BONABEAUAPPS

Eric Bonabeau · Apps & explorations

Simple apps for complex systems.

Interactive tools and data-driven explorations built by a complexity scientist — turning questions about markets, medicine, and longevity into things you can actually play with.

Browse the apps

Eric Bonabeau has spent three decades studying how simple rules produce intelligent collective behavior — in ant colonies, in markets, and in machines.

Eric is a complex systems scientist, an entrepreneur, and an educator.

150
scientific publications
50,000+
citations
~300
analytics & ML projects for Fortune Global 1000 clients
14
granted patents
The full story

Trained as a theoretical physicist (PhD, Université Paris-Sud; alumnus of École Polytechnique and Télécom ParisTech), he has published 150 scientific articles — in Science, Nature, and PNAS, as well as Harvard Business Review and MIT Sloan Management Review — collecting more than 50,000 citations along the way. His book Swarm Intelligence (Oxford University Press) has been a scientific bestseller for 25 years and inspired Michael Crichton's novel Prey; he also co-authored Self-Organization in Biological Systems (Princeton University Press), a leading graduate biology textbook, and was the founding editor-in-chief of two scientific journals.

As Chairman and Chief Scientific Officer of Icosystem for 17 years, he grew a single idea into a firm that executed nearly 300 analytics and machine learning projects for Fortune Global 1000 companies — while spinning off four companies along the way. He went on to serve as Chief Scientific Officer of Telepathy Labs, a customer service and analytics AI company, and as founding CTO of BiomEdit, a bacterial discovery and engineering company.

As the founding Dean of Computational Sciences at Minerva University — the #1 most innovative university in the world — he designed the curriculum, carried it through accreditation, and taught its first cohorts of students.

Today, Eric is Chief Scientific Officer at Discontinuity Capital, an AI-focused hedge fund, and Professor of Practice at Arizona State University (ASU–Santa Fe Institute Center for Biosocial Complexity). This site is his workshop: small, focused apps that take one hard question at a time and make it explorable.

He also once designed a prototype food replicator for NASA's long-term space missions, based on the science of molecular gastronomy.

Trajectory
  • Now
    Chief Scientific Officer, Discontinuity Capital
    An AI-focused hedge fund.
  • 2014 – Present
    Professor of Practice, Arizona State University
    ASU–Santa Fe Institute Center for Biosocial Complexity.
  • 2022 – 2025
    Founding CTO, BiomEdit
    Bacterial discovery and engineering for animal health; raised $55M in equity and ~$10M in grants.
  • 2018 – 2022
    Chief Scientific Officer, Telepathy Labs
    Directed all machine learning and AI across the US, Europe, and Asia for conversational, customer-service and analytics AI.
  • 2013 – 2017
    Founding Dean of Computational Sciences, Minerva University
    Designed the curriculum, led it through accreditation, and taught the first cohorts; later Dean of Global Affairs.
  • 2000 – 2017
    Founder, Chairman & Chief Scientific Officer, Icosystem
    Nearly 300 analytics and ML projects for Fortune Global 1000 companies, governments, and foundations; four spin-off companies.
  • 1996 – 1999
    Interval Research Fellow, Santa Fe Institute
    Research on collective behavior in biological and man-made systems.

Each app is a self-contained exploration — a question, the data behind it, and an interface that lets you reach your own conclusions. Pick a theme to narrow the list; most apps belong to more than one.

