Every scribblers’ room I’ve ever sat in has a version of the same wall. Index cards in different colors for different characters. String sometimes, if the mythology runs weird enough that somebody needs to see the connections instead of just say them out loud.

Doesn’t matter if it’s Alias or Lost or Heroes or a pilot that never got picked up. The wall is the room’s memory, because nobody at that table, no matter how good the pitching gets on a given Tuesday, can hold an entire season in their head at once.

I need that wall more than most people at the table. I’ve got ADD, diagnosed late, well into my career, after my son got tested for it and I recognized every symptom on his list as my own. Craft doesn’t stick in my working memory the way it does for most scribblers.

So a few years back, on a hiatus between shows, I sat down and built myself a personal version of that wall. Every fundamental I’d ever learned about scribbling a scene, scribbled out by hand: track the character emotionally, make it cinematic, the power of the time cut. A mnemonic I rewrote every morning until it stuck.

I wasn’t inventing anything. I was just doing by hand what a whole category of software wants to sell scribblers as a product now.

Two Volcano Movies, an Asteroid and a Comet

Hollywood’s had this problem since before I got in the business. Dante’s Peak and Volcano hit theaters the same year, months apart. Armageddon and Deep Impact hit the same summer after that, an asteroid picture and a comet picture.

Every so often the town develops a shared unconscious, and two studios greenlight the identical high concept in different buildings, neither aware the other exists ‘til the trades break it.

Something close to that just happened in AI scribbling software, except it happened three times to the same word.

Storyflow.so is a visual workspace, notes and research and story material and images sitting on a canvas the model can read. Storyflow-ai.com takes you from script to storyboard to AI-generated video.

StoryFlow Editor, out of a small studio called Soulstices, is a branching-narrative scripting tool that landed in Steam Early Access this past November and plugs straight into Unity, Unreal, and Godot.

Three companies. Three different jobs. Funny collision.

Ask why it happened and you land somewhere more useful than the joke. An entire industry asked a language model for something like story plus workflow, and everybody’s model handed back the same portmanteau.

We spent the first couple years of this technology arguing about which chatbot to use. That argument’s basically over. The one that matters now is what gets to walk into the room with actual authority once the model’s done proposing things.

What the Room Actually Does

BookBub surveyed something like twelve hundred authors last year and found about 45% using generative AI somewhere in the process. Skews hard toward working commercial self-published scribblers, so don’t read that as gospel for every scribbler alive.

But look at what the AI-using half actually does with it. Eighty-one percent for research. Outlining, marketing copy, organizing a plot, a second opinion on a scene that isn’t working. ChatGPT led at 85%, Claude a distant second at 54%.

A newer survey out of Kit this year, 550 working creators, found nearly the same shape: ChatGPT 73%, Claude 70%, Gemini 40%, people running two or three at once instead of picking one and going steady. And 89.2% of Kit’s respondents said they always review and edit the output before it goes near a reader.

Not one respondent claimed they used it straight, no notes. That’s every table read I’ve ever sat in. The room generates. Somebody with actual standing still calls it, works or doesn’t, before anybody downstream trusts it.

Sudowrite’s Story Bible is close kin to the mnemonic I built by hand on that hiatus, except it lives in software and the model can read it directly: characters, world rules, synopsis, beats.

Even Sudowrite’s own advice, once you’ve got something screenplay-shaped out of it, is to walk it over to Final Draft, Highland, or WriterSolo and let dedicated software handle the formatting a reader actually trusts.

Smart split. You want the guessing half of the stack good at guessing, and the counting half never guessing at all.

Novelcrafter takes the split further and makes it explicit. Its Codex is a project wiki, characters and places and lore, and you can point almost any brain at it: GPT, Claude, Gemini, Llama, Mistral, or a few hundred more through OpenRouter.

The story bible stays put. Whatever’s plugged into it is a rental. And Final Draft, sitting at the far end of that pipeline doing the least glamorous job in the building, still runs on something called a Beat Board. A digital corkboard. Software finally caught up to the wall.

The Guild Already Drew the Line

I’ve carried a WGA card since 2001. The 2026 MBA keeps the AI language the Guild fought for in the 2023 strike, and it’s worth restating, because people outside the business don’t understand how deliberately that language sits.

AI is not a writer under our contract. Material a company hands a scribbler that came out of a model doesn’t count as source material when the Guild decides whose name goes on screen.

A studio can’t make using it a condition of employment. If a scribbler wants to use it anyway, the company must consent, and whatever policy company has still governs on top of the contract.

The new agreement adds a piece I like. If a studio licenses our scripts, our life’s actual work, to train a commercial model that generates new material, the company must notify the WGA. The Guild can then sit down and talk about what scribblers are owed for it.

The Dead King Walks Into the Tavern

Games don’t get the luxury of fading out. A script can sit wrong for a week until somebody catches it in the next draft. A character inside a live game might be talking to somebody at two in the morning and has to remember, mid-conversation, who’s dead, which door is actually locked, what that particular player already knows and hasn’t been told twice.

Inworld built its name on exactly that problem, AI characters and NPCs. By this year the company describes itself several floors further down: realtime voice, speech recognition, model routing, context management, an orchestration layer moving across more than two hundred models from a pile of different providers. They’ve raised close to $120 million chasing a question that used to just be a narrative designer’s headache and is now infrastructure.

I’ve never had an AI character forget the rules on VALORANT. We don’t run generative dialogue there. But I’ve had a human scribbler do it. Pitched a scene deep in an Alias room once, built around a piece of Rambaldi tech our mythology said didn’t exist on our timeline. Happens in every room that runs long enough. The difference is somebody catches it before the table read.

That’s the company LoreWeaver is chasing, and their pitch line is blunter than most of what crosses my desk: large language models are fluent, not faithful. Their examples read like a narrative designer’s blackout nightmare.

A king you killed three scenes back wanders into the tavern and starts making conversation.

A scholar scribbled gentle as a single adjective in a design doc threatens somebody with violence because the sentence sounded plausible in the moment and nothing in the system stopped it.

Their fix sits a validation layer between what the model wants to say and what a player actually hears: pull the real canon, generate inside it, check the output against the rules, kill whatever breaks them. The model proposes. Your rules dispose.

The Note About the Note About the Note

The failure that actually worries me isn’t the dead king. You catch that one. It’s loud, it announces itself. It’s the slow leak.

Every scribbler who’s worked inside a network system knows this move already. A note comes down from the network. The head of current passes it to the exec assigned to your show, with a little spin of their own added on. That exec passes it to your showrunner, more spin. By the time it lands on you in the room, three people have already reshaped it.

A note about a note about a note, and every single person in that chain swears they’re just relaying what the last person told them.

AI scribbling systems are building the identical chain into their own architecture, on purpose. No model holds an entire long project in its context window at once.

Long projects lean on summaries, retrieval, character memories, compressed versions of scenes that happened three hundred pages back. The model generates something a little bit off. A second pass summarizes it. A third system retrieves that summary later and treats it as fact.

Nobody flags the original mistake three hops downstream. By then it doesn’t look like an error. It looks like established character.

Four Jobs, New Hardware

Strip the product names off everything above and you get the same four jobs every show I’ve ever run already had, just running on different hardware.

Somebody generates. That’s the room, or now, on a slow morning, a model sitting in for the room.

Somebody remembers. That’s the bible: a whiteboard, a Codex, a Beat Board, or a mnemonic scribbled out by hand on a hiatus nobody but me remembers.

Somebody turns pages into a thing that can actually get shot, printed, or shipped.

And somebody has to decide if the scene actually works, if the character would really say that, if breaking your own canon is a mistake or the best idea anybody’s had all week.

Adobe ran a survey this past May, better than sixteen thousand creators worldwide. Seventy-five percent called AI integrated or essential to how they already work. Eighty-five percent said the final creative call should always stay with the human being.

Read those two numbers side by side and they’re not fighting each other. The second one is probably what makes the first one survivable. Nobody hands over the wheel unless they trust there’s still a hand on it.

Guard the Wall

So this is where I’ve landed, and it’s a smaller point than the software companies are selling it as. The wall doesn’t go away because the pushpins can argue with you now. If anything it matters more, because it’s holding against something that can generate a thousand plausible index cards a minute and has no idea, none, which ones are true.

