Grilling my own essay

Matt Pocock’s /grill-me is “a relent­less inter­view to sharpen a plan or design” before you commit to it. When I ran it in June, it asked one con­testable ques­tion at a time, until the design was sharp. I pointed it at a fin­ished 99rabbits essay instead. It sur­pris­ingly worked.

The essay was about two pat­terns of how AI agents join human con­ver­sa­tions — inter­cep­tors and demons. One ques­tion got the genre wrong, in a way that’s worth writ­ing down.

Q1 — what is the actual thesis?

The first ques­tion was: which single sen­tence would a reader quote after­wards and say this is what the post argued? It listed five can­di­dates from the draft, ranked them, rec­om­mended the strongest one, and asked me to either endorse it or name a sixth.

My answer was that none of them was the thesis, because the piece doesn’t have one. It’s observational. It’s not argu­ing for a pre­dic­tion or pre­scrib­ing a behavior. It’s saying: here are two struc­turally dif­fer­ent pat­terns, here’s how to tell them apart, here are the con­se­quences each one carries. The con­tri­bu­tion is the dis­tinc­tion itself, not a claim about which side wins.

The skill accepted that imme­di­ately and re-grounded:

Observational essays are a legit­i­mate form: they don’t argue, they distinguish. The con­tri­bu­tion is the dis­tinc­tion itself. Think Christopher Alexander on pat­terns, or any good tax­on­omy essay — before this, you couldn’t cleanly sep­a­rate X from Y; after, you can.

Then it asked a dif­fer­ent ques­tion, scoped to the new genre: does the open­ing deliver the dis­tinc­tion sharply enough? Which is the right ques­tion for an obser­va­tional essay and the wrong one for an argu­men­ta­tive one.

Q2 through Q4 — once the genre was right

Once the skill knew it was grilling an obser­va­tional essay, the rest landed cleanly.

Q2. Second para­graph said “two struc­turally dif­fer­ent pat­terns are in play, and I sus­pect that dif­fer­ence is worth think­ing about.” The skill called out — cor­rectly — that “worth think­ing about” is doing none of the work. It offered three sharper open­ings, each with a dif­fer­ent promise to the reader: clar­ity, recog­ni­tion, design leverage. I picked.

Q3 was visual symmetry. The Pattern B sec­tion had a dia­gram and a screenshot. The Pattern A sec­tion had nothing. The skill named both the case for leav­ing it (form matches con­tent — Pattern A is invis­i­ble) and the case for fixing it, and rec­om­mended fixing. I agreed. Inserted the diagram.

Q4 raised the strongest counter-objection a sharp reader would make; I judged the exist­ing essay already defused it and overruled. Fine — the skill sur­faces con­testable choices, it doesn’t make them.

The skill held up

Three ques­tions pro­duced direct edits, one I overruled. The first cost a cycle to genre-correct, then re-routed. For fin­ished prose, treat the first ques­tion or two as ori­en­ta­tion — the grilling tight­ens once the skill knows what it’s grilling.

If you write non­fic­tion prose, try /grill-me at your next fin­ished draft. Tell it the genre first; the rest ports cleanly.