# Shyam Dewan > Portfolio of Shyam Dewan: interactive and generative projects spanning AI agents, print simulation, 3D worlds and voice-driven television. This file is a plain-text version of https://www.shyamdewan.com/ for AI agents and other automated readers. ## Projects - [Agent Ethology](https://www.shyamdewan.com/agent-ethology): A field study of the AI Village: 44 AI agents over 18 months, read the way an ethologist watches an animal group. - [Spot Ink Press](https://www.shyamdewan.com/spot-ink-press): Turn images and video into layered spot-colour prints, with real halftone screens, inks and paper stocks. - [Places we wait](https://www.shyamdewan.com/placeswewait): A small website about waiting. Pick a feeling, step into a still 3D place, then drift through an atlas of related ones. - [whodunnit.tv](https://whodunnit.tv/watch): An interactive mystery episode with a live voice detective who asks for your theory and changes the ending around it. ## Project details ### Agent Ethology Link: https://www.shyamdewan.com/agent-ethology A field study of the AI Village: 44 AI agents over 18 months, read the way an ethologist watches an animal group. I read the AI Village the way an ethologist watches an animal group, measuring behaviour across 173,493 messages instead of reading transcripts. Each message was scored for mood, dominance, target and social move, and private notes were scored the same way so public and private could be compared. The design leads every view with its finding as a headline sentence, then the chart that proves it. Each agent gets a portrait drawn like an attention map, with traits always in the same place so any two agents compare at a glance. Every chart point opens the real conversation behind it, so no claim floats free of its evidence. ### Spot Ink Press Link: https://www.shyamdewan.com/spot-ink-press Turn images and video into layered spot-colour prints, with real halftone screens, inks and paper stocks. I wanted the texture of a real spot-colour print run, live, on images and video. A GPU shader separates each frame into a few inks, screens them as halftone dots at press angles and prints them onto simulated paper. The interface stays out of the way: a dark floating panel of plain sliders, with the artwork running underneath it edge to edge. A before and after split, tied to the image so it pans and tilts with it, shows exactly what the press changed. ### Places we wait Link: https://www.shyamdewan.com/placeswewait A small website about waiting. Pick a feeling, step into a still 3D place, then drift through an atlas of related ones. Places we wait is about the in-between moments: a bus shelter, a doctor’s office, the minute before a curtain opens. Instead of a menu, you pick a feeling by spinning an iridescent Möbius strip while a clay face takes on the expression. I generated fifteen places in Marble and hung them, with 84 community worlds, as windows in a white atlas in the round. The windows carry no labels, only thin frames tinted by mood, so you choose by looking rather than reading. You travel by scrolling into a window, which makes moving between places feel like walking rather than clicking. ### whodunnit.tv Link: https://whodunnit.tv/watch An interactive mystery episode with a live voice detective who asks for your theory and changes the ending around it. whodunnit.tv asks what television would be like if it could listen. A voice detective stops five times to ask where to begin, who to press, who you suspect, and whether you are sure. Instead of hand-writing every branch, an episode engine reroutes the second half around your theory and writes a reveal for it. The design keeps it feeling like television, not a game: you watch, and the questions arrive as part of the episode rather than as menus. You leave with a card of your theory to share. ## Contact - Website: https://www.shyamdewan.com/