Pick your model. Swap it out next year. I’m not precious about ChatGPT versus Claude versus whatever ships next, and neither should you be.

But build the wall first, in your own hand if that’s what it takes, before you let anything start pinning cards to it for you.

I had two folders full of blog posts just sitting on my desktop. Fifteen on the fundamentals of visual storytelling, one topic a week. Seven more distilling a longer guide I’d written about building short, emotional films. Good material. Read once, then buried in the archive like everything else on my blog/journal.

So instead of leaving them there, I turned them into free courses.

Filmmaking 101

Fifteen lessons: character entrances, point of view, stakes, reaction shots, establishing shots, the 180-degree rule, transitions, intention and obstacles, setups and payoffs, active characters, time cuts, dialogue, exposition, theme, and finally, audience. Free, self-paced, no signup.

Progress saves in your browser so you can pick it back up where you left off. Read it straight through, or jump to the one thing that’s bugging you about a scene right now.

Scribbler’s Playbook

Seven lessons built around one idea: everything in a short film or video should be pointing at the same emotional question. Opening moves, scene work, four structure recipes sorted by running time, camera and edit choices, and a final alignment check for when something feels broken and you can’t say why.

This one also ships with nine fill-in worksheets. Not printouts of questions to copy onto a legal pad, actual forms. Type into one, it saves as you go, print it when you want a copy on the desk. Seven come straight from the Playbook’s own checklists. Two are worksheets I’ve used on my own projects for years (character development, location development) that never had anywhere to live online until now.

Both are free at orbitalranch.studio/course.

You’ve Probably Got One of These in a Folder Somewhere

If you’ve been doing this a while, you’ve got a version of what I had. A folder of posts. A PDF you keep attaching to emails. Notes you paste into every consult because you’re tired of retyping them. That’s not content you haven’t published yet. That’s a course you haven’t assembled yet.

A few things that made the difference between a pile of posts and something people can actually learn from:

The order you wrote in isn’t the order someone should learn it in. My posts went up whenever a topic occurred to me. The course goes foundation, then grammar, then structure, then craft, then strategy, because that’s the order that actually builds understanding.

Reading isn’t doing. The Playbook lessons work fine on their own, but the worksheets are what make someone apply it to their own project instead of nodding along and closing the tab.

If your knowledge has a checklist or a diagnostic hiding in it somewhere, pull it out and let people fill it in.

Track progress or it’s just a long scroll. Fifteen lessons with no sense of where you are is homework. Fifteen lessons with a bar that fills up is a course you finish.

I’m not a developer, and I didn’t need to become one.

Describe what you want to Claude, steer by feel, let it handle the code while you handle the part you actually know.

If you’ve got a domain of expertise you’ve earned the hard way, screenwriting or anything else, the posts/docs/emails explaining it are probably already scribbled somewhere on your computer.

So why not go see what they add up to?

Always. Be. Scribbling.

A corkboard built to hold pushpins and index cards now stocks itself. A shy scholar coded gentle in a design doc threatens violence out of nowhere because a language model got confident about the wrong sentence.

Three separate companies are shipping software called some version of Storyflow, and not one appears to know about the other two.

Three true stories. Same year, screen side and game side both.

Everybody’s Startup Is Named Storyflow

The debate over whether scribblers should touch AI tools ended sometime before this year started. One June field guide on the space put a number on it: by its count, only about 4% of working scribblers use AI to produce a piece start to finish.

Everybody else treats it like a research assistant, a fast reader, or a first-pass machine, useful for one stage of the job and nothing more.

What’s replaced that debate is a pile-up. Search “AI screenwriting tools” and you surface three unrelated products fighting over the same name.

Storyflow.so: an AI visual workspace pitched at novelists and scribblers.

Storyflow-ai.com: script-to-storyboard-to-video, a different company entirely.

StoryFlow Editor: built by a small studio called Soulstices, sold on Steam for branching game dialogue.

None share ownership. All three shipped in the same rough window. You can bet a scribbler doing five minutes of research picks the wrong one at least once.

The Page Side: Sudowrite and the Board

For the actual script, the working pattern splits the job in two. One tool holds story-world context, the treatment, the bible, the character sheets, all at once, and drafts prose inside that frame.

A second tool handles the format nobody wants to fight about, the sluglines, the margins, the industry-standard PDF a studio expects on delivery.

Sudowrite’s own 2026 pitch to scribblers draws that line directly: draft where the AI has full context, format elsewhere where precision matters, then hand the pages to Final Draft or Highland. Final Draft’s Beat Board, the software’s index-card corkboard, has been the industry’s go-to outline tool since long before any of this started.

It didn’t need AI to earn that spot. It’s picking up AI features now anyway.

One flag before the number: Sudowrite’s cited 92% completion-speed bump comes from the company’s own user survey. Company research, published on the company’s blog. Real names stand behind the product anyway.

Emmy winner Bernie Su uses it. Silo author Hugh Howey calls it “scary good.” Take a company’s read on its own product for exactly what it is.

The Room Side: Inworld and the Dead King

Game narrative runs the same split, with the second tool doing something screenwriting software never has to: keep a character in character after the player stops watching.

Inworld AI is the name every roundup lands on. The funding backs up the hype, more than a hundred million raised, advisors that include a Disney Imagineering SVP and the guy who coined the word “metaverse.”

Its Character Engine already turns up as fan-built mods inside Skyrim and Grand Theft Auto V, plus native support inside Unity, Unreal, Roblox, and Minecraft. Ninety-nine percent of the gamers in Inworld’s own “Future of NPCs” survey said smarter NPCs would improve the games they play. Same flag as Sudowrite’s number: that’s Inworld’s own research, about Inworld’s own space.

LoreWeaver built its entire pitch around a harder problem. In the company’s own words: “Large language models are fluent, not faithful. They generate what sounds right, not what is true in your world.”

Their list of failure modes reads like a lore bible’s nightmares. A dead king who keeps talking three scenes after you killed him. A shy scholar who threatens violence out of nowhere. Gods and wars invented on the spot because the model needed something to say.

Guard consequence like it’s the show’s currency, or somebody else spends it for you. Heroes lost the thread when memory and death stopped meaning anything. LoreWeaver’s engineers are chasing the game-narrative version of that same fix, and they put it in five words: “the model proposes, your rules dispose.”

The Model Proposes, Your Rules Dispose

Every roundup this year skips the part that decides whether any of this works: the stack is only as good as the rules you hand the second tool.

This column runs on that exact split.

Raw dictation goes in one end, notes, tangents, half a rule scribbled between takes. A structured brief sits on the other end holding the guardrails, the facts that can’t be invented, the words that don’t get used, the lexicon that turns every “writer” into a scribbler before it hits the page.

The AI drafts inside that frame the same way Storyflow drafts inside a series bible, the same way a Character Engine drafts inside its own rules about who a shy scholar is allowed to be. Different room. Same architecture. A project brief and a Character Brain are both just a corkboard with better memory.

Pick Two, Every Time

Every roundup this year admits the same limit. A format tool can’t tell you whether the scene works. A Character Engine can’t tell you whether the scholar had a reason to snap. Only a scribbler sitting in the room, screen side or game side, makes that call.

Two tools, sometimes three, depending on the day and the platform. One set of rules that doesn’t move no matter how fluent the draft sounds.

Pick your two. Guard your rules.

My first real TV job put me in the Alias room, 2001. J.J. Abrams broke story out loud at the head of the table, fast, half his ideas contradicting the other half until the room collectively wrestled the good ones into a beat sheet.

I’d take that sheet down the hall and generate pages off it. He’d rewrite half of what came back into the episode that shot.

Nobody in that room had a name for what I was doing. I do now. I was functioning like a language model, for a showrunner running a job description nobody had scribbled down yet either.

Flip the chair. You’re the one breaking story out loud these days. The collaborator across the table doesn’t get tired, doesn’t have a bad day, never needs a note explained twice, and still can’t tell a good scene from a competent one without you.

Tulsea and Ormax Media put out the second edition of a survey back in February: The Right Draft: 2026, 254 working screen scribblers in India. Variety ran the numbers under an exclusive, and most of the coverage buried the number that matters under the top-line.

