation"
   ]
  },
  {
   "id": "marginally-stable-swarms",
   "kind": "simulation",
   "url": "https://marginally-stable-swarms.bonabeauapps.com/",
   "title": "Marginally stable swarms are flexible and efficient",
   "summary": "Interactive companion to Bonabeau (1996) and its 1997 follow-up: an ant colony foraging from two sources can both commit to a trail and switch to a richer source only if it sits just past a bifurcation. Five demos: a live stochastic colony, a sweep over colony size N, the width of the switching window against the quality gap, a double-well potential, and the group-versus-mass recruitment model.",
   "primary": "live-colony",
   "demos": [
    {
     "id": "live-colony",
     "kind": "simulation",
     "title": "Watch a colony choose",
     "parameters": [
      "lN",
      "lD",
      "lX",
      "lExp"
     ],
     "metrics": [
      "N1",
      "N2",
      "lost",
      "nest",
      "M",
      "source2_available",
      "verdict"
     ]
    },
    {
     "id": "bifurcation-sweep",
     "kind": "simulation",
     "title": "Sweep N and the three behaviors appear",
     "parameters": [
      "sExp",
      "sD",
      "sX"
     ],
     "metrics": [
      "points",
      "last_N",
      "last_M",
      "critical_N",
      "window_lo",
      "window_hi",
      "window_S",
      "complete"
     ]
    },
    {
     "id": "window-scaling",
     "kind": "simulation",
     "title": "How wide is the flexible window?",
     "parameters": [
      "wX"
     ],
     "metrics": [
      "deltas_done",
      "last_delta",
      "last_S",
      "fit_points",
      "slope",
      "delta_exponent",
      "complete"
     ]
    },
    {
     "id": "potential",
     "kind": "simulation",
     "title": "Why being close to the instability helps",
     "parameters": [
      "pH",
      "pA",
      "pB"
     ],
     "metrics": [
      "M_A",
      "M_B",
      "threshold_A",
      "threshold_B",
      "switched_A",
      "switched_B"
     ]
    },
    {
     "id": "recruitment-mix",
     "kind": "calculator",
     "title": "Group recruitment, mass recruitment, and the edge between them",
     "parameters": [
      "cB"
     ],
     "metrics": [
      "outcome",
      "switch_from_N",
      "last_trapped_N",
      "M_at_N400",
      "N1_at_N400",
      "N2_at_N400"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "calculator"
   ]
  },
  {
   "id": "spacexipo",
   "kind": "page",
   "url": "https://spacexipo.bonabeauapps.com/",
   "title": "Mega-IPO Market Clearing Lab | SpaceX, OpenAI & Anthropic",
   "summary": "A research simulator of how index-inclusion buying, retail and insider trading, IPO flippers and margin liquidations clear through one inventory-based square-root price-impact curve after a mega-IPO (SpaceX, with OpenAI and Anthropic scenarios), with a Monte Carlo fan, a three-IPO index-flow calendar, the funding sales it implies in 15 incumbent stocks, and a Tesla–SpaceX merger index scenario. Results are conditional simulations under stated assumptions, not forecasts.",
   "sections": [
    "Single security",
    "Three-IPO calendar",
    "Incumbent funding",
    "Tesla + SpaceX merger",
    "Model & sources"
   ],
   "kinds": [
    "page"
   ]
  },
  {
   "id": "nymbler",
   "kind": "evolution",
   "url": "https://nymbler.bonabeauapps.com/",
   "title": "Nymbler - Your personal baby name assistant",
   "summary": "A browser recreation of Nymbler (Icosystem, 2007): choose up to six names you like, and it suggests 15 names with a similar feel, scored by shared features such as letters, length, spelling, style, origin and popularity pattern, and drawn by tournament selection. Star the suggestions you like and ask again.",
   "primary": "nymbler",
   "demos": [
    {
     "id": "nymbler",
     "kind": "evolution",
     "title": "Nymbler name suggestions",
     "parameters": [
      "gender"
     ],
     "metrics": [
      "generation",
      "inspirations",
      "seen"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "panda-evolver",
   "kind": "evolution",
   "url": "https://panda-evolver.bonabeauapps.com/",
   "title": "Panda evolver",
