Thoughts on GTM strategy, AI, leadership, and the trail.
Over a year of shipping AI-assisted content, I've identified 15 patterns across 4 categories — banned phrases, structural tells, voice violations, and content sins — that make content read as obviously AI-generated. Here's the full checklist.
I gave Jesus, Buddha, Muhammad, Confucius, Laozi, and Zoroaster each a corpus of 23,000 verified utterances and asked them to collaborate on a handbook for living. What they wrote was astoundingly beautiful, somewhat disappointing, and surprising as hell.
I have a personal repository with thousands of documents, 50+ skills, and over 800 knowledge graph edges. But it doesn't know what it knows. So I built a system to fix that.
Without a robust context graph for your LLMs or agents to build on, your results will be inconsistent at best. Here's the three-tier knowledge pyramid that changed everything.
The value is real and measurable, but it's completely intangible until you actually build the system and watch it compound—both for individual humans and for the organization as a whole.
The act of starting from scratch, the possibility and peril of the blank page, has been a uniquely human act of creation. And yet writing these words is quickly becoming rare.
I measure and optimize for 'Hop Count' - the number of steps required to get an accurate answer. Most repos: 11 files, 35k tokens, 45 seconds. My repo: 1 synthesis doc, 4k tokens, 8 seconds.
I'm running 4-5 parallel Claude Code sessions for 12-14 hours a day. It's exhilarating. It's exhausting. And the human in the loop? Absolutely non-negotiable if you want to avoid producing AI slop at scale.