Forty-one percent already use AI somewhere in the process. Forty-seven percent don’t see it as a threat to the work itself. Sixty-eight percent think it’s quietly shrinking how much producers value what a human brings to a page in the first place.

Line those up and you’ve got the whole argument. Adoption’s already happened. Fear hasn’t caught up to it. And the number that should actually keep people up at night isn’t the top line. It’s how the job gets valued once the tool’s already in the building.

Assign It, Read It, Cut It

Every showrunner I’ve worked under did some version of what Abrams did in that first room. They can’t scribble every scene of a twenty-two episode order. Many can’t manage it on an eight-episode order either. So they walk in with a strong point of view, assign the work, read what comes back, decide what stays, what gets cut, and what needs more passes before it goes to the network.

That’s the exact job a screen scribbler does with an AI collaborator now. Nobody had to invent a new skill for this. Every scribbler in that position just had to notice they’d already been trained for it, usually by whoever ran the last room they sat in.

Four things carry that job: taste, structure, voice, judgment.

I’ve had twenty-five years to build all four before any model could string together a line of dialogue worth keeping. Each one holds up a little differently now that the collaborator across the table stopped being a person.

Competent, Polished, Forgettable

My dad spent close to forty years as an automotive journalist. Never designed a car in his life. Didn’t need to. Ten minutes into a drive he could tell you exactly where an engineer had cut a corner. That’s taste. It doesn’t require you to have built the thing. It requires having sat with ten thousand versions of the thing and kept score on which ones were true.

Ask a model for a scene in Hannibal’s register and it’ll hit every surface cue Bryan Fuller spent three seasons training audiences to expect: the operatic dialogue, the slow-motion dread, the food shot like a crime scene. It still won’t be Hannibal.

Taste is what tells you that in the first four lines, the difference between a scene that’s alive and one that’s a very fluent forgery of something you already watched.

The model doesn’t know the difference. It’s averaging across everything it’s ever read, and the average lands on competent, polished, and forgettable, which are the exact three words that get a script sent back for a pass in any room I’ve sat in.

What Season One Buried

A model can generate a scene. It can’t hold a season.

Lost ran three, four, five storylines at once across a nonlinear timeline, flashback, flash-forward, eventually flash-sideways, and the corkboard existed because nobody at that table could hold all of it in working memory either, let alone a system that forgets everything outside its own context window.

A reveal in season three only worked because of what got planted, quietly, in season one. That’s architecture, and it was always a showrunner’s job for exactly that reason.

Every other scribbler in the room gets assigned an episode, a sequence, a scene. The showrunner is the one required to know how all of it locks together.

That’s the gap a model leaves wide open right now, and precisely the skill that turns a pile of generated pages into a season instead of just a pile of paper.

Hannibal Doesn’t Sound Like Heroes

Hannibal sounds like Hannibal no matter whose name is on a given script, because Bryan Fuller set the register on day one and defended it for three seasons.

Heroes, a completely different animal, ran on convergence and forward motion instead of dread and patience. Put the same scribbler in both rooms in the same week and the sentences themselves would have to change shape.

A model can hold a character’s speech pattern if you feed it enough of that character’s dialogue.

It cannot decide what a show’s voice is.

That’s still the showrunner’s call, same as it was when the newest voice in the room belonged to a staff scribbler three weeks into the job and hungry to please. A model has no opinion about anything until somebody hands it one.

The Failure Wearing Different Clothes

This is the skill that actually separates a showrunner from a staff scribbler, and it’s the one that gets exposed fastest working with AI.

Judgment means trusting the material and still checking it, every time, against pattern recognition you’ve spent decades earning.

It means asking what happens three scenes downstream if you take the note the model just handed you, instead of taking it because it sounded smart in isolation.

It means rarely getting surprised, because you’ve watched this exact failure mode before, just wearing different clothes.

I’ve got ADD, diagnosed late. My first instinct on a page is sometimes wrong, so I had to build the habit of sitting with a read before I trust it, checking my own gut the same way I’d check anybody else’s draft.

Turns out that exact habit is the whole job now.

A showrunner who stops reading every page, who mistakes silence for quality because nobody’s pushing back anymore, ends up running a show that’s already flatlined without knowing it. I’ve watched that collapse happen more than once. It always starts the same way: somebody decided they didn’t need to check the work anymore.

Direct the Model Like Staff

None of this works as an abstraction. The mechanical version looks like this.

Treat the model like a specialized scribbler in your room.

Prime it with your actual expertise instead of a generic prompt.

Ask it a directed question, the way you’d hand a staff scribbler a specific assignment. Skip the vague, undirected prompt.

Build every follow-up on what it already gave you instead of starting cold each time.

Run all of it through judgment you already own, the same filter you’d run a green scribbler’s first draft through before it ever reached the showrunner above you.

The Number Every Recap Skipped

The part of the Tulsea-Ormax numbers that should worry working scribblers more than the adoption figure ever will: seventy-four percent now feel unfairly compensated, up from sixty-three percent in the 2023 edition of the same survey.

Fifty percent report producers expecting faster turnaround now, on the assumption the AI already did some of the work.

Did it? Doesn’t matter to the schedule. Line that up next to the 41% adoption number and you’re not looking at separate facts.

You’re looking at a negotiation producers are already having with themselves, on a timeline nobody invited scribblers to sit in on.

Shailesh Kapoor, who runs Ormax, put it plainly: writers “sit at the core of the storytelling ecosystem, yet too many friction points remain structural rather than episodic.”

He’s right, and structural is the part that should worry you, because structural doesn’t fix itself once the trade press moves on to next week’s headline.

I’ve carried a WGA card since 2001, and the Guild’s contract already settled the credit question: a model doesn’t get called a writer, and what it generates doesn’t count as source material.

That protects something real. But it does nothing about a producer who read this same survey and figured out AI is a good enough excuse to ask for a first draft in half the time, for the same money, from somebody too worried about the next job to say no.

Protecting what your judgment is worth was always part of running a room too. Showrunners negotiate. They hold a line when a note isn’t worth taking and let one go when it is.

That instinct doesn’t retire because the collaborator handing you a first draft is a model instead of a person. It matters more now, because you’ll spend more of your career proving your judgment did the work, to people who’d rather believe the tool did it alone.

Nobody Handed You a New Job

The trade press keeps asking if AI is coming for the scribbler’s job. Wrong fight.

The part of the job that was ever just words on a page was always the smallest, most replaceable part of it, and every showrunner who ever rewrote your draft on a Tuesday night already knew that before any of this technology existed.

Taste, structure, voice, judgment, applied under deadline, to material that wasn’t always yours to begin with. You’ve been training for that since your first day as a scribbler in a room somebody else was running.

Phil Tippett got called extinct by his own movie in 1993, then spent the rest of the shoot proving it wrong. A three-time Oscar winner just ran the identical play on Dune: Part Two, thirty-one years later. It’s a clear read on how this business survives contact with a render farm.

“It looks like we’re out of a job.” “Don’t you mean extinct?”

That’s Alan Grant and Ian Malcolm in Jurassic Park, and the second line wasn’t invented for a laugh.

Phil Tippett said almost exactly that for real, standing in a screening room in 1993, watching Industrial Light & Magic’s first photoreal CG dinosaur test: “I think I’m extinct.” Tippett was the film’s credited Dinosaur Supervisor, one of the best stop-motion and go-motion animators alive, hired to build every dinosaur in that movie one frame at a time, by hand. Spielberg showed him what the computer artists had built instead, and Tippett’s own reaction ended up written into the finished film almost word for word.

I grew up on the man Tippett learned from. Ray Harryhausen, skeleton warriors clashing on a beach in Jason and the Argonauts, animated one frame at a time on a rig in his garage, wired straight into whatever part of my brain later decided telling stories for a living was a reasonable plan.

Watching a render farm generate in a weekend what Tippett’s career had been built on should have ended him right there in that screening room.

It didn’t.

Spielberg kept him on the film and put him in charge of the one thing the computer artists still couldn’t do: making the dinosaurs move and behave like animals instead of monsters. Weight. Gait. The way a predator actually watches a jeep instead of just menacing it for the camera.

The physical animating passed to other hands. Tippett kept the eye, the part that knew what a believable predator looks like. That eye turned out to be the whole job, and Tippett spent the rest of the shoot teaching a T-Rex how to hunt.