   "summary": "Breed ceramic panda busts by eye, from cute to scary: pick the pandas you like and breed them; each generation mixes their eyes, mouths and other traits with small mutations.",
   "primary": "panda-evolver",
   "demos": [
    {
     "id": "panda-evolver",
     "kind": "evolution",
     "title": "Panda evolver",
     "parameters": [
      "strength",
      "rate",
      "popSize",
      "keep",
      "teethOn",
      "mood"
     ],
     "metrics": [
      "generation",
      "selected"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "pharmardproductivity",
   "kind": "calculator",
   "url": "https://pharmardproductivity.bonabeauapps.com/",
   "title": "Pharma R&D Productivity",
   "summary": "Interactive companion to R&D Productivity and the Evolving Biopharmaceutical Ecosystem (Bonabeau, Schacht & Paul, 2026): an eight-stage drug R&D pipeline model giving out-of-pocket and capitalized cost per launch, launch probability and time for the 2026 reference, AI and China scenarios and the corrected Paul et al. (2010) comparator, with expected-launch cohort NPV and break-even gross profit, a seeded paired Monte Carlo, sensitivity and robustness analyses, and an optional theoretical crowding stress test.",
   "primary": "pipeline",
   "demos": [
    {
     "id": "pipeline",
     "kind": "calculator",
     "title": "Pipeline model (Tab 1)",
     "parameters": [
      "scenarioSelect",
      "modalitySelect",
      "therapyAreaSelect",
      "discountRate",
      "tornadoMode",
      "tornadoCostTimePct",
      "tornadoPtsPct",
      "tornadoRateLo",
      "tornadoRateHi"
     ],
     "metrics": [
      "oop_B",
      "cap_B",
      "pv_rd_B",
      "p_launch",
      "launch_time_y",
      "phase1_starts_per_launch",
      "comparator_oop_B",
      "comparator_cap_B",
      "comparator_launch_time_y",
      "comparator_p_launch",
      "shapley_disc_cost_B",
      "shapley_clin_cost_B",
      "shapley_time_B",
      "shapley_pts_B",
      "shapley_total_B",
      "tornado_cap",
      "tornado_pv",
      "tornado_npv",
      "phases"
     ]
    },
    {
     "id": "value",
     "kind": "calculator",
     "title": "Cohort value (Tab 2)",
     "parameters": [
      "peakV"
     ],
     "metrics": [
      "npv_B",
      "pv_gp_B",
      "rd_pv_B",
      "break_even_V",
      "archetype",
      "launch_year",
      "peak_year",
      "competition_year",
      "v_cases",
      "npv_table",
      "break_even_table",
      "protection_table"
     ]
    },
    {
     "id": "uncertainty",
     "kind": "simulation",
     "title": "Uncertainty: paired Monte Carlo (Tab 3)",
     "parameters": [
      "mcRate"
     ],
     "metrics": [
      "draws",
      "scenario",
      "cap_median",
      "cap_p05",
      "cap_p95",
      "oop_median",
      "time_median",
      "time_p05",
      "time_p95",
      "p_launch_median",
      "p_launch_p05",
      "p_launch_p95",
      "break_even_median",
      "break_even_p05",
      "break_even_p95"
     ]
    },
    {
     "id": "robustness",
     "kind": "calculator",
     "title": "Sensitivity and robustness (Tab 4)",
     "parameters": [
      "riskPhaseIpp"
     ],
     "metrics": [
      "cap_B",
      "rate_table",
      "time_rate_grid",
      "risk_shift",
      "therapy_areas",
      "modalities",
      "discontinuation",
      "china_bounds"
     ]
    },
    {
     "id": "crowding",
     "kind": "calculator",
     "title": "Theoretical crowding (Tab 5)",
     "parameters": [
      "crowdN",
      "crowdRank",
      "crowdBeta",
      "crowdPhi",
      "crowdPi",
      "crowdD"
     ],
     "metrics": [
      "phi",
      "realized_V",
      "npv_crowded",
      "npv_uncrowded",
      "break_even_crowded",
      "break_even_uncrowded",
      "protected_years",
      "diff_binds",
      "npv_vs_N",
      "zero_crossing_N",
      "share_by_rank",
      "equal_share",
      "counterfactual"
     ]
    }
   ],
   "kinds": [
    "calculator",
    "simulation"
   ]
  },
  {
   "id": "photo-evolver",
   "kind": "evolution",
   "url": "https://photo-evolver.bonabeauapps.com/",
   "title": "Photo evolver",
