Systems

Leverage behind the distribution work

A non-engineer learning to code and building what used to require considerably more calendar coordination.

Production workflow

From conversation to published idea

AI handles repeatable movement. Judgment, context, and taste remain annoyingly human.

  1. 01Meeting transcript
  2. 02Signal extraction
  3. 03Structured Notion entries
  4. 04Drafted copy
  5. 05Publish via Typefully CLI

Used in production

Tools with receipts

USED IN PRODUCTION

Typefully CLI

Custom Bash CLI with 1Password-integrated authentication for publishing across BungeeExchange, SOCKETProtocol, and BungeeCEO.

Approximately 1,400 lines · 3 managed accounts

USED IN PRODUCTION

Marketing memory system

Structured context across Notion and persistent memory files, designed around progressive disclosure rather than loading everything everywhere.

4 Notion databases · 17 persistent memory files

USED IN PRODUCTION

Token-optimized agent architecture

Router-based context loading reduced documented workflow size while retaining task-relevant memory.

Roughly 30K → 3K tokens per message · roughly 600–700K saved per documented 40-message session

Active practice

Systems still learning in public

ACTIVE PRACTICE

Narrative monitoring agent

Tracks crypto narratives, trending topics, and ecosystem or partner activity.

ACTIVE PRACTICE

Meeting-to-task pipeline

Processes Google Meet transcripts through Gemini structured notes before work enters operating systems.

ACTIVE PRACTICE

Ralph loop

Agents audit and upgrade markdown specs, skills, CLIs, and workflows.

LEARNING IN PUBLIC

Programmatic video

Building repeatable brand-video workflows with Remotion while learning to code.