The Squeeze Has a Number Now

I’ve felt a version of that same squeeze the last two years, minus the render farm. Gigs that would have landed easily two years ago went to a smaller room, sometimes no room at all, and the free time that opened up did something I didn’t see coming.

I started building. Random games. Worlds. Assets, stories, merch, whatever a given afternoon wanted to become. Somewhere in that pile I started suspecting my own ikigai is sitting inside a transmedia world I build myself, which meant actually learning what these generative tools do, past the demo reel.

The GDC crowd already knows the number: 63% of game design and narrative professionals hold an unfavorable view of generative AI, second-highest of any discipline in the business behind visual and technical art, ahead of programming.

GDC’s 2026 State of the Game Industry report ran that stat next to a few others that explain the mood: a third of US game industry professionals laid off in the last two years, two out of three people at AAA studios watching their own company do the cutting, 74% of students entering the field worried there won’t be a job waiting when they finish the degree.

Fifty-two percent of the industry now calls generative AI’s impact negative, up from thirty percent the year before and eighteen the year before that.

Wrong Link

Type a sentence into Sora, Veo, Runway, or Kling and thirty seconds later you’re holding a moving image that didn’t exist an hour ago. That’s the public argument: prompt in, movie out, actors and crew and continuity all optional. Great demo reel.

It has almost nothing to do with how a real pipeline gets built in 2026.

The actual chain runs seven links: script, design, performance, generative asset creation, compositing, editorial, conventional finishing.

Generative tools own exactly one of those. Sora and Veo and Runway and Kling, the same names that just generated a full scene from one sentence, do real work at that link too, but they get pointed at a single plate, a single extension, a single background element the production already storyboarded.

The script, the production design, the performance an actor delivers on a stage or in a mocap volume: people in a room, making choices, author every bit of it.

Compositing, editorial, the finish a studio signs off on: all of it still runs through the same craft it always has.

The Blue Eyes, Thirty-One Years Later

Paul Lambert ran Tippett’s exact play on Dune: Part Two, just with better tools and three Oscars already on his shelf, one of them for the first Dune.

The flaw he needed to fix was a color: the right shade of blue in every Fremen’s eyes across Arrakis. His team hand-rotoscoped 260 shots from the first film into a training set of 30,000 pairs of eyes, built a machine learning model out of it, then pointed the model at more than a thousand new shots of Zendaya’s Chani and every other Fremen in the sequel.

It didn’t just work and walk away, either. Eye size and shape change shot to shot, so plenty of frames still needed a compositor’s hand, roto’d the old way by the artists at WylieCo, then fed back into the model to make the next pass smarter.

Nobody generated Chani’s performance. Nobody generated the blocking, the sandworm chase around her, or a single choice Zendaya made in that scene.

Lambert built one tool, aimed it at a single correctable, deeply boring detail, and kept everything else exactly where it already was: in the hands of the same artists who’d have painted those eyes by hand a decade earlier.

AI Production Design

Name the skill and it stops sounding mysterious: AI production design. Same two words the art department has used for a century.

Knowing which pieces of an authored scene can get generated without losing character, performance, or continuity. Knowing, just as precisely, which pieces never leave a human’s hands.

Tippett protected the walk. Lambert protected the eyes. A screen-scribbler protects the scene itself, on the page, before a director, a compositor, or a generative tool ever gets a turn.

Write a character’s choices sharp enough on the page and everyone downstream knows what can’t be touched without the scene stopping being that character’s scene. That’s the job opening up in this business right now, and almost nobody’s training for it.

The Sixty-Second Test

Most people testing this stuff try one shot. That only proves a tool can generate a pretty picture. It proves nothing about a workflow’s ability to hold a production together, because a single shot has no character to keep consistent, no location to match across a cut, no prop that has to be the same prop twice.

Build sixty to ninety seconds instead. One recurring character. One location. Props that show up more than once. Cuts that must hold editorial continuity the way an actual scene does. Run a workflow through that, and every weak seam will show up.

Forget whether the tool can make one shot look expensive. Ask if the character still looks like the character, the room still looks like the room, the cut still lands like a cut, sixty seconds in.

New Tools, Old Instincts

Tippett didn’t stop working after that screening room in 1993. Lambert’s name is still in Dune’s credits thirty-one years later for the identical reason: he knew exactly which detail belonged to the machine, and which one never left his artists’ hands.

I think about that split constantly now, building worlds out of whatever a given afternoon hands me, deciding by gut which piece I generate and which piece I sit down and actually scribble myself, frame by frame, like Tippett with that T-Rex.

Learn that same read on your own scene, and the credit still has your name on it.

Three facts from this year. A seventy-year-old billionaire working out how much of the world’s work should stay in human hands by arguing it through with a chatbot.

A tax code that bills a company for hiring a person and hands it a deduction for buying the machine that replaces them.

A guild contract ratified in April whose headline fight was health insurance.

One fence runs underneath all three.

A Nature Reserve for Work

Gates published six thousand words on August 26 called “The turbulent AI era is here. The choices we make now are critical.”

Three ideas of note.

1. New national coordinating bodies plus an international one assembled out of pieces of nuclear weapons inspections, aviation rules, and the ozone treaties.

2. A tax on AI tokens and robots.

3. A category of work society deliberately keeps for people after machines can do it. He calls that third one Human Reserved.

The analogy is a nature reserve. Roads and buildings could go there and don’t, because what’s standing is worth more. His example is a robot delivering news you have an incurable disease.

Gates writes that there’s “no technical reason why it couldn’t. Yet it shouldn’t.” The idea came out of watching the paid caregivers who looked after his father through Alzheimer’s, work he puts permanently outside a machine’s business.

Gates doesn’t pretend to have it solved. Who decides what gets reserved, by what criteria, how you stop a company from quietly using robots anyway, and what happens to trade between a country that allows it and one that doesn’t, all of that sits in the essay as open questions to be worked out in public.

He told GeekWire he’s been running the math through Claude, testing shorter workdays and earlier retirement as ways to push the share of work reserved for people up toward 40% and spread what’s left around. GeekWire

That reads as abstract policy right up until you run it as a story problem. Then it’s the most useful thing anybody has said about this business all year.

Everything Before the Contract

The Writers Guild drew a line in 2023 and members renewed it this spring. A model doesn’t get to write or rewrite literary material. What a model produces doesn’t count as source material, so nobody can use it to erode a scribbler’s credit or separated rights.

A company can’t force a scribbler to use AI software, and it has to disclose when material handed over came out of a model.

The 2026 MBA, ratified April 24 with 90.38% in favor and running through May 1, 2030, preserves every one of those protections and adds one: written notice to the Guild when a company licenses scribblers’ work to train a commercial generative system, with the Guild able to request a conversation about the license and about paying scribblers for it. Writers Guild of America EastWriters Guild of America

Every one of those provisions protects credit, compensation, and consent on covered projects.

None of them says a word about who conceives the premise, who breaks the season, who decides what a character wants badly enough to lie for it. Outside the covered zone the fence doesn’t exist at all. Development. Unrepped features. Games. Everything a screen scribbler does before a contract even shows up.

The Guild bought ground it could hold in a year when the health fund was on fire, and members paid for it. Active participants carried no monthly premium at all, and a flat $50 for any number of dependents, frozen for twenty-three years. That changes in 2027, along with higher deductibles and out-of-pocket maximums.

Somebody still needs to draw the rest of the map, and it starts with naming what a machine can produce and what it shouldn’t. Story conception. Character arc. Voice.

Ability is not permission, and a business that can’t name three things it refuses to automate has already agreed to automate all of them by default. Hmmm.

If You Don't, a Start-Up Will

Gates names the mechanism and calls it a vicious cycle. It runs in three moves. Somebody automates, turns the savings into a lower price, and every competitor feels immense pressure to answer. Sitting it out solves nothing, because if the existing companies don’t move, start-ups will. New American Journal

Swap in studios and the sentence needs no rewrite.

One financier produces cheap coverage, cheap concept art, cheap previz, cheap first drafts, and the margin pressure travels through the entire market inside a quarter.

Volume over quality, on a loop, with nobody able to step off first.