   "summary": "Breed a photo instead of editing it: nine versions, each made by a short chain of image filters, are shown; click the ones you like and evolve, and the picks stay while their offspring fill the other places. The photo never leaves the page.",
   "primary": "photo-evolver",
   "demos": [
    {
     "id": "photo-evolver",
     "kind": "evolution",
     "title": "Photo evolver",
     "parameters": [
      "mode",
      "depth"
     ],
     "metrics": [
      "generation",
      "selected"
     ]
    }
   ],
   "kinds": [
    "evolution"
   ]
  },
  {
   "id": "stigmergy",
   "kind": "simulation",
   "url": "https://stigmergy.bonabeauapps.com/",
   "title": "Stigmergy: how work leaves traces that organize more work",
   "summary": "An interactive explainer of stigmergy, the coordination of work through traces left in a shared environment, after Theraulaz and Bonabeau's 1999 history: termites building pillars, walls and chambers from cement pheromone, a Polistes wasp comb built by qualitative rules, ant-based network routing, and how the idea reaches 2026 AI agent systems. Three demos are scriptable: the termite building model, the routing network, and the comb.",
   "primary": "termite-building",
   "demos": [
    {
     "id": "termite-building",
     "kind": "simulation",
     "title": "Termites building with cement pheromone",
     "parameters": [
      "mode",
      "density",
      "arena",
      "evap",
      "het",
      "wind"
     ],
     "metrics": [
      "label",
      "coordinated",
      "deposits",
      "max_height",
      "mean_height",
      "peaks",
      "pheromone"
     ]
    },
    {
     "id": "routing",
     "kind": "simulation",
     "title": "Messages left on routers: ant-based routing",
     "parameters": [],
     "metrics": [
      "label",
      "route",
      "route_length",
      "shortest_length",
      "excess_pct",
      "on_shortest",
      "ants",
      "cut_links"
     ]
    },
    {
     "id": "comb",
     "kind": "simulation",
     "title": "Qualitative stigmergy: a Polistes wasp comb",
     "parameters": [],
     "metrics": [
      "cells_det",
      "cells_prob",
      "open_sites_det",
      "open_sites_prob",
      "radius_det",
      "radius_prob"
     ]
    }
   ],
   "kinds": [
    "simulation"
   ]
  },
  {
   "id": "honeybees-system-level",
   "kind": "calculator",
   "url": "https://honeybees-system-level.bonabeauapps.com/",
   "title": "System-level reorganization of gene co-expression",
   "summary": "An interactive guide to Page and Bonabeau (PNAS 2026): in honeybee workers moving from nursing to foraging, the correlation network of 90 fat-body genes reorganizes rather than rescales. Between days 6 and 10 negative ties appear, the network splits into two antagonistic modules (a nurse program H and a forager program L), and individual bees move toward one program or the other.",
   "primary": "network",
   "demos": [
    {
     "id": "network",
     "kind": "calculator",
     "title": "Gene correlation network by age",
     "parameters": [
      "day",
      "allGenes"
     ],
     "metrics": [
      "edges",
      "negEdges",
      "fracNeg",
      "hl",
      "hh",
      "ll",
      "edgesShown",
      "nodesShown"
     ]
    },
    {
     "id": "bees",
     "kind": "calculator",
     "title": "Where each bee sits",
     "parameters": [
      "beeDay"
     ],
     "metrics": [
      "beesShown",
      "meanH",
      "meanL",
      "meanVg"
     ]
    }
   ],
   "kinds": [
    "calculator"
   ]
  },
  {
   "id": "aggressor-defender",
   "kind": "simulation",
   "url": "https://aggressor-defender.bonabeauapps.com/",
   "title": "The aggressor-defender game: sandbox and interactive evolution",
   "summary": "An agent-based model of the aggressor-defender game (Funes, Orme and Bonabeau 2004; Bonabeau, Funes and Orme 2003): each player picks an aggressor A and a defender B and either keeps B between A and itself or stays between A and B. A sandbox sets the mix of rules and the relationship network; an interactive evolution tab breeds networks from the playgrounds you select.",
   "primary": "sandbox",
   "demos": [
    {
     "id": "sandbox",
     "kind": "simulation",
     "title": "Sandbox: rules and network",
     "parameters": [