38 apps

Longevity What the science of aging actually says — separating evidence from hype in the quest for a longer healthspan. BiotechExplainers Pharma R&D Productivity Why discovering a new drug keeps getting more expensive — a data-driven look at the productivity puzzle in pharmaceutical R&D. BiotechFinancial models SpaceX IPO What happens when a giant goes public — how a mega-IPO would ripple through markets, reshape the indices, and reshuffle investor portfolios. Financial modelsAgent-based models Information Theory Information is what you can rule out — eight small games that take you from a jar of marbles to Shannon's noisy channel coding theorem, one bit of intuition at a time. Explainers Estimating Agent-Based Models A simulation is easy to run and hard to fit — ask the real data and the simulated data the same randomly chosen questions, then tune the parameters until the answers agree. Interactive figures built around a small epidemic model. ExplainersAgent-based models The AI Financing Web Who owes what to whom in the AI build-out — a quarterly simulator of the capital network linking frontier labs, hyperscalers, neoclouds, chip and power suppliers, and their lenders, with named scenarios and a scenario-weighted Monte Carlo to see where the defaults land. Financial modelsAgent-based models Why Lynx Started Eating Kites Drought and rabbit disease, together and only together, turned Doñana's Iberian lynx into predators of black kite nestlings. An interactive predator-prey model built from one foraging rule reproduces 30 years of field data, with the model paper and the original Nature article alongside. ExplainersBiological models Agent-Based Pattern Formation From ants clustering their dead to agents that carry out a Turing-style activator–inhibitor rule: run the models of Bonabeau's 1997 Artificial Life paper, load its figures as presets, and watch spots, stripes and labyrinths emerge from local rules. The paper is one tab away. Biological modelsSwarm & self-organizationAgent-based models How Nurse Bees Become Foragers When a worker honeybee leaves brood care to forage, its genes don't just switch on and off: the whole correlation network reorganizes. An interactive guide to Page & Bonabeau's 2026 PNAS paper, with every network recomputed from the data and the paper alongside. Biological models The Aggressor-Defender Game Each player picks an aggressor and a defender, then either hides behind the defender or puts itself between the two. Hiding scatters the group; shielding packs it into a single cluster; mixes and relationship networks give everything in between. Play in the sandbox, breed new groups by picking the patterns you like, and read the three papers behind the game. ExplainersAgent-based modelsEvolutionary computing Swarm Building How do wasps build a nest without a blueprint? Agents wander a 3D lattice and drop a brick only when their local neighborhood matches a rule, and the structure itself coordinates the work. Rebuild the published nests, evolve new architectures, and read the Science, J. Theor. Biol. and BioSystems papers. Biological modelsSwarm & self-organizationAgent-based modelsEvolutionary computing evoDesign: Evolving Tiles Design by selection rather than specification: pick the tiles you like, mutate or cross them, and watch seamless patterns evolve to your taste. A working reproduction of an interactive evolutionary design tool, with the book chapter that describes it. Evolutionary computing Army Ant Raids Thousands of blind workers, one pheromone rule, and a raid that fans out or narrows depending on the prey. Run the trail model live for three Eciton species, then let a search algorithm find the parameters that bring the most food home per ant. Biological modelsSwarm & self-organizationAgent-based modelsEvolutionary computing Ant Colony Optimization Artificial ants build tours of the traveling salesman problem, lay pheromone in proportion to tour quality, and let trails evaporate. Drop in cities, tune the Ant System parameters, and watch good routes emerge from stigmergy. ExplainersSwarm & self-organization Evolving Drug Candidates A genetic algorithm searches compounds built from chemist-chosen fragments, computational models score them, and the chemist steers the search by reshaping the models and rating what comes back. A working model of interactive evolution in molecular space. BiotechEvolutionary computingApplications Stigmergy Work leaves traces; traces direct more work. How termites build a cathedral while each one behaves as if alone: the history of the idea since Grassé, live models of pillar building and trail laying, and where stigmergy shows up in engineering today. Biological modelsSwarm & self-organizationAgent-based modelsAI Marginally Stable Swarms A colony that forages well has committed to a trail; one that can change its mind has not committed too hard. Simulate the foraging model and see why the sweet spot sits just past a bifurcation, ordered enough to build a trail and close enough to abandon it. Biological modelsSwarm & self-organization The Jagged Frontier 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. Every figure is interactive. ExplainersAI Interactive Inversion Which parameters produced this picture? Be the observer who picks the candidate that looks right, generation after generation, to invert models of a market frenzy, mantle convection, fault spacing and an aquifer, then hand the job to a model. Evolutionary