Gates thinks the lever is tax, since a company that hires a worker pays payroll taxes while a company that buys a robot writes it off right away.

He first floated a robot tax nine years ago and mostly got told it was a strange idea. He’s still for it. He’ll grant the tax costs the economy something, then argue we can carry that cost as the price of keeping people working. GeekWire

This industry already has that lever built and running.

California’s Film and Television Tax Credit Program 4.0 puts $750 million a year on the table through June 2030, and applications get ranked on a jobs ratio: qualified wages plus 35% of other qualified expenditures, divided by the estimated credit.

Payroll’s a scoring metric. Geography is what the state is buying.

Maureen Kerr laid out the problem in Forbes on August 19. The AI-native outfits she profiled last year, Promise, Phantom X, Gossip Goblin, run on small permanent teams, contracted specialists, and more cash in compute and rights than crew payroll.

That doesn’t make them ineligible. It makes the program irrelevant to the economics that define them. Forbes + 3

The jobs ratio is a human-reserved policy. Nobody’s calling it one yet. Somebody should, before the next revision quietly stops caring who did the work.

Six Years Before Anybody Notices

The line that should scare this town sits nowhere near the tax argument.

Gates writes that the jobs at most risk are entry and mid-level, that the new jobs coming will mostly require skills that take many years to learn, and that he’s especially worried about young people entering a workforce with fewer entry-level openings.

Inser apprentice jobs here. Scribblers’ assistants. Script coordinators. Staff scribblers. Junior story editors. Now name what those jobs consist of: coverage, research, transcription, tracking continuity, a polish pass nobody wants to do twice.

That’s the exact task list a model eats first and the cheapest.

Kill the first rung and the ladder stands there looking fine for about six years. Then the industry goes looking for its next generation of showrunners and finds a cohort that never sat in a room while two executive producers argued for ninety minutes about whether a character would really make that phone call.

Nobody learns that from a wiki.

Showrunners and EPs are the ones who protect those entry level seats or don’t. That decision gets made a hundred times a year, quietly, in budget meetings, by people who could go either way.

What Stays Scarce When the Picture Gets Cheap

A budget meeting grades cost. The market pays whoever cuts fastest. Which is why the counter-cyclical bet is worth making.

Kerr has been putting the same question to every AI filmmaker she profiles: what stays scarce once making the picture gets cheap.

The answers keep coming back ownership, voice, audience trust, control of the release. Not geography, and not speed. Forbes

Synthetic content will saturate every low-friction feed on earth over the next few years, because the cost of filling a feed is verging toward zero.

Everything with a person behind it gets scarcer at exactly the same rate. Hyper-specific, messy, first-person work no model can assemble from training data. Stuff somebody scribbled with intent.

What’s the work you’d refuse to hand off even if refusing cost you the gig? And if you’re the one signing a budget, when’s the last time you fought to keep a baby scribbler’s seat in your room?

ABS. Always. Be. Scribbling.

Runway’s terms of use warn, in capital letters, that outputs may not be unique and similar outputs may be generated for other users.

Luma assigns you the rights to your shot on the condition that you generated it while holding a paid plan permitting commercial use.

Generate on the free tier and the watermark stays welded to that asset forever, whatever you upgrade to later.

Higgsfield rewrote its own terms inside of three days in July, after a thread breaking them down cleared 331,000 views on X. Everybody who signed up before that rewrite has been bound by the new version since August 7.

Three video companies, one summer, and not one of those sentences has anything to do with how the footage looks.

I Haven’t Generated a Frame Yet

Straight up front, so nobody mistakes this for a review. Everything below is my own research, done while I work out which of these tools I’m actually going to pay for in time and money.

The reason I’m doing the research is a project called Warranty Void. Five operators running freelance work through Lost Angeles in 2066, inside the Terra Conflictus world I’ve been building. The whole thing is scribbled to exist as found footage. In-world video posted by the crew themselves, operator logs, recovered files, fixed cameras nobody was supposed to be looking at.

That format comes with a technical demand welded into it. The same five faces have to come back artifact after artifact. The same streets. The same gear on the same shoulders. Recurring characters in a recurring city collapse the second shot two looks like a different person than shot one, and that failure mode is the one every honest review of these tools keeps circling.

I’ve scribbled about this stuff already. Phil Tippett walking into Jurassic Park as Dinosaur Supervisor and walking out having taught a T-Rex how to hunt. Paul Lambert taking 280 hand-rotoscoped Fremen shots out of the first Dune, cropping and augmenting them into a training set of 30,000 eyes, then pointing that model at a thousand more shots in the sequel.

I made the case that the job opening up right now is AI production design, knowing exactly which portion of an authored scene can be generated and which portion never leaves a human’s hands.

Making that case and running the pipeline are different jobs. Warranty Void needs sixty seconds of moving picture standing on its own before anybody’s going to care about page one.

So this is a scribbler at the starting line, having my AI minions read contracts and comparison charts, trying to figure out what a stack is supposed to look like before diving in.

Nobody Buys a Camera Package

Two decades of production teaches you one useful reflex here. You build a department. Every piece of gear on that truck earns its slot by doing one job nobody else on the truck can do.

The AI video market keeps getting covered like a horse race. Which model wins. Which one just leapfrogged the other. That framing sells clicks and tells a working scribbler almost nothing, because these tools are not interchangeable and the differences that decide a project sit in camera control, character consistency, editability, and license terms.

Three roles, already visible in how people actually work.

The board room. Image-led previs. Lookbooks, character explorations, scene boards, the visual bible that keeps every downstream shot pointed the same direction. Firefly and an Adobe-adjacent image workflow fit here.

The primary workspace. One place where you iterate on hero shots, keep your references, version your variants, and assemble something. Runway carries the strongest reputation for this because it operates as a production-facing environment with reference-driven character consistency built to hold a face across a whole sequence.

The specialists. A second and third engine you point at the shots your primary tool handles worst. Luma when a shot lives or dies on believable motion through real space. Kling when a strong source still is driving the image. Higgsfield when the whole idea of the shot is a camera move.

Thirty-Five Millimeter Handheld, Push Through the Arcade

Higgsfield is one I keep circling, and the reason is vocabulary.

Its Cinema Studio ships north of seventy named camera presets. Dolly in, crane up, 360 orbit, crash zoom, Dutch angle, Snorricam, FPV drone. You pick the move off a menu the way you’d call it on a set, and the platform routes the job to whichever engine it thinks fits, Kling or Sora 2 or Veo 3.1 or Seedance.

Reviewers keep landing on the same two-part read: best-in-class directorial camera control, and a credit system that empties faster than anybody plans for, with premium models running forty to seventy credits a clip and credits expiring in ninety days.

Here’s what that means for anybody who came up on a set or in a scribblers’ room. You already have the language. Thirty-five millimeter handheld, push through a crowded arcade, sharp rack focus to the protagonist. That sentence is legible to a controllable camera system in a way it never was to a plain prompt box. Coverage, lenses, blocking, screen direction, motivation. Every one of those is a control surface now.

The unfair advantage a screen scribbler brings into this market is knowing what to ask the camera for.

Three Questions the Demo Reel Won’t Answer

Every platform’s marketing page runs similar sizzle. The terms of service is where the differences live, and it comes down to three questions worth asking before you upload anything you care about.

Who owns the output. All three of these say they don’t claim ownership of your inputs or your outputs. Luma goes further and assigns you its own right, title and interest in the output. Fine as far as it goes, and it doesn’t go as far as it sounds.

Can you sell it. Generally yes on a paid plan. Luma ties commercial use directly to the subscription that was active on the day the asset was generated, and those rights vest permanently in that specific asset. Free and Lite generate watermarked, personal-use-only material, permanently, and hand Luma a much broader license back on top of it.

Can somebody else make the same thing. No platform will promise you otherwise. Runway says so in its own all-caps: outputs may not be unique and similar outputs may be generated for other users.

Then the one that should stop a working professional cold. None of those answers touch clearance. Platform terms cannot hand you rights in a recognizable actor’s face, a logo in the corner of your frame, a needle drop, a copyrighted character, or a reference image you pulled off somebody’s portfolio.

Ownership of output and clearance of what’s inside it are separate problems, and only one of them shows up in the terms of service.

