Descriptions:
Nate B. Jones walks through building a personal token burn dashboard using OpenAI Codex, framing 800 million tokens consumed in a single day not as a boast but as a lens for understanding how AI usage habits shape productivity. The dashboard tracks consumption across Codex, Claude, and ChatGPT sessions, rendered in a GitHub contribution-style chart using an open-source Tufty data visualization skill.
A recurring technical thread is the gap in Claude’s tooling: unlike Codex, Claude Code and Claude Chat do not expose token counts to users outside the API. Jones describes using Codex itself to reason from session artifacts and logs to approximate Claude usage — an irony he acknowledges directly and flags as something Anthropic should address. The build also incorporates the /workflows feature from Claude Code’s Opus 4 release, which dynamically generates an orchestration plan and spins up sub-agents to execute parallel tasks, and which Jones ported from a community open-source skill into Codex.
The broader argument is that visibility into AI usage patterns enables genuine self-improvement: Jones’s dashboard revealed that his highest-value days correlated with delegating structured database work to Codex, a behavioral insight that wouldn’t have surfaced without the chart. The video also covers using Codex to batch-label and organize screenshot archives, illustrating how delegated file management fits into a fully AI-assisted computing workflow.
📺 Source: AI News & Strategy Daily | Nate B Jones · Published June 05, 2026
🏷️ Format: Workflow Case Study