computingAI Letting the Workers Divide the Work The planner knows what is efficient; only the people doing the work know what they would rather do. The computer tunes six plans, each person ranks only their own piece, and those rankings choose which plans breed next. Mail carriers and a nurse roster, after a CEC 2004 paper. Evolutionary computingApplications Nymbler You can’t describe the baby name you want, but you would recognize it. Star a few names you like and Nymbler finds others with the same feel, weighing letters, spelling, style, origin and popularity history. Evolutionary computing Photo Evolver Breed your photo instead of editing it. Nine versions, each a short chain of filters: click the ones you like and evolve, and the rest are replaced by their offspring. Your photo never leaves the page. Evolutionary computing Biomorphs Dawkins’s branching figures, bred by eye. Click the offspring you prefer and it becomes the next parent; add symmetry, segmentation and gradients, or breed from several parents at once. ExplainersEvolutionary computing Ant Synchronization Why ant colonies work in pulses. Recruitment only takes hold above a critical number of active foragers, so a small colony that spends the same effort in synchronized bursts can forage where a steady one cannot. With the 1998 BioSystems paper. Biological modelsSwarm & self-organization Margin-Liquidation Cascades How a price drop can sell itself: a step-by-step walk through forced selling by leveraged traders, from one position to finite-size effects, heavy tails and fading price impact. With the Physical Review E paper. Financial modelsAgent-based models Information Geometry of Evolution The invariance defect: how much of selection a mean-and-variance model of a trait misses, and why the missing part is a rate of adaptation. An interactive companion to the replicator-dynamics manuscript. Biological modelsEvolutionary computing Interactive Evolution of Images Karl Sims’s genetic images, rebuilt in the browser. Each picture is a small Lisp expression evaluated at every pixel: pick one to mutate or two to mate, and breed images nobody would think to write by hand. ExplainersEvolutionary computing Facial Composites Build a suspect’s face from memory the way EvoFIT does: study a target, hide it, then breed a likeness by choosing among whole faces rather than picking noses and eyes one at a time. ExplainersEvolutionary computingApplications Panda Evolver Ceramic panda busts, from cute to scary. Pick the ones you like and breed them: each generation mixes eyes, mouths and other traits, with small mutations. Evolutionary computing Interactive Evolution of Reaction-Diffusion Twelve Gray-Scott simulations running side by side, each with its own feed and kill rates, scale, flow and colors. Pick the patterns you like and breed the next generation from them, after Karl Sims. ExplainersSwarm & self-organizationEvolutionary computing Evolving Particle Swarms Map everything a swarm can do. A quality-diversity search fills a grid of collective behaviors (flocking, milling, clustering) with the most stable rule it finds for each, live in your browser. With the Physica A paper. Biological modelsSwarm & self-organizationAgent-based modelsEvolutionary computing Choice Decision Choice tasks built so that one pick reveals the decision strategy: a genetic algorithm designs each respondent’s product comparison so every option is the pick of a different decision rule. With the 2014 JBDM paper. Evolutionary computingApplications Ant Scheduling Ants versus a market in a truck paint shop. Booths either bid for each truck like traders or specialize the way ant workers do, with response thresholds that learn. Run both rules side by side and see which wastes less paint. With the Adaptive Behavior and Proc. R. Soc. B papers. Biological modelsSwarm & self-organizationApplications Breeding Network Metamaterials A mechanical metamaterial grown from a short genome can be bred. Pick the specimens you would keep from a sheet of scale-rich networks, each squeezed in your browser, and breed toward auxetic, isotropic or payload-shielding designs, in 2D and 3D. With the Nature paper and a manuscript documenting the models. ExplainersEvolutionary computingAIApplications Interactive Data Exploration When you do not know what you are looking for, you cannot write a fitness function. Evolve distance metrics, projections and views of a dataset by eye, and see in nine live studies where a modern automatic criterion can replace the person and where it cannot. With the accompanying article. Evolutionary computingAI Emoji Evolution Pick the faces you like and they become the parents of the next generation: hair, expression, accessories and setting mix and mutate. Describe the face you want and the breeding steers toward it. Evolutionary computingAI Evolving Car Silhouettes Evolve a car silhouette by eye, inside rules you draw. Two parent designs are recombined into six offspring, and constraint zones on the drawing board, such as a roof limit or the driver’s sight line, decide which ones are allowed to exist. Evolutionary computingApplications Term Sheet Simulator Change a clause and see who gets paid what. SAFEs, convertible notes and priced rounds from pre-seed to Series A, played out across the futures a company might face. Financial modelsApplications

A question about one of the apps, an idea worth exploring, or an invitation to speak — send a note and it goes straight to Eric.