Perpetual Is Already in Everybody’s Terms

The word that torched Higgsfield in July was perpetual. Go read Runway’s terms of use. Section four grants Runway a license over any inputs and outputs generated through the service, described as non-exclusive, irrevocable, perpetual, worldwide, royalty-free, fully paid, transferable and sublicensable.

That’s the same string of adjectives, in the same order, sitting in the terms of the tool with the best reputation in this market.

Luma’s grant runs a shorter list aimed at a narrower target. Worldwide, non-exclusive, irrevocable, perpetual, royalty-free, sublicensable through multiple tiers, covering the processing, hosting and storage of output produced during an active subscription, for the purpose of improving the service and training the models. Every tier below Enterprise is inside that.

Nobody made a thread about either one. Higgsfield got dragged because a company sent a terms-update email at the exact moment enough people were paying attention to read it. The clause itself is standard practice across this entire category.

You are not choosing between a company that takes a license and companies that don't. All three take one. You are choosing how wide it runs, how long it survives you, and whether anybody happened to be reading the day it changed.

Three Days in July

Higgsfield ran the lesson in public, on a clock anybody can check.

The July 23 update handed the company a license over user inputs and outputs described as perpetual, irrevocable, transferable and sublicensable, covering uploaded faces used for character generation.

A thread from Theoretically Media breaking it down went wide and cleared 331,000 views. By July 26 a revised version was up with the perpetual and irrevocable language deleted, rights now ending when a user deletes their content or account, and promotional use limited to material a user made public or consented to. New users were bound immediately. Existing users were bound as of August 7.

The training grant survived the rewrite word for word. Content still trains the models by default, the only consumer opt-out is deletion, and enterprise gets carved out. Model weights that already learned from your material don’t unlearn it.

Four provisions got quietly tightened in the company’s favor on the way through, including a fifteen-day notice guarantee downgraded to reasonable notice.

The contract you agreed to in the spring is not the contract you’re working under in August.

Terms move, sometimes in three days, sometimes in your favor because a few thousand people made noise, sometimes the other direction while everybody’s watching the demo reels.

Check the effective date on the page you’re actually bound by.

The Boring Folder That Gets Your Movie Distributed

Chain of title is the unglamorous folder that decides whether anybody can distribute your picture.

Same discipline applies here, and building it from day one costs nothing. Log the exact terms version in force when you generated, the plan tier and the receipt, generation dates and prompts, licenses for every uploaded source, raw outputs sitting alongside the human edit files, script drafts, boards, sound licenses, and a shot log naming your creative choices on every single shot.

That packet is worth more than any claim about which model made what, because human authorship is the part you can actually prove.

A scribbler who can produce a documented trail from script draft to boards to shot log to final cut is holding something. A scribbler who can only say the AI made it is holding a file.

Sixty Seconds, Six Shots, Every Engine

I’ve argued before that a single shot is the wrong unit for testing this stuff. One shot proves a tool can make a pretty picture. It has no character to keep consistent, no location to match across a cut, no prop that has to be the same prop twice.

So the plan is sixty to ninety seconds, run through identical inputs on every engine I’m considering.

One locked character reference, because a crew of five has to survive the cut.

Then six shots: interior dialogue setup, full-body walk and turn, an establishing shot, a complex camera move, an action transition, and a repeat of the first shot after everything else is done.

Then score the boring stuff. Continuity across cuts, usable duration, artifact rate, cost per keeper. And whether I can reproduce the same look two weeks later when the model has quietly updated underneath me.

Reproducibility separates a production tool from a slot machine.

I did time in the Lost room. Years in the Heroes room after that. Both shows had tentpoles: big, fixed points the whole season got built around, with a clear sense of where the year ended. A lot of open road stayed between those points, left there on purpose.

That gap is where a story gets room to surprise you. It surprised us plenty of times, sitting around that table.

Both shows became hits. The industry took the wrong “Why? lesson.

The Big Nope

Once destination-driven serialized TV proved it could hold an audience for years, networks wanted the route and road map on paper before they’d greenlight anything.

Every beat scheduled. Every reveal, every character turn, set in stone season by season before a single scene got scribbled.

The shows that came out of that mandate often felt rote, uncompelling, marking time to hit a plot point the room had promised in a document nobody in the audience would ever read.

A bible that specifies every turn doesn’t protect your story. It embalms it.

Your scribblers will stop discovering the season and start delivering a memo. Audiences can tell the difference between a turn a character earned and one a spreadsheet demanded on schedule.

Let The Destination Do The Work

A destination is one commitment. A fully mapped route on how to reach it is another. Mixing them up is a critical failure.

Knowing where a story ends tells everyone in the scribble room that the questions raised early will matter later.

Drawing what’s on every bit of road before leaving the garage means nobody discovers anything once the greenlight flashes.

Jeff Pinkner did a stretch as a scribbler and executive producer on Lost before taking over, years later, as showrunner on From, MGM+’s slow burn horror series about a town nobody can leave.

Three seasons in, that show was piling up unanswered questions in exactly the way that had sunk shows like it. Then, this spring, MGM+ committed to a fifth season and called it the final one.

That’s your destination. Every question From raises now serves an ending the room has already fixed.

The scribblers getting finales right this year keep reaching the same shape: destination first, road map unrolling along the way.

Vince Gilligan put it plainly this past fall, talking about Pluribus. “I feel like we have kind of a general roadmap for where the show should go,” he said, then: “I always will quickly get rid of a good idea for a better idea. If better ideas come along, I’ll happily discard what I’m thinking of right now.”

What Buyers Want Now

The current conversation among scribblers, the one showing up in pitch-prep panels and scribblers’ groups this year, says streaming buyers want the bible earlier than ever. Sometimes before they’ll take a first meeting at all. Worth taking seriously either way. The specific complaint about timing might not hold up in every room, but the mood behind it checks out.

Ben Silverman, who runs Propagate and used to co-chair NBC Entertainment, said it plainly last year: “There’s just so much fear among so many executives about taking risk and doing anything that just doesn’t have a massive star or a massive IP.” A buyer that scared of a swing wants documentation that feels like proof.

A finished bible reads like proof. Doesn’t matter if it’s the kind of evidence that will actually produce a good season of televsion.

Set Your Destination First

A bible built for this climate doesn’t need every psalm in stone. It needs the goal shouted loud enough that a buyer can hear it.

State the ending. One paragraph, full confidence, no hedging language that reads like doubt to somebody scared of a greenlight.

Name the engine, the thing that generates weekly story, the reason episode nine has somewhere to go that episode one didn’t spoil.

Mark three or four tentpoles the whole season actually depends on.

I’ve had a version of this conversation on the game side too, doing narrative design work for live-service titles. A season roadmap for a game runs on the same logic. Nobody needs the whole patch schedule the day the game ships. They need to know the story has a baseline under it and a reason to log back in next month.

Leave everything between those tentpoles loose. Alive. Argued over. Undecided until the season needs it decided. A buyer wants to know you’re not lost. They don’t need to see the whole map. They just need to believe that you know where you’re going.

The Map Still Gets Drawn in the Room

Lost never needed the whole route drawn before the pilot shot. It needed an island, a hatch, an empty wheelchair, and eventually, once the network finally let the room name an end date, a destination everyone could drive toward with purpose.

Heroes needed New York, an explosion, and a handful of ordinary people converging on it. Everything else got discovered on-route, which is the only reason either show is still worth arguing about.

NBC hung “Save the Cheerleader, Save the World” on bus benches before Heroes aired a single hour. Strangers who had never watched a frame could recite the premise back at you at a bus stop. That is a genuine marketing win.

A tagline sells what an audience wants. The hour must deliver what they need. Marketing bought the appointment. The room earned every following minute. Most rooms have never landed a line that clean.

On August 16th, Dario Amodei did the corporate version of that admission on his own account. He runs Anthropic, one of the largest AI labs on the planet, and he posted the sentence his industry has spent years avoiding out loud:

“At this point, saying that AI will cure cancer is more a cliché than it is inspiring, and most people think it is deceptive. The thing that will work is actually curing cancer.”

Show the proof. Not the demo, not the keynote, not the trailer cut to make investors lean forward in their chairs. Fair enough.

Then Steve Hilton read fourteen thousand words of Amodei’s and ran a different test. He counted the word happiness.

Zero.

“It’s striking that the word ‘happiness’ does not appear once in your entire 14,000 word essay, the one thing that, in my experience, parents most want for their children. The closest you get to it is ‘well-being’ and a ‘more fulfilling day-to-day experience,’ which rather proves my point.”

Every scribbler alive has taken that note. The word your story is supposedly about goes missing from the pages, and the absence tells the room what you actually understand.

Amodei can cure the cancer and still fail the scene, because curing the cancer answers the want. Hilton asks about the need.

The Third Act Nobody Roots For

Amodei’s own essay concedes that “most or all humans may not be able to contribute meaningfully to a sufficiently advanced AI-driven economy.” His proposed fix runs from “a large universal basic income for everyone,” which he grants is likely “only a small part of a solution,” to this:

“A capitalist economy of AI systems, which then give out resources … to humans based on some secondary economy of what the AI systems think makes sense to reward in humans.”

Read that as a page. An outside force arrives and hands the protagonist the thing they were fighting for.

We have a name for this. Aristotle flagged it in the Poetics and the note has held for centuries: the resolution must come out of the plot itself. A machine that lowers down and dispenses what it has decided humans deserve is the crane, the god, the wire.

Turn that ending in and it comes back with the same note it has always come back with.

Hilton, whose parents fled Hungary after the 1956 Soviet invasion, read the same passage and called it what it looked like to him. Central planning. Different vocabulary. Same structural complaint. Somebody offstage is deciding the outcome, and the people in the scene have nothing left to do.

His counter runs straight through the middle of every story you’ve ever loved:

“In each case people still had to provide for themselves and their families, and it is exactly that caring and providing that gives life meaning for so many of us.”

Not a political position. More like a dramatic construction. A character who provides for someone has stakes. A character who is provided for has a montage.

Contention Is the Product

Cure a character’s cancer in act two and you’ve solved a problem. You haven’t built a scene. Health, money, safety, and no reason to contend gets you a sequence audiences check their phones during.

Amodei’s answer to the hollowness is chemical. He points at caffeine, modafinil, Ritalin, and suggests “much more is likely to be possible.” Hilton clocks the reference immediately and names it: Brave New World, minus even the centrifugal bumble-puppy.

Huxley wrote the definitive version of that world in 1932 and nobody has ever pitched it as the happy ending. The reason is craft. Take away the fight and you take away the reason to watch.

Nobody gets converted by a screenplay a machine generates in thirty seconds. Speed stopped impressing this town a while back.

True conversion means making a stranger feel something they never agreed in advance to feel, and those emotions are built out of watching a person contend with something hard and pay for it.

The Four Tests, Run Fast

Does anything bleed? AI builds an outline, fakes a genre’s voice, hits the rhythm. Adequate is the most dangerous output a machine produces, because it’s CosPlay scribbling and skips everything that made the costume necessary. A scribbler clocks this one the page. The outfit is right. But nothing underneath is bleeding.

Would Nicholson fight for the line? Actors fight over pressure, subtext, an unexpected reversal. We’ve had actors call into a room wanting one more pass at a scene because the version on the page hadn’t been earned. Sidney Lumet shot 12 Angry Men in one room with no effects and it stands as the reference point for pressure-cooker drama sixty-nine years later. Aaron Sorkin built a courtroom monologue into A Few Good Men that acting classes still cold-read thirty-four years on. Jack Nicholson never raises his voicet, until “you can’t handle the truth” lands like a verdict.

Can it shut a room up? Rutger Hauer trimmed his own death speech in Blade Runner down to four lines on the day they shot it. The film underperformed that opening weekend. The speech outlived the weekend by four decades and counting. An actor earned that silence on set, in real time, cutting his own scripted lines to get there.

Does it sound like Elmore Leonard? Blend ten famous scribbling voices into one manuscript and the result reads like nobody. Tear the cover off a Leonard book and a reader clocks his rhythm inside a page, then comes back for the next one because of it. A sentence carries a fingerprint. Nobody is prompting their LLMs to write more like AI.

Four tests, one question underneath all of them. Did a person contend with something on this page, and did it cost them.

The Pitch Nobody’s Made

Hilton ends where a scribbler would start. “There IS an inspiring story to tell about AI.” He’s right.

Here’s the pitch, from someone who has scribbled for their supper.

Put the machine in the scene as pressure. Let it raise the floor, take the drudgery, hand back hours, and leave the fight fully intact.

The protagonist still needs to provide. Must still choose, and pay, and carry.

That story’s been told about the plow, printing press, and camera, and every version of it works, because the human stayed in the process doing something with consequences.

The version Amodei sketched has the machine deciding what humans deserve. Fourteen thousand words and the word happiness never appears once. I think this omission helps us understand the author more than they understand themselves.

Whatever your story claims to be about, search the doc for that word. If it doesn’t appear, and no scene puts a character in a position to earn it… What the hell are you even scribbling about?

A superintelligent ape stands convicted of killing his father, the King of Gorilla City. A Daily Planet reporter reopens the case. Eight episodes, true crime grammar, in mockumentary format.

That’s The People v. Gorilla Grodd, ordered straight to series at HBO Max from Tony Yacenda and Dan Perrault, the American Vandal team, with Yacenda directing every episode and James Gunn and Peter Safran producing. Skyler Gisondo plays Jimmy Olsen. Jimmy Tatro plays Grodd. Wendell Pierce plays Perry White, with Beck Bennett as Steve Lombard and Mikaela Hoover as Cat Grant.

Y & P stated their goal in the announcement: it should feel “less like an entry into the DCU and more like an artifact from it.”

On paper, it’s a great pitch. I’m sure it’ll be fun and look forward to watching. But I’m a man of a certain age. A DC fan from my days wearing footie pajamas. But I can’t help wondering —

— Why are we disassembling DC icons before this generation has seen them assembled?

Deconstruction Is a Second Draft

Frank Miller published The Dark Knight Returns in 1986. It landed like a bomb because readers had spent forty-seven years with another Batman. Alan Moore had the same advantage with Watchmen. Both books spent a debt the culture had already paid, and the payment was decades of earnest, unironic, cape-and-cowl belief.

Marvel ran the identical math. Guardians of the Galaxy arrived in 2014, six years and nine pictures after audiences watched Steve Rogers throw himself on a grenade and mean it. Gunn got to make the joke because somebody else had already made the promise.

Subversion is a response to something already established.

Irony Fatigue

People keep blaming superhero fatigue for the softening box office. The diagnosis is off by one word. Audiences still want the heroes. It’s the winking that’s wearing them out.

When every project elbows the audience in the ribs, dissects its own tropes, and runs world-saving heroics through a mockumentary filter, the message reaching the not-cheap seats is that the people making this stuff find the material a little embarrassing.

Marvel spent years on forced meta humor and taught a generation that sincerity is the setup and the joke is the payoff.

All that irony is cheap armor. Nobody can call your hero corny if you called him corny first.

Gunn Already Built the Straight Man

Gunn’s Superman was sincere. Big blue boy scout, cape, dog, hope. It made $618.7 million worldwide. Gunn and Safran said the Grodd series exists because of “Skyler’s wonderful performance and the fantastic chemistry of the Daily Planet cast,” which is a straight line from earnestness to spinoff.

So they did lead with the myth. Then they immediately started turning into the dark alley behind it.

Second Swing is a Miss

Supergirl opened June 26, 2026, on a budget reported between $170 and $186 million. Craig Gillespie directed, Milly Alcock starred, and the pitch was a road movie in outer space with a darker tone than Superman. Gillespie named True Grit, Logan and John Wick as his touchstones, a Venn diagram with no cape in the overlap.

It opened to $38 million domestic, $68 million worldwide, and finished at $125.8 million worldwide. Superman earned that entire run in its opening weekend. Critics landed at 54 percent on Rotten Tomatoes across 303 reviews. Audiences gave it a B minus CinemaScore, and a B minus on a superhero picture is the sound of a room walking out unsure what it just watched.

Alcock got good notices. I thought she was terrific. Her performance kept turning up in the positive reviews. The footing beneath her feet is what moved away from the exact thing that had worked-ish eleven months earlier.

The Grodd series appears to be making the same error in a different key. Superman played it straight and an audience showed up. But Gunn’s Superman wasn’t an origin story. There was no space for young DC newcomers who might want super PJ’s or figures.

The second entry went grim, the third goes mockumentary, and neither one is the sincere version of anything. Two swings away from the register that barely got ‘em across the finish line, and DC’s second foundational figure still hasn’t been introduced.

One Hero Is Not a Pantheon

I don’t think I’m alone over my skies when I submit that DC has only three foundational figures. Superman. Batman. Wonderwoman.

Let’s look at where the other two sit right now. Paradise Lost, the Themyscira series that was supposed to carry Wonder Woman into this universe, is no longer in development at DC Studios.

The Brave and the Bold has no date, and Gunn has said why: “I’m dependent on when there’s an actionable script ready so there is no way of me guessing this. Also, frankly, we’re well into Batman 2, and I wouldn’t want to cloud the Batsphere until after that.”

One of three has been introduced. The other two are a shelved series and a script that isn’t finished. The up in the air WB/Paramount merger can’t be making the dev process any easier.

Start filing edges off a mythology in that condition and the satire arrives before the audience has anything to measure it against. Grodd’s a punchline in a world the audience visited once.

In Defense of Gunn’s Troma Trigger

Give Gunn his due. He treats a comic universe as an ecosystem, a place where bizarre people actually live, pay rent, and get on each other’s nerves. That’s the Troma gene, the Gen-X indie gene, and it produced Guardians, Suicide Squad, Peacemaker and Creature Commandos, which are genuinely compelling and authentically his.

World-first, hero-second is a legitimate strategy. It builds texture fast. It makes a universe feel inhabited.

It also depends on the audience already caring who these heroes are. Texture is the second layer. You can’t start with weathering.

Who Is DC For?

Two audiences.

There’s the ten-year-old who wants awe. There’s the thirty-five-year-old who listens to true crime podcasts and loved Spinal Tap.

There’s room for both, only one can be served first, because the ten-year-old is where the next thirty-five-year-old comes from.

I miss Adam West and Burt Ward playing it absolutely straight in the middle of pure lunacy. I miss Lynda Carter’s Diana Prince, who never once apologized for the tiara. I miss Donner’s Superman, which believed in its own goodness hard enough to make an entire planet believe with it. The Lego Batman Movie works, and that’s a feature-length joke, because the foundation of the myth underneath was strong enough to carry the weight.

I’m only just realizing I forgot to mention Lanterns which is what initially put me in this headspace. Can you guess? It disassembles the myth of a character only Geoff Johns and a few others grok.

Your Order of Operations

Rules for any scribbler walking into an established mythology.

Build the sincere version first, and let that sincerity ride all the way down. No hedge, no wink, no character standing offstage rolling their eyes at the premise. Let earnestness be the wall the rest of it hangs on. Young audiences can forgive a corny hero. But will they attach to one who’s embarrassed to be there?

Wait until the audience loves it without your help. The clock on deconstruction starts when the straight version becomes common property, when people quote it, CosPlay it, and fanboys cannonize it. Miller waited forty-seven years. Marvel waited six.

You can aim the knife at the myth. Watchmen takes heroism seriously enough to break it in public. But satire that mostly signals how cool the creators are can turn the audience into the mark, and an audience that feels like the mark won’t be amused.

Deconstruction is one of the sharpest tools in a scribblers toolbox. One that requires something already be well built.

DC put up one standing set and drew a crowd (ish). Now they’re scribbling a mockuseries about the goofy and grimy adjacent alley.

Maybe before we explore the funny and weird corners of Metropolis, we stand out in the light for a minute and believe a man can fly.

I’m looking forward to seeing newly divorced BatMite, Mister Mxysptik, and Bizzaro working on their Tinder profiles while poolside at The Sportsmen’s Lodge.

The George Washington webcomic had laser hands. Four panels, sitting right there on NBC’s Heroes site, and not one word of it had come out of our scribblers’ room.

That was the winter the WGA strike had us off the lot and on the picket line, November of oh-seven into the following February, while NBC licensing kept the machine running without us. We came back to a corner of our own universe we’d never signed off on. Historical figures with powers nobody in the room invented. Plot threads snagging on our own mythology.

We did the only thing you can do with a mess like that on deadline. Pretended it never happened.

The job is managing an ecosystem, and every ecosystem loses track of its own canon eventually. Entertainment just does it in public, in front of the people who notice first.

Toy Soldiers

A presentation crossed my desk recently, notes from a talk on scribbling fiction for a mass-combat miniatures wargame. Little plastic and metal armies. Rulebooks. Painted regiments shot for a hobby magazine. About as far from a network drama as this business gets, on the surface.

Underneath, this outfit solved the George Washington problem on purpose instead of by accident, and built the toy-box lesson into an actual system, one worth stealing if you’re scribbling for somebody else’s IP.

Toy Box and Sandbox

Any scribbler walking into somebody else’s IP should know which one they’re standing in before they scribble a word.

The toy box holds what’s too important to change. Major players, core canon. Whatever happens to them in your story, they go back in the box exactly how they came out, ready for the next scribbler’s turn. Spock stays Spock. A Riot character’s kit and personality survive your side story intact. Will Graham’s arc has to land somewhere Thomas Harris already wrote.

The sandbox is everything else, self-contained stories, minor characters, elements free to break, bend, or bury, as long as none of it leaks into the box next door. I’ve been running that split since I was about twelve. I refereed a home D&D campaign where the module’s named NPCs stayed exactly where the module left them, and everything in the empty hexes on the map was mine to wreck.

A company selling toy soldiers by the thousands landed on the identical rule from the opposite direction, backed by sales numbers instead of a twelve-year-old’s gut instinct.

Ask which box you’re in before you ask anything else. Most new hires don’t think to ask until they’ve already broken something.

You’ve Seen These Boxes Before

I grew up on Star Wars, the original trilogy, so I notice this split every time Lucasfilm makes a move. The Skywalker Saga is the toy box. Whatever era a spinoff visits, it doesn’t get to contradict what happens across those nine films.

Then look at Visions, the anime anthology where a dozen different studios get to blow up the mythology any way they want because none of it counts as canon. Or look at the High Republic, an entire new sandbox built centuries earlier so scribblers can invent factions and burn down a status quo without anybody worrying it’ll wreck the main timeline.

DC runs the same move under a different label. DC protects its main continuity, the one where Batman and Superman actually live, and quietly resets it whenever a change goes too far. Elseworlds is the release valve, a standing invitation to take the same characters somewhere the main line never could. Superman lands in the Soviet Union instead of Kansas in Red Son. Batman hunts Jack the Ripper through gaslit Gotham in Gotham by Gaslight.

Nobody needs to argue about canon. The label already told you it doesn’t count on this one, so the story’s free to go all the way.

Same architecture as my kid’s D&D table and a company selling toy soldiers by the thousands.

“You fought in the Clone Wars?”

Good worldbuilding plants flags it doesn’t dig up right away. A regiment mentioned once and never shown. A species named in a sentence and gone. A world referenced but never visited. None of it fleshed out, all of it sitting there for whoever needs a thread to pull next season, next expansion, next film. Ob Wan Kenobi casually mentioning the “Clone Wars” set the table for this one.

Lore the audience won’t cash in for years still pays off later. It’s the layer that rewards the fan on a third rewatch, a tenth match, a hundredth mission. A world worth returning to keeps giving you more to notice.

VALORANT runs this on a clock: seasonal content pulls old characters back into focus on a schedule players learn to expect and look forward to.

Never Say Never

Absolute language is a trap dressed up as authority. Always. Never. Every. Lock a rule down that hard in the text and you’ve told every future scribbler, licensee, and fan artist exactly where the fence sits. Fences get climbed. Leave the wiggle room.

Fans don’t just watch a universe, they move into it. Boxing them out of that ownership with an absolute you didn’t need costs nothing today and everything in three years.

I learned the other side of that lesson on Alias, Lost and Heroes, the expensive way. We built a mythology that ran on ambiguity, powers with soft edges, rules the audience filled in themselves.

It worked right up until network executives wanted harder answers on screen, more certainty, fewer open questions. Narrowing what used to be possible loses something that doesn’t come back.

Tokusatsu heroes never break character on screen for a reason. The