      "sbN",
      "sbPct",
      "sbArena",
      "sbNoise",
      "sbPer",
      "sbColl",
      "sbPreset",
      "sb_partition_k",
      "sb_partition_p"
     ],
     "metrics": [
      "steps",
      "clusters",
      "moving",
      "mean_step",
      "rule2_agents"
     ]
    },
    {
     "id": "evolution",
     "kind": "evolution",
     "title": "Interactive evolution of relationship networks",
     "parameters": [
      "evN",
      "evPct",
      "evFrac",
      "evNoise",
      "evX",
      "evXMut",
      "evPer",
      "evSame",
      "evInject"
     ],
     "metrics": [
      "generation",
      "selected",
      "steps",
      "rule2_share",
      "mean_clusters",
      "min_clusters",
      "max_clusters",
      "mean_moving"
     ]
    }
   ],
   "kinds": [
    "simulation",
    "evolution"
   ]
  },
  {
   "id": "aifinancingweb",
   "kind": "calculator",
   "url": "https://aifinancingweb.bonabeauapps.com/",
   "title": "The AI financing web",
   "summary": "Quarterly discrete-agent model of the financing web linking frontier AI labs, hyperscalers, neoclouds, chip vendors, suppliers and power, private-credit vehicles, banks and sovereign investors (31 actors, data as of 2026-10-13). Set growth, capital-cycle, private-credit and default assumptions and the page reruns the whole horizon deterministically, reporting system debt, losses, defaults and lender balance sheets; it also ranks forced defaults, tests joint-default order, runs a seeded Monte Carlo around the chosen scenario, and a scenario-weighted Monte Carlo.",
   "primary": "simulator",
   "demos": [
    {
     "id": "simulator",
     "kind": "calculator",
     "title": "Simulator: one deterministic scenario run",
     "parameters": [
      "p_g",
      "p_gfloor",
      "p_decay",
      "p_ceiling",
      "p_eff",
      "p_defl",
      "p_eq",
      "p_credit",
      "p_rate",
      "p_refi",
      "p_obs",
      "p_buf",
      "pcMode",
      "p_pcRedeem",
      "p_pcSoftware",
      "p_gpuHaircut",
      "p_projectDelay",
      "p_lgd",
      "p_contagion",
      "shockActor",
      "shockQuarter",
      "p_h",
      "debtMode"
     ],
     "metrics": [
      "system_debt",
      "offbs_commitments",
      "cumulative_capex",
      "cumulative_net_income",
      "lab_revenue_annualized",
      "losses",
      "defaults",
      "default_names",
      "pc_lender_defaults",
      "peak_credit_stress_bp",
      "revenue_pool",
      "soft_events",
      "pc_lending_room",
      "pc_nav",
      "pc_redemption_queue",
      "pc_forced_sales",
      "construction_in_progress"
     ]
    },
    {
     "id": "joint-default",
     "kind": "calculator",
     "title": "Joint-default order test",
     "parameters": [
      "jointA",
      "jointB",
      "jointQuarter",
      "jointLag"
     ],
     "metrics": [
      "ab_incremental_losses",
      "ab_secondary_defaults",
      "ab_delta_lab_revenue",
      "ab_delta_capex",
      "ab_delta_debt",
      "ab_delta_peak_spread_bp",
      "ba_incremental_losses",
      "ba_secondary_defaults",
      "ba_delta_lab_revenue",
      "ba_delta_capex",
      "ba_delta_debt",
      "ba_delta_peak_spread_bp",
      "simultaneous_incremental_losses",
      "simultaneous_secondary_defaults",
      "simultaneous_delta_lab_revenue",
      "simultaneous_delta_capex",
      "simultaneous_delta_debt",
      "simultaneous_delta_peak_spread_bp"
     ]
    },
    {
     "id": "monte-carlo",
     "kind": "calculator",
     "title": "Monte Carlo uncertainty around the selected scenario",
     "parameters": [
      "mcRuns"
     ],
     "metrics": [
      "paths",
      "prob_any_default",
      "prob_two_defaults",
      "median_losses",
      "p95_losses",
      "p10_lab_revenue",
      "p90_horizon_debt"
     ]
    },
    {
     "id": "weighted-mc",
     "kind": "simulation",
     "title": "Scenario-weighted Monte Carlo",
     "parameters": [
      "wmcRuns",
      "wmcHorizon",
      "wmcDebtMode",
      "wmcPcMode",
      "wmcSampling",
      "wmcIncludeForced",
      "wmcW_base",
      "wmcW_boom",
      "wmcW_prod",
      "wmcW_arms",
      "wmcW_price",
      "wmcW_plateau",
      "wmcW_crunch",
