Seance.AI — Summoning Ideas Into Reality

What if bringing an idea to life were as simple as summoning an entire team?

Meet Seance.AI, a Halloween-inspired, AI-powered project orchestration platform that bridges two worlds: human imagination and autonomous AI execution.

While conventional AI assistants generate answers, Seance.AI is designed to turn ambitious goals into actionable workflows powered by specialized AI agents, real-world tools, and an interactive visual execution graph.

Whether you’re a student trying to master a difficult subject, a community organizer launching an initiative, or an entrepreneur developing a business, Seance.AI makes agentic AI more accessible, understandable, and useful.

Three Worlds. Endless Possibilities.

Community — Bring People Together

Discover local organizations, explore volunteering opportunities, research community needs, and develop actionable plans for real-world initiatives using location-aware tools and AI-powered research.

Education — Bring Knowledge to Life

Transform learning goals into structured study plans, interactive knowledge graphs, recommended educational resources, flashcards, and quizzes.

Business — Bring Ideas to Market

Explore business concepts through AI-powered competitor research, launch planning, marketing strategies, and structured deliverables.

Each workspace shares the same underlying orchestration architecture while providing domain-specific tools and outputs.

The Technology Behind the Séance

Google DeepMind — Gemma 4: The Intelligence

At the heart of Seance.AI is Gemma 4 26B A4B, accessed through Google’s Gemini API.

Gemma powers our goal interpretation and structured planning workflow, translating natural-language requests into task graphs with dependencies, specialized agent assignments, and tool selections.

Rather than returning only conversational text, the application is designed to transform model outputs into structured work that can be validated, executed, and inspected.

Google DeepMind enables us to explore powerful open-weight intelligence as the foundation for an accessible agentic product.

Maritime (YC F26) — The Execution Infrastructure

Maritime serves as the agentic infrastructure layer of our architecture.

Our Maritime integration is designed to support isolated, persistent agent workloads, separating AI execution from the user interface and creating a foundation for resumable workflows.

As projects evolve, this architecture is intended to support more reliable autonomous execution, workload isolation, and scalable orchestration.

Maritime lets us focus on building the human experience around autonomous AI instead of reinventing the infrastructure required to run it.

DigitalOcean & Open-Source AI — The Foundation for Accessibility

We believe autonomous AI should be accessible to more people than just those who know how to configure agents, build infrastructure, or write code.

Seance.AI combines an open-weight language model with a modular application architecture designed to make AI-powered workflows extensible and understandable.

By sharing our implementation publicly, we hope to enable other developers to adapt the platform for new educational, community, and entrepreneurial use cases.

DigitalOcean’s commitment to open-source innovation aligns with our goal of making advanced AI systems easier to build with, inspect, and extend.

Our Engineering Stack

Frontend & 3D Experience

• React, TypeScript, Vite, HTML5, CSS3
• Three.js, React Three Fiber, Drei, GSAP
• GLB/GLTF animated 3D models
• Custom cinematic camera transitions and responsive UI

Workflow Visualization & Domain Tools

• React Flow (@xyflow/react) for interactive task dependency graphs
• React Router for workspace navigation
• Leaflet and React Leaflet for geographic exploration
• React Markdown for rendering agent-generated deliverables
• Domain-specific research, educational resources, and planning components

AI & Backend

• Google DeepMind Gemma 4 26B A4B
• Gemini API for model inference
• Node.js and JavaScript backend services
• Structured task schemas, agent orchestration, and tool adapters
• Maritime integration for persistent agent workloads

Hardware & Developer Infrastructure

• Arduino Uno and Python serial communication bridge for physical execution-status feedback (hardware integration, if included in the submitted version)
• Git, GitHub, npm, and Vercel
• Automated tests and TypeScript production builds

What Makes Seance.AI Different?

1. Agentic AI for Everyone

Users interact with goals rather than complex agent configurations. Seance.AI presents execution through an understandable visual interface.

2. The Living Graph

Instead of hiding automation behind a conversation, Seance.AI visualizes task dependencies, agent responsibilities, execution statuses, and generated results.

3. One Engine, Three Specialized Experiences

Community, Education, and Business share reusable orchestration components while delivering different domain-specific experiences.

4. An Immersive Frontend That Explains the Product

Our interactive haunted forest, house, Ouija board, and horror-themed transitions turn the concept of “summoning” an AI workforce into a memorable user experience.

5. Built Toward Human–AI Collaboration

Our longer-term vision is a collaborative workspace where people and AI agents share plans, revise tasks, approve actions, and resume work together.

Challenges We Tackled

• Combining real-time 3D rendering with a responsive React application
• Building an interactive dependency graph for AI-generated workflows
• Designing structured execution pipelines rather than relying solely on free-form chatbot output
• Integrating model inference and external tools across three different application domains
• Exploring persistent agent infrastructure for autonomous workloads
• Balancing ambitious AI capabilities with usability and performance

Impact & What’s Next

We’re building toward a world where someone with an idea doesn’t need an engineering team, a complex software stack, or extensive AI knowledge to begin making progress.

Our roadmap includes real-time collaborative planning, human approval checkpoints, resumable project execution, more rigorous verification of agent-generated results, and additional community-contributed workflow templates.

Google DeepMind gives Seance.AI its intelligence. Maritime provides the infrastructure for autonomous execution. Open-source AI makes the approach extensible. And Seance.AI connects those capabilities to the people and projects that need them.

Seance.AI: Stop leaving your ideas haunting your to-do list. Summon the tools to bring them to life.