In the final part of Art, Automation, and AI, Tim Grahl and Shawn Coyne bring the series home. Tim starts with his own confession that when ChatGPT arrived, he worried a machine would write a book before he did. Shawn explains why asking a large language model for feedback on your writing keeps you in a wicked learning environment: it’s an inference machine built to satisfy you, not a deterministic one built to tell you the truth. From there they walk through the Story Grid Dojo, where writers start with three sentences, work through the five commandments across the three levels of reality, and get consistent, standardized feedback within 24 hours. Shawn describes the daily practice he wishes he’d had at 25 and why every student who shows up is the one.
Tim opens by returning to a story from earlier in the series. When ChatGPT 3.5 first hit, he told his wife that if a machine learned to write a book before he did, he was done. That led to the question underneath the whole series: why are you actually doing this? He recaps the first three parts, then goes back to a loose end from Part 3. People are using LLMs for feedback on their writing. Why does that keep them stuck?
Shawn’s answer is that an LLM is trained to satisfy the person asking. It’s a confirmation bias machine. Ask it whether your scene has the five commandments and it will find them. Ask whether you write like F. Scott Fitzgerald and it will tell you how close you are and offer a revision. It cites authority convincingly, but it has no understanding of truth and no capacity for artifice. It isn’t lying on purpose; it doesn’t know it’s lying.
Tim shares his own experience: the same scene with the same prompt gets completely different feedback the next day, and an elaborate “edit like Shawn Coyne” prompt made things worse. He connects this to the research on learning environments. Random feedback from non-experts is bad. Random feedback from something that presents itself as an expert is worse than no feedback at all.
Shawn explains why. An inference machine calculates probabilities, letter by letter, and will never give the same answer twice. A deterministic machine runs on consistent rules, like a calculator. One plus one is two every time, all day, every day. His foundational principle is that artifacts like sentences and myths have deterministic features that can be reliably repeated, which goes back to Claude Shannon’s information theory: clarify the signal and remove the noise so the message survives the channel between writer and reader.
That is the basis of the Story Grid software: not an LLM, but a deterministic system Shawn has built over decades, grounded in the formal proofs of Church, Turing, Deutsch, and Shannon. Submit the same thing and you get the same response, currently around 96.7 percent of the time. When a wobble shows up, Tim reports it and Shawn fixes it. Students submit in a standardized form and get standardized feedback, and Shawn deliberately doesn’t know most of them, because an editor who likes a writer goes easier on them.
Tim plays the whole argument back. Myth is an objective standard. Shawn has pulled out the repeatable patterns that make a story mythic, from the sentence up to the epic, and encoded them as a long series of if-then gates. Shawn confirms it and adds that he’s not the first to look for the commonality in stories that outlast their creators; Jung, Campbell, and Aristotle did the same. The Odyssey still works because we all want to go home. Myth is the deep memory of the species, and that’s why Story Grid asks students to find a memory with a story’s shape before they write anything.
From there they get specific about the Dojo. Students begin with three sentences: an inciting incident driven by antagonism, a protagonist’s reflexive response, and an escalation back from antagonism. Progressive complication has been the hardest thing to teach for eight years, so the Dojo starts there, before any of the five commandments. Students write those three sentences on each of the three levels of reality: the physical body, social belonging, and ideological belief.
Shawn then walks through the four scales of antagonism. Individual antagonism is direct and targets your body. Group antagonism is diffuse and makes you question where you belong, like the locker room scene in Carrie before the ringleader steps forward. Institutional antagonism is more diffuse still, like a letter from the IRS with no signature. Environmental antagonism wraps around all three, like a storm hitting your picnic, and it can individuate, as when a gust of wind throws your wife into you and off a cliff. Tim frames all of it as a toolbox: you don’t use every tool every time, but you have to be able to use each one when the writing calls for it.
By the end of the fourth lesson suite, a student has 15 working sentences, the five commandments across all three levels of reality, that form the core of a scene and resolve with uncertainty so the reader wants the next one.
Shawn closes with how he’d use the Dojo himself: 20 to 30 minutes a day on an assignment, predict which sentences will pass, get feedback within 24 hours, revise, and resubmit. It’s a warmup before tackling the manuscript in the closet, and like the gym, every day makes you a little stronger. He named it the Dojo years ago in The Story Grid Universe, borrowing from The Matrix. Downloading kung fu into Neo’s head didn’t mean he could fight. He had to get in the dojo with Morpheus and hear “nope” until he couldn’t be beaten. Morpheus never lied to Neo, and he believed he was the one. That is how Story Grid feels about every student.