Welcome to a special edition of GenTalks, streamed live from The Lighthouse in Venice, CA. This session experiments with Discord, a new multi-camera setup, and a format that blends community check-ins with high-level guest discussions on the future of creative AI.

We’re joined by two innovators approaching generative media from radically different directions:

1. Parth Patil — Agentic Workflows & Vibe Coding

Parth walks through his journey from Clubhouse to building coding agents full-time. He discusses:

Using Claude Code to build fully automated creative pipelines

Local agent workflows that can handle editing, clipping, captioning, and video assembly

Why coding agents matter more for non-technical creators

The origin story behind the “Reid Hoffman talks to Reid Hoffman” avatar experiment

His long-term focus on agent-driven games, world models, and emergent storytelling

2. Jiajian Min — AI/XR Performance, Motion Capture & World Models

Jiajian presents extraordinary work combining dance, calligraphy, projection mapping, and generative models. Highlights include:

Motion-capture-driven AI calligraphy created in collaboration with dancers

Hand-tracked real-time AI animation using Leap Motion + custom diffusion models

Award-winning interactive installations presented at ACM Multimedia

Harvard research digitizing Buddhist caves into immersive spatial environments

Panoramic world-model theatre created for MIT: nested dreams, 360° projection, and live camera compositing

This session explores how creative AI is splitting into two powerful directions:

(1) Agentic automation for scalable personal workflows

(2) Embodied, immersive, spatial art beyond the 2D frame

Key Topics Covered

Verticalization of creative toolchains (Figma, Weavy, Adobe, etc.)

Agent-based coding and why local agents matter

The difference between Claude Code vs LM Studio vs typical web apps

Motion capture + diffusion models in live performance

World models and panoramic generative imagery

Future of AI-native theatre and spatial storytelling

How AI augments the human body in ritual and performance

Why games and interactive worlds may be the next frontier for creators

Tools Mentioned

Claude Code (Anthropic)

Leap Motion / Ultraleap

TouchDesigner

OptiTrack

Kinect

WorldLab

Early avatar systems (Hour One-style pipelines)

Opus Clips (as reference)

Descript

Weavy

00:00 — Discord setup chaos

03:20 — Community check-in

09:10 — Introduction of guest: Parth Patil

12:00 — Parth’s GPT-4 origin story

20:30 — Building “Reid Talks to Reid”

28:00 — Why Claude Code is a creative game-changer

35:00 — Automated podcast/pipeline example

44:10 — Q&A on agents, apps, and iPhone development

48:30 — Introduction of guest: Jiajian Min

50:00 — Embodi Ink (dance + calligraphy)

58:15 — Hand-tracked diffusion animation

1:05:20 — Harvard Buddhist cave immersive

1:10:40 — MIT panoramic theatre

1:17:20 — Artistic philosophy of embodied AI

1:21:00 — Final thoughts & upcoming sessions

1:24:00 — Closing + next-day RoboDojo announcement