Direct. Clear. Meaningful. Let AI write the boilerplate, but make sure YOU learn the craft.
The Problem
AI coding tools (GitHub Copilot, Cursor, Claude Code) have created a silent paradox:
You ship faster. But you learn slower.
Every day, AI writes thousands of lines you accept without truly understanding — complex async patterns, architecture decisions, subtle edge guards. Three days later, you can't reproduce or debug them independently.
LFAI closes that gap. It automatically turns your daily AI-assisted code into a personalized 5-minute learning session — not generic tutorials, but knowledge extracted from your own work.
✨ Features
⚡ Dual-Engine Workflow: Work Delivery + Tech Growth
LFAI recognizes that developers don't need another chore at 6 PM. It bridges work and learning seamlessly:
For Work (Public Standup & PR): Automatically distills today's engineering changes into 3-5 high-signal bullet points. One click on "📋 Copy Standup/PR" copies a clean markdown summary ready for your daily standup, team report, or GitHub PR description.
For Growth (Private Knowledge Base): Drops all boilerplate and filters out boring CRUD to extract high-density technical cards into your personal knowledge base.
🎯 Staff-Level Deep Probing Pillars
Unlike generic code commentators that parrot what the code literally does, LFAI enforces negative filters (no basic syntax, no CRUD storytelling) and applies 4 probing dimensions:
💡 The Why (机制剖析): Behind-the-scenes architectural intent — why choose this specific pattern or data flow over naive approaches?
⚠️ Destructive Test (破坏性推演): What breaks if you delete this line of defensive code? Under what extreme concurrency, race conditions, or unexpected inputs will it silently crash?
📖 Deep Mechanics (底层原理): Event loop microtasks, V8 optimizations, buffer overflow limits, and framework internals.
✨ Best Practices (生产级实践): Production trade-offs and modern modern language idioms.
❓ Active Recall Quiz (实战拷问): Interactive Anki-style quiz testing real mental models with reveal-on-demand explanations.
🔍 Automatic Daily Code Digest
Scans today's Git incremental changes, filters noise (lockfiles, build artifacts, formatting-only diffs), and surfaces the real learning signal.
🖥️ Native VS Code Sidebar Panel
Filter cards by category with one click (Gotchas, Architecture, Syntax, Quiz)
One-click copy for daily standup / PR deliverables
Expandable deep-dive blocks (The Why, Destructive Test, Best Practices)
Interactive quiz mode with reveal-on-demand answers
Copy code snippets directly to clipboard
📝 Markdown Knowledge Base
Every daily digest auto-saves to .lfai/daily/YYYY-MM-DD.md — cleanly organized with a Work Deliverables section and a Technical Deep Dive section. Compatible with Obsidian, Logseq, Notion, and any markdown note-taking tool.
🤖 Zero-Config LLM Support
VS Code LM API (default): Uses your existing GitHub Copilot subscription — zero API key required
Custom LLM: DeepSeek, OpenAI, Claude, or any local Ollama endpoint
✂️ Selection-Based Instant Learning
Right-click any code selection → "LFAI: Extract Learning from Selection" — on-demand deep analysis of any snippet.
