AI
A transparent look at how I use AI. I work with AI agents every day, for building, debugging systems, writing, and thinking through problems. Rather than just talk about it, I figured I’d show the actual numbers.
This page pulls live telemetry from my agent sessions and surfaces the patterns: how often, how long, and what kind of work, this is generated using sigil.
· Claude Code handles nearly all the heavy lifting — 99% of tokens, with a median session of 24 minutes. OpenClaw sessions are fewer but run much longer (median 5+ hours): quick iterative work vs long autonomous runs.
· 86% of sessions involve tool usage. Most interactions are "doing" — editing files, running commands, searching code — rather than pure conversation.
· Mornings are the peak across both tools. Weekdays account for 80% of all sessions, with Tuesday being the busiest day.
· Feb 22 was the most active day this period with 24 sessions. Only 5 days had zero activity — roughly 83% daily coverage.
· Compared to the prior 30 days, token share shifted dramatically toward Claude Code (41% → 99%), with OpenClaw dropping from 59% to 1%.
AI Policy #
All the ideas, arguments and opinions on this site are mine. The thinking is mine. The mistakes are mine too.
I use AI tools extensively in my writing process. Editing, restructuring, catching errors, tightening prose. If you’ve read something here that flows well, there’s a good chance an LLM helped me get it there. But the core of every piece, what I’m trying to say and why, that’s always me.
Things I use AI for on this site:
- Editing and proofreading drafts
- Restructuring narrative flow
- Editorial feedback on tone and clarity
- Minor site maintenance
Things I don’t use AI for:
- Generating ideas or opinions
- Writing drafts from scratch
- Deciding what’s worth writing about
Inspired by Vicki Boykis’ AI page.