You Don’t Need SaaS. The $0.10 System That Replaced My AI Workflow
Source: YouTube — Nate B Jones, published 2026-03-02 Link: https://www.youtube.com/watch?v=2JiMmye2ezg
Summary
The definitive OpenBrain architecture guide. Nate B Jones lays out why every AI user needs their own memory infrastructure, details the full Postgres + PGVector + Supabase + MCP architecture, and argues that the gap between “I use AI sometimes” and “AI is embedded in how I think” is the career gap of this decade — and it comes down to memory infrastructure.
The Walled Garden Problem
Claude’s memory doesn’t know what you told ChatGPT. ChatGPT’s memory doesn’t follow you into Cursor. Every platform has built a walled garden of memory — five separate piles of sticky notes on five separate desks. Memory is designed as lock-in. This spawned an entire VC-backed industry (Memsync, One Context).
Human Web vs Agent Web
The internet is forking: the human web (fonts, layouts, reading) and the agent web (APIs, structured data, machine-to-machine). Your Notion workspace is built for human eyes — beautiful, but useless for AI agents that search by meaning, not folder structure. Second brain tools solved the capture/organization problem but not the agent-readable problem.
The Architecture
Capture: Type a thought (Slack, any messaging app) → Supabase edge function generates vector embedding + extracts metadata (people, topics, type, action items) → stores in Postgres with PGVector. Round trip under 10 seconds.
Retrieval: MCP server connects to any compatible AI client. Three tools: (1) semantic search (find by meaning), (2) list recent (browse this week), (3) stats (see patterns). Works from Claude, Claude Code, ChatGPT, Cursor, VS Code — anywhere.
Cost: $0.10–0.30/month on Supabase free tier, ~20 thoughts/day in API calls.
Setup: ~45 minutes, no coding required. Tested with non-technical user.
Four Lifecycle Prompts
- Memory migration — extract existing memory from Claude, ChatGPT, etc. into OpenBrain. Run once after setup.
- OpenBrain Spark — interview prompt that discovers how the system fits your workflow; generates personalized capture suggestions by category
- Quick capture templates — 5-sentence starters for clean metadata extraction: decision capture, person note, insight, meeting debrief
- Weekly review — end-of-week synthesis clustering by topic, scanning for unresolved action items, detecting cross-day patterns, finding connections missed
The Compounding Advantage
Person A opens Claude, spends 4 min explaining context, gets good answer. Person B opens Claude — it already knows her role, projects, constraints, team, decisions from last week via MCP. Every thought captured makes the next search smarter. The gap compounds weekly.
“Memory architecture determines agent capabilities much more than model selection does.”
Key Correction to OpenBrain Concept Page
The /loop video described OpenBrain as “SQL database, no vector store, no embeddings.” This video reveals the full architecture does use vector embeddings (PGVector) for semantic search. The earlier description was simplified.
Pages Created or Updated
- OpenBrain — major update with full architecture
- Nate B Jones — updated
See Also
- loop — the proactivity primitive paired with OpenBrain
- OpenClaw — what OpenBrain + /loop replaces
- Four Prompting Disciplines — the framework OpenBrain serves (context/specification engineering)
- RAG vs Wiki — OpenBrain uses semantic search (embeddings) for personal knowledge