Enai seeks to create a computing environment that organizes itself around your intent to protect your attention and save you time.
The current Enai version is an experimental prototype named after Robinson Jeffers, a poet from Carmel whose work focuses on the cold, ferocious beauty of wild animals and the Pacific Coast. https://literaryamerica.net/LiteraryAmerica_RobinsonJeffers.pdf
Most of the people who collectively invented personal computing imagined computers as extensions of human thought:
- Vannevar Bush's Memex (1945): An augmented memory device where information forms associative trails, not isolated files
- Douglas Engelbart's NLS (1968): A system for augmenting human intellect through collaborative knowledge work
- Alan Kay's Smalltalk (1972): A living environment where users shape their tools, not the reverse
But the average amount of time people can focus has dropped precipitously in the past few years, and knowledge workers now spend half their time organizing stuff and getting distracted. Something went wrong.
The Alto-Mac paradigm defined in the 60s and 70s is aging. And it's based on assumptions that are no longer true: isolated devices, siloed native apps, and a different kind of computer intelligence. A paradigm shift in computing is happening. From a clean piece of paper, what kind of computer would we build knowing what we know now?
Everything is an object with contextual awareness and memory
Files (emails, spreadsheets, images...), conversations, traditional web content, apps, and AI applets are 'cognitive objects' that remember their context, relationships, and usage patterns.
The browser is infrastructure, not an app
Like display layers in traditional computing, the net is an ambient layer deeply integrated in the Enai environment. We're not trying to build an electron browser, but to use the browser engine as the foundation for a holistic, web-first computing experience.
An intelligent substrate
AI isn't bolted on through chatbots or copilots—it's woven into how objects are understood, related, and transformed. Enai gives AI agents the ability to respond primarily not with text, but by composing and recomposing your information environment based on your stated and implicit intent and preferences.
The environment learns and adapts
Your patterns of thought and work reshape the environment, creating a truly personal computer. Information is organized in one dynamic layer by meaning and intent instead of application or file format.
Second brain cognitive computing Enai’s data model is a machine to augment and mirror the human brain. It consists of an intent stream, working memory, long term memory, and an ontological model. Each of these parts works together to extend human intelligence while providing a powerful context for agent orchestration.
The current implementation explores these ideas through a kernel-based architecture:
Cognitive Kernel
├── Object System — Everything is a live object with behavior
├── Memory Layers — Working memory (WOM) and long-term memory (LOM)
├── View System — Objects can present themselves in multiple ways
├── Intelligence — AI-native understanding and transformation
└── Message Passing — Objects communicate and coordinate
Instead of an app, a computing environment is a container for websites, docs and apps.
Calm: Paperlike textures, warm colors, and human-paced interactions respect your perceptual wellbeing.
Focused: Supporting your intent and attention. The environment brings together everything you need, maintaining context across sessions.
Interpersonal: Share not just documents but entire contexts. Collaborate in shared knowledge spaces that organically preserve semantic relationships.
- Runtime: Electron 35.1.5 with Next.js 15.3.0
- Storage: SQLite (better-sqlite3) with LanceDB for vectors
- Intelligence: OpenAI models via LangChain
- Language: TypeScript throughout
- State: Zustand with IPC persistence
Prerequisites: Node.js 20+, npm 10+
git clone https://github.com/yourusername/enai.git
cd enai
npm installCreate a .env file:
OPENAI_API_KEY=your_openai_api_key
# Optional
EXA_API_KEY=your_exa_api_key # Web search
BROWSERBASE_API_KEY=your_key # Web scraping
npm run dev # Start development environment
npm run lint # Run linting
npm run typecheck # TypeScript checks
npm test # Run test suite- Object Management: Lifecycle, persistence, and identity of cognitive objects
- Memory Management: Working memory (WOM) and long-term memory (LOM) layers
- View System: Multiple representations via WebContentsView
- Intelligence: Embeddings, understanding, transformation
- Message Bus: Inter-object communication
/electron/ Main process and IPC handlers
/src/ Renderer process (Next.js)
/models/ Data models (SQLite)
/services/ Business logic
/shared/types/ TypeScript definitions
The first commit for v2 was 18 April 2025. This was an experiment to see if it's possible for me to build Enai on my own. The next goal is to take a first step towards designing an architecture that can support that claims being made about a computing environment. In particular:
- Object protocol: Complete cognitive object AI pipeline and introduce object sourcing from Gmail and Gdrive APIs
- Intent system: Evolve natural language interaction
- Local search enhancements: Complete a RAG pipeline that finds and outputs appropriate data to hydrate composable objects and applets
- Browser as infrastructure: Incremental first steps towards chromium based object rendering (not just "websites") and general electron browser robustness
See CONTRIBUTING.md for guidelines and CLAUDE.md for AI-assisted development instructions. Your collaboration is welcome.
This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.