Overview
Syllabus
Why I’m ditching “chat interface” AI for real work
The new era: OS-level AI agents and why it matters
Agent #1: OpenClaw setup + “Hermy” custom UI attempt failed
Agent #2: “New Agent” prototype working UI, messy bugs
The audit method: stress-testing plans across multiple models
The breakthrough: 80% of customization doesn’t need a fancy UI
Safety talk: don’t install sketchy agents on your machine
Project Meridian: the portable AI Brain folders + .md only
Live HUD dashboard + cognitive sliders dials & gauges
File structure tour: gauges, identities, memory protocol, personalities
Demo: the agent refuses to overwrite itself safety rails
MasterSpec.md: capability handshake + strict output order
Memory retrieval + persistence protocol write memories every turn
Download from GitHub + how to “drop in” the brain anywhere
Live setup demo: “Rowan” bio researcher + synthetic data test
Auto-personality creation + memory files being written
Error/crash + recovery + web research test run
Wrap-up + what’s next and no more scripts in 2026
Taught by
MattVidPro AI