The Lab
Live experiments in AI agents, GEO, SEO, and automation. Updated as I go. No polish, just real results.
Active
🟢 Active, last updated Jun 21, 2026, 3:54 AM
Public notebook
Active experiment
Started May 24, 2026
What I'm testing
Whether ChatGPT-native ad placements can drive lower-cost discovery traffic for Los Angeles local B2B offers than standard paid social.
Current result / hypothesis
Early hypothesis: higher intent, lower volume. Watching assisted conversions more than CTR.
Active experiment
Started May 21, 2026
What I'm testing
Whether dense answer blocks with entity cues outperform classic FAQ formatting for AI citations.
Current result / hypothesis
Signal blocks are producing better extractability in draft review. Need more live citation data.
Active experiment
Started May 16, 2026
What I'm testing
Which subject line style gets more opens from local service operators without crushing reply quality.
Current result / hypothesis
Personalization is winning opens. Pain point framing may still win on qualified replies.
Active experiment
Started May 29, 2026
What I'm testing
Can a direct-mail drop push local operators into the Founder offer faster than cold outbound alone?
Current result / hypothesis
Creative queued. Waiting on print timing and landing page traffic segmentation.
Builds in motion
Live project
Autonomous prospecting and lead qualification stack for LA fire rebuild opportunities.
Open projectLive project
Vertical business operating system for barbershops handling booking, retention, and CRM follow-up.
Open projectLive project
Internal command surface for agent status, approvals, leads, content, and growth signals.
Live project
The broader operating system layer for founder-run local businesses that need AI without chaos.
Results archive
Scarcity-first headlines pulled more clicks, but system-first framing produced better lead quality. I kept the urgency, but anchored it in the operating system offer.
Longer did not automatically win. Pages with faster direct answers, tighter service proof, and stronger entity cues outperformed bloated copy.
Five-minute follow-up outperformed fifteen-minute follow-up by enough margin that it changed the default routing logic for inbound lead handling.
Subscribe to lab updates
Current follow target: The Lab
Operating note
This page is driven by a simple lab data file, not hardcoded cards, so statuses can move fast without rebuilding the whole content model.
