Behavioral and computational scientist. I study how information is distributed, interpreted, and transformed through networks, and how communication structure shapes judgment, coordination, and collective behavior across human, organizational, and AI systems.
The work combines experiments, network analysis, measurement, simulation, statistical inference, and computational systems. Four lines are active: machine-agent networks, information transformation and provenance, simulation as a way of separating mechanisms that produce similar visible outcomes, and measurement of cultural structure.
| Repository | Paper |
|---|---|
| m2m-collective-belief-dynamics | Communication content and local connectivity shape collective accuracy in networks of machine agents |
| m2m-bavelas-groups | Communication topology shapes collective problem solving in networks of machine agents |
| shared-ancestry-lm-reports | Shared ancestry among language-model reports: code, prompts, constructed worlds, and every stored model call |
| Information mutation code and recoding release (OSF) | Information mutation across implemented AI-agent communication architectures |
| bfg-corrective-replicator | A corrective constraint increases multi-target attainment under finite adaptive search |
| compositional-latent-diffusion | Diffusion before adoption: A model of compositional latent diffusion |
| meme-family-contrastive-signatures | Measuring meme-family structure with contrastive sparse features |
| memes-structural-signals | Memes as structural signals: A chemical model of interpretive compatibility |
Earlier snapshots: m2m-information-mutation (early development version of the information-mutation simulation) and structural-legibility-prototype (superseded prototype of the shared-ancestry study).
manuscript-u recommends where to submit an academic manuscript, returning ranked journals with fit scores, desk-rejection risk, and revision culture notes grounded in data fetched at query time.
org-signal is browser-based network analysis for teaching and research. It builds networks from message exports, surveys, drawings, and generated organizations with planted structure, and its calculations are checked against NetworkX and closed-form references (accuracy report). Live at orgsignal.graystoneindustries.co.
sandia-automated-presentation is a fifteen-minute talk on credibility and evaluation in systems of interacting agents, built as a self-running browser presentation. Other talks and recorded walkthroughs, each with its own repository, are listed at graystoneindustries.co/talks.
Browser-native recommendation tools, each a two-stage reasoning pipeline over live web search with explicit confidence scoring. No backend, no build step.
- draft-and-beam, used sailboats
- prime-rigs, camera and lighting packages
- garden-mind, plant selection by zone, soil, and sun
- what-should-i-ride, bicycle tires
- employee-training-finder, training programs
daily-digest-script assembles a daily email and Slack digest.