Geist

Company: Geist

Self-reported basis: The entire analysis is based on a single project description provided by the caller — a self-reported write-up from the author (Kevin Fischer) of a hackathon submission to the OpenAI 2026 hackathon. No external verification or historical data are available.

What it appears to be: A screen-aware, AI-powered desktop companion that exists as an animated ghost-like entity within the user’s desktop environment. It observes what happens on screen, thinks in real time, and communicates with the user through speech and visual cues.

What changed: The project is a prototype submitted for a hackathon. There is no evidence of prior development or commercial traction beyond this one submission.

Single most important open question: Is there any evidence that Geist has moved beyond the prototype stage, or whether it has been tested with users in real-world conditions?

The description states that Geist is:

• A fullscreen transparent Electron app
• An embodied screen pet that "floats over the desktop"
• Equipped with vision and speech capabilities, including:
  • Screen awareness
  • Camera context
  • Realtime voice interaction
  • Multimodal perception (visual + auditory)
• A custom animated renderer for its body, face, gaze, mood, idle motion, and movement
• A thought panel that surfaces internal states like perception, planning, and thoughts
• Built using:
  • Electron
  • OpenAI Realtime API
  • GPT-5 (as declared)
  • JavaScript
  • Three.js
  • Voice-AI tools

Inferred: The product is a desktop-based AI companion, not a chatbot or web app. It is designed to be present and reactive within the user’s desktop environment.

The author states:

• Geist was built as an experiment in alternative interfaces for AI
• It aims to move beyond traditional chatbox assistants
• The goal was to explore a spatial, reactive, emotional, and alive interface for AI
• It is described as a screen-aware ghost pet, not just a tool

Inferred: The positioning is exploratory and conceptual — it does not claim to be a commercial product or a mainstream assistant. It is framed as a research prototype or design experiment.

Not evidenced.

The description does not state:

• Who the intended users are
• What problem they solve for
• Whether there is an identified customer segment
• Any user personas or buyer profiles

Not evidenced.

There is no mention of:

• Revenue model
• Pricing structure
• Monetization strategy
• Customer acquisition plans
• Sales process or go-to-market approach

The description states:

• Built with Electron, Three.js, OpenAI Realtime API
• Uses GPT-5 (as declared)
• Combines vision, perception loops, voice, planner, brain panel, local memory controls
• Designed for realtime interaction and embodiment
• Developed using AI pair-programming tools (Claude, Codex)

Inferred: The technical stack suggests a desktop application with AI integration, likely targeting developers or early adopters of AI interfaces. The use of Electron implies cross-platform compatibility.

Not evidenced.

There is no evidence of:

• Revenue
• Customers
• User adoption
• Product usage metrics
• Iteration history beyond this one submission
• Any commercial or user testing beyond the hackathon context

Not evidenced.

The description does not mention:

• Competitors in the AI assistant space
• Similar products or platforms
• Market positioning relative to existing tools
• Differentiation from other desktop or AI companions

Inferred risks:

1. Prototype-only: The project is a hackathon submission with no evidence of further development.
2. Unproven commercial viability: No revenue, customer base, or monetization strategy are evident.
3. Technical coherence challenges: The author notes that embodiment is hard and requires many systems to align — this may be a significant technical risk if not resolved.
4. No user feedback or testing: There is no evidence of real-world usage or iteration based on user input.
5. Unverified claims: The use of GPT-5 is self-declared, with no verification.

1. What was the intended user base for Geist beyond the hackathon?
2. Has there been any testing or feedback from users outside the development team?
3. What are the plans for further development beyond this prototype?
4. Is there a path to monetization or commercial use?
5. How does Geist’s design handle privacy and data security in a desktop environment?
6. What were the key technical challenges that were not resolved in this version?