A 20-page partner-skinned Quick Audit is not done when the model says done. Deterministic checks, judgment nudges, UNKNOWN vs zero, and the private GitHub note.
How We Stopped Handing Agents a Paste
A Cursor agent wrote a perfect prompt instead of publishing. We connected MCP, hit the card, and made the rule a public URL.
If You Can Do It, You Should
Agents do the work. You hit the auth card. Completeness is free now, so boil the ocean instead of handing back a paste.
How we put Greatest Hits on every topic wheel and money tree
Gold three-card strips on the topic wheel and money tree, plus the canonical method at /greatest-hits/.
How We Harmonized Skills and Routines
xAI published skill vs routine. We already had Claude scheduled tasks. The Grok Bot pages named the roster and skipped the two buttons.
A Skill Is How. A Routine Is When.
Claude scheduled task = Grok Bot routine = Cursor Automation. Do the job once, save the method, test, then put it on a clock.
How we put a topic wheel on every personal brand site
Every personal brand site now ships a colorful clickable topic wheel paired to its money tree. George Paladichuk is the worked example — 22 chips, NaiL in the center.
Grok Bot Is a Roster, Not Four Experts
Cursor is the workshop. Grok Bot is the named staff. SuperGrok Heavy’s four-at-once trick is a smarter one-reply, not twelve teammates with photos.
How We Stopped Being the Team’s VA
The always-reply essays already existed. We made one two-minute standard for A-players, and a filter for people no SOP will fix.
Always Reply
If the question is to you, answer on that thread within 24 hours. Another thread is not a reply. Dennis is not your VA.
How We Taught Magic Berry by Example
Recut: 205-row public harvest, 25 highlight berries, two scores over 20, embeds on the wall.
Agents Need a Line, Not a List
The Task Library was a flat list of 240 SOPs. We scored every task 1 to 5, put the Content Factory line on the dashboard, and made Claude-only or Grok-only a first-class path.