21 daily use cases after 2.54 billion tokens
Matthew Berman spent 2.54 billion tokens refining his OpenClaw setup into 21 daily workflows. This is one of the most heavily iterated setups in the community.
What His Agent Handles
- CRM: contact tracking, follow-up reminders, relationship scoring
- Memory: structured markdown for short-term and long-term memory
- Business: email triage, scheduling, content, research
- Projects: active tracking, deadlines, deliverables, status updates
How He Built It
Each of the 21 workflows went through multiple iterations of prompt refinement, memory structure optimization, and edge case handling. The 2.54B tokens represent the sheer volume of interaction needed to tune it properly. MD file structure and memory architecture shared publicly.
Results: 21 daily workflows refined through 2.54B tokens. Memory architecture shared publicly.
I've spent 2.54 BILLION tokens perfecting OpenClaw.