Notes from the Field: How to Build a Second Brain with Claude Code (Slides)

This is an abstract of a Chinese-language presentation deck (13 slides) covering the Software 2.0 second-brain system. It frames the problem (every conversation contains valuable knowledge that vanishes), the architecture (extraction → classification → storage → continuous learning), the five-category intent classifier (decisions / concepts / problems / learnings / projects), the daily 23:00 automation loop, and the agent-butler decision tiers (L1/L2/L3). The deck pairs with the longer technical write-up “Building a Second Brain from Claude Code Conversations.” Useful as a 10-minute orientation for someone new to the Synapse knowledge-management philosophy.

Key Takeaways

• Software 1.0 vs Software 2.0: Software 1.0 relies on human-defined tags and manual links; Software 2.0 lets AI learn implicit rules from usage patterns and discover associations automatically.
• 100% conversation coverage vs the 0.2% achieved by manual daily-record habits. The system pulls directly from Claude Code’s .jsonl files, leaving nothing on the table.
• Real numbers from one day (April 8): 526 high-value knowledge items extracted from 1,272 messages across 72 sessions. Scale is non-trivial.
• Three-tier decision routing: L1 routine operations auto-execute; L2 plans need butler approval; L3 complex/irreversible actions escalate to human.

Why This Matters

Most “AI knowledge management” pitches miss the cycle: capture → classify → retrieve → re-train. The Software 2.0 framing matters because it inverts the assumption: instead of humans maintaining a tag taxonomy, the AI maintains it from observed patterns. The closing line of the deck — “every conversation is a learning event; every day makes the second brain smarter” — is the mission statement of the Synapse OBS layer. This deck is the cleanest 13-slide summary we have.

---

This is an abstract. Read the full article in Chinese → 如何用 AI 打造第二大脑 - Claude Code 实战分享