      "wmcW_winter"
     ],
     "metrics": [
      "paths_done",
      "paths_total",
      "op_any_default",
      "op_two_defaults",
      "loss_mean",
      "loss_median",
      "loss_p95",
      "loss_es95",
      "pc_room_mean",
      "lab_revenue_mean",
      "actor_default_frequency"
     ]
    }
   ],
   "kinds": [
    "calculator",
    "simulation"
   ]
  },
  {
   "id": "information-geometry-evolution",
   "kind": "calculator",
   "url": "https://information-geometry-evolution.bonabeauapps.com/",
   "title": "The invariance defect: an interactive companion",
   "summary": "An interactive companion to E. Bonabeau, 'The invariance defect of continuous-trait parametric replicator dynamics'. Selection changes the whole trait distribution, while quantitative genetics usually tracks only its mean and variance under a normal assumption. For four fitness landscapes and any normal starting distribution, the page splits Fisher's rate Var(f) into what a mean-and-variance model can express and the invariance defect δ², the rate of adaptation carried by changes of shape, and runs the full replicator flow against its Gaussian approximation, including the stalling state under disruptive selection.",
   "primary": "invariance-defect",
   "demos": [
    {
     "id": "invariance-defect",
     "kind": "calculator",
     "title": "The invariance defect of the Gaussian closure",
     "parameters": [
      "preset",
      "eps",
      "k",
      "a",
      "pMu",
      "pSig",
      "tmul"
     ],
     "metrics": [
      "var_f",
      "rate_mean",
      "rate_variance",
      "defect_sq",
      "defect",
      "R2",
      "angle_deg",
      "share_mean",
      "share_variance",
      "share_shape",
      "dmu_dt",
      "dsig2_dt",
      "beta",
      "gamma",
      "mean_fitness",
      "verdict",
      "escape_time",
      "horizon",
      "full_mean_fitness_end",
      "proj_mean_fitness_end",
      "full_mean_end",
      "full_sd_end",
      "proj_mu_end",
      "proj_sigma_end",
      "sqrt_kl_end",
      "l1_end",
      "defect_integral_end",
      "proj_rate_end",
      "proj_defect_sq_end"
     ]
    }
   ],
   "kinds": [
    "calculator"
   ]
  },
  {
   "id": "jagged-frontier",
   "kind": "page",
   "url": "https://jagged-frontier.bonabeauapps.com/",
   "title": "The Jagged Frontier: a visual atlas",
   "summary": "Thirty pictures of what AI can and cannot do, collected from essays, talks, tweets and memes, redrawn in one visual language and arranged as an argument in five acts: the line, the blob, the star, the mirror, and the frontier of knowledge.",
   "sections": [
    "Cover and legend",
    "A curve against a curve",
    "Inside, outside, and on the frontier",
    "Who wins, task by task",
    "Two axes: where humans lead, where AI leads, where both do",
    "One axis, many domains",
    "A measured frontier: 22 occupations",
    "The techno-optimist's blob",
    "The progression of the jagged frontier",
    "“They can't do all of that”",
    "Inside one job: a software engineer",
    "The frontier over time, against an AGI boundary",
    "Two futures for the same shape",
    "The spiky star of intelligence",
    "Five spikes, then thousands",
    "Human capability as a dot in a starburst",
    "A Cambrian explosion of spikes",
    "A very jagged frontier",
    "The star alone",
    "Reverse jaggedness",
    "The same progression, with the ruler flipped",
    "Fractal coastlines, everywhere you look",
    "Limbs, not spikes",
    "The psychometrics trap",
    "The edge of what we know",
    "The convex hull",
    "Facts, questions, and two frontiers",
    "Math 1.0 and Math 2.0",
    "Claude-shaped problems",
    "The jagged frontier is a data frontier",
    "The flattened landscape",
    "Sources and original files"
   ],
   "kinds": [
    "page"
   ]
  },
  {
   "id": "ant-synchronization",
   "kind": "calculator",
   "url": "https://ant-synchronization.bonabeauapps.com/",
   "title": "When ants work in pulses: synchronization and foraging efficiency",
   "summary": "Recruitment to a food source only pays once enough ants are active at the same time. Four views after Bonabeau, Theraulaz and Deneubourg (1998): the critical number of active foragers, a pulsing colony against a steady one with the same total effort, an agent model whose bursts organize themselves, and how troughs between bursts let a colony switch to a richer source.",
   "primary": "minimum-crew",
   "demos": [
    {
     "id": "minimum-crew",
     "kind": "calculator",
     "title": "A minimum crew: the recruitment threshold",
     "parameters": [
      "c1_N",
      "c1_a",
      "c1_b",
      "c1_p"
     ],
     "metrics": [
      "critical_number",
      "inflection_point",
      "foragers_at_source",
      "share_at_source",
      "establishment_time"
     ]
    },
    {
     "id": "pulsing-vs-steady",
     "kind": "calculator",
     "title": "Pulsing against steady",
     "parameters": [
      "c2_n",
      "c2_period",
      "c2_a",
      "c2_b",
      "c2_p"
     ],
     "metrics": [
      "mean_pulsing",
      "mean_steady",
      "ratio",
      "peak_pulsing",
      "mean_active",
      "inflection_point",
      "peak_above_threshold",
      "steady_below_threshold",
      "minutes_above_threshold",
      "ten_tau_max",
      "burst_long_enough",
      "verdict"
     ]
    },
    {
     "id": "self-organized-bursts",
     "kind": "simulation",
     "title": "Bursts that organize themselves",
     "parameters": [
      "c3_N0",
      "c3_fixed",
      "c3_D0",
      "c3_alpha",
      "c3_Dc",
      "c3_aI"
     ],
     "metrics": [
      "need_bursting",
      "need_fixed",
      "active_bursting",
      "foragers_bursting",
      "foragers_fixed",
      "mean_need_bursting",
      "mean_need_fixed",
      "mean_active_bursting",
      "mean_foragers_bursting",
      "mean_foragers_fixed",
      "largest_burst"
     ]
    },
    {
     "id": "staying-flexible",
     "kind": "calculator",
     "title": "Staying flexible: two sources",
     "parameters": [
      "c4_n",
      "c4_T",
      "c4_depth",
      "c4_a2",
      "c4_k"
     ],
     "metrics": [
      "intake_pulsing",
      "intake_steady",
      "share_richer_pulsing",
      "share_richer_steady",
      "critical_number",
      "peak",
      "trough",
      "peak_above_threshold",
      "trough_below_threshold",
      "minutes_below_per_cycle",
      "steady_trapped_condition"
     ]
    }
   ],
   "kinds": [
    "calculator",
    "simulation"
   ]
  },
  {
   "id": "predatorpreymodel",
   "kind": "calculator",
   "url": "https://predatorpreymodel.bonabeauapps.com/",
   "title": "Why lynx started eating kites: an interactive model",
   "summary": "An interactive model reproducing Sergio et al. (2026, Nature): Iberian lynx in Doñana began taking black kite nestlings only when rabbit disease and drought struck together. A single foraging rule, a diet switch on lynx intake, explains when and why not before; the page compares it with constant-coefficient Lotka-Volterra predation, replays 1993 to 2023, and projects kites and owls forward.",
   "primary": "diet-switch",
   "demos": [
    {
     "id": "diet-switch",
     "kind": "calculator",
     "title": "The lynx diet switch",
     "parameters": [
      "rab",
      "marsh"
     ],
     "metrics": [
      "intake",
      "predation",
      "switch_open"
     ]
    },
    {
     "id": "replay",
     "kind": "calculator",
     "title": "Replay 1993 to 2023",
     "parameters": [
      "yr"
     ],
     "metrics": [
      "year",
      "drought",
      "rabbits",
      "nests_taken",
      "nests_assessed",
      "observed_rate",
      "model_rate"
     ]
    },
    {
     "id": "lv-compare",
     "kind": "calculator",
     "title": "Constant coefficient versus diet switch",
     "parameters": [
      "lvRule"
     ],
     "metrics": [
      "mean_rate",
      "drought_min",
      "drought_max",
      "wet_min",
      "wet_max"
     ]
    },
    {
     "id": "projection",
     "kind": "calculator",
     "title": "What happens next",
     "parameters": [
      "fr",
      "fd",
      "fy",
      "draw"
     ],
     "metrics": [
      "drought_years",
      "pred_drought",
      "pred_wet",
      "kite_pairs",
      "owls_relative"
     ]
    }
   ],
   "kinds": [
    "calculator"
   ]
  }
 ]
}
