Prism

Next-gen desktop AI workspace powered by Gemini. Features multi-model chat, system-wide Quick AI, browser automation, and MCP tools wrapped in a high-performance, local-first interface.

Comment

Inspiration & The Story of Building Prism

Digital workers, researchers, and students lose hours every day to cognitive friction and context-switching, toggling endlessly between web browser tabs, code editors, flashcard applications, and third-party AI portals. I built Prism to make intelligence ambient, zero-friction, and OS-native. Powered by Google Gemini, Prism brings system-wide writing, autonomous tool execution, and connected study workflows directly into your native desktop environment.

Starting as a solo developer during the hackathon, I set out to build an AI-native company where intelligent agents handle continuous operations while human creativity drives architectural direction. Operating with minimal overhead, I engineered a high-performance desktop workspace that connects directly to Google Cloud Vertex AI and the Gemini API, enabling deep reasoning and ambient intelligence without the latency of clunky web applications.

What It Does & Category Impact: Education and Human Potential

In the category of Education and Human Potential, Prism unlocks human capability by removing the barriers between thought and execution:

  • System-Wide Quick AI Overlay: Summon Gemini over any active application with a global hotkey (Ctrl + Space) to stream responses, rewrite text, or analyze multi-modal attachments mid-sentence with zero window toggling.
  • Gemini 3.6 Flash & Vertex AI Integration: Low-latency Server-Sent Event (SSE) streaming with live, animated thought-reasoning process rendering.
  • Model Context Protocol (MCP) Registry: Autonomous @-tool mentions for local file creation, workspace agents, and custom API execution.
  • Browser Automation Engine: Localized Node.js automation server (localhost:9090) driving Playwright and Puppeteer to execute agentic web scraping and DOM tasks with live action logs.
  • Connected Learning Suite: Interactive Quiz Me mode, spaced-repetition SM-2 Flashcards, and native LaTeX math rendering: $$\int_0^\infty e^{-x^2} dx = \frac{\sqrt{\pi}}{2} \quad \text{and} \quad i\hbar\frac{\partial}{\partial t}\psi = \hat{H}\psi$$
  • PrismCode Companion Editor: Native code editor target with integrated terminal (SwiftTerm) and Git version control.

By embedding Gemini directly into the operating system, students and researchers transform passive reading into active mastery through automatically generated quizzes, spaced-repetition schedules, and synthesized deep research reports with verifiable citations.

How I Built It & Day-to-Day AI Operations

As a solo founder, AI is integral to my day-to-day operations and development:

  • Frontend / Native OS: Swift, SwiftUI, Liquid Glass UI engine, and macOS Accessibility APIs for system-wide text overlays.
  • CLI & Tooling: TypeScript monorepo (prism-cli), Node.js, esbuild, and Vitest.
  • Backend Infrastructure: Appwrite Cloud Functions for authentication, student .edu verification, and team workspaces; Azure Functions for hosted chat streaming.
  • Windows & Mobile: Native WinUI 3 / .NET 9 Windows companion app and Jetpack Compose Android companion app.
  • Day-to-Day Development Operations: I use agentic coding tools (Cursor, Copilot, Devin, Antigravity) alongside Gemini to run continuous unit tests, draft cross-platform bindings, and triage daily error logs.

Human vs. AI Division of Labor

Prism operates with a clear separation of responsibility between human judgment and autonomous AI execution:

  • What AI Does: In production, AI ToolRouters classify user intent in real time, route tasks between local file systems and external MCP servers, autonomously navigate complex web pages to extract research data, structure dynamic study flashcards using SM-2 intervals, and generate multi-modal responses. Operationally, AI assists in parsing customer telemetry and monitoring API health.
  • What Humans Do: As the founder, I focus on core system architecture, user experience design, security boundaries, payment compliance, customer relationship management, and strategic roadmap planning. Human oversight ensures that AI tool execution remains sandboxed, private, and aligned with user consent.

Economic Opportunity and Job Creation Beyond the Founder

Prism creates tangible economic opportunities for individuals and businesses beyond the founding team:

  • Student & Educator Enablement: By reducing research and study times by over 40%, students accelerate time-to-graduation and skill acquisition in high-value technical and quantitative fields. Verified .edu discounts make high-end AI tools accessible to underserved learners.
  • Freelancers & Knowledge Workers: Independent researchers, technical writers, and consultants use Prism to automate browser workflows and compile deep research dossiers, multiplying their billable output and enabling solo operators to offer enterprise-grade research services.
  • Ecosystem Job Creation: As Prism expands into Team Workspaces and third-party MCP tool registries, external developers can build and monetize specialized MCP plugins, creating a marketplace economy for independent tool creators.

Challenges I Faced

Engineering low-latency SSE streaming through proxy layers, building a sandboxed Playwright browser automation bridge, and rendering complex LaTeX mathematical formulas inside fluid SwiftUI glass containers without performance degradation.

Accomplishments & Real-World Traction

Built for production, Prism reached 82 registered signups and 19 paid customer transactions, generating $543.50 in gross revenue at a 98% gross margin with zero paid marketing spend. Active subscribers are currently renewing three months running.

Built With

Share this project:

Updates