End-user-focused, local-first, privacy-focused, personalized AI experiences.

Maibook is a local, private community of AI members personalized based on your interests, discussing things you care about. Free to try.

• Personalized AI agents that respond to your every request
• Works behind the scenes so you can focus on your priorities
• Local-first desktop app for Mac and Windows running local AI models

• Personalized - end-user focused. Focus on the end user - whether at work, at home, or as a parent.
• Local-first AI. On-device inference where it works (Apple MLX, Ollama, local LLMs); hosted models when the task demands it. Your data stays on your machine by default.
• Privacy by default. No ad telemetry. No accounts that extract you. Tools that respect that you own your data and your attention.
• Often early. Twenty years of independently arriving at ideas before they became mainstream - no-code, visual knowledge graphs, multi-LLM councils, AI second brains.

• macOS - Windows Maibook Your third brain. A private community of personalized AI agents expanding on your every activity. Local-first desktop app.
• Web - $1+ Claude Code CLI Viewer Dark-mode viewer for Claude Code CLI sessions. Browse, search, and visualize AI coding sessions.
• 68 - 9 kindle Haixu Visual Guides AI-powered educational comics spanning science, history, culture, technology. 68+ published on Gumroad, 9 books on Amazon Kindle.
• macOS - MLX ahai Local AI idea discovery for Mac. Scans markdown files to surface forgotten ideas. 100% private, Apple Silicon optimized.

Research and architecture threads we follow.

• World models - Yann LeCun's V-JEPA and world-model program.
• Thousand Brains - Jeff Hawkins and Numenta. Cortical columns, voting-based representation.
• Andrej Karpathy - Projects, posts, and explainers on LLMs and learning.
• Simon Willison - Daily-updated blog on LLMs, tools, and practical AI.
• Computational / cognitive neuroscience - The biological side of intelligence.
• Neurosymbolic AI - Combining learned representations with symbolic reasoning.
• Reinforcement learning - Sutton & Barto - the canonical text.
• The Bitter Lesson - Sutton.
• Arxiv papers - Reading abstracts at the current rate of publication.
• Local-first AI - Apple MLX. Ollama.

• Claude (Code + apps) - Anthropic. Used daily across coding, knowledge work, and Obsidian workflows.
• Gemini Nano-Banana - Google. Image generation and editing.
• Google NotebookLM - Google. Paper clusters and long sources.
• Asta - Ai2. Literature-discovery agent over Semantic Scholar.

• May 2026 Rich personalized UI with AI Four modes for generating personalized AI UI - single-page HTML, web servers, native wrappers, and source-driven auto-generation. The thesis behind Maibook and the rest of MLAI's lineup.
• May 2026 How to use AI effectively - the missing manual Pick the best model you have, invest in the first prompt, debug by fixing the prompt and re-running. A practical guide to getting senior-level work from any AI assistant.
• Jan 2026 The Best Harness Beats the Best Model The code+prompt layer matters more than the model in 2026. Small and open models in a good harness rival frontier models for most tasks.
• Dec 2025 Building ahai - what, how, why A transparent decision log for ahai: local-first MLX, Mac-only, $19+ one-time pricing, no subscriptions, MLX over Ollama, US-origin default model, and why the "open-source as bait" model is morally worse than transparent paid.
• Nov 2025 My AI coding mindset A compact manifesto on working with AI assistants - end-to-end ownership, recursive task decomposition, tests as immutable contracts, dual-doc patterns, parallel sessions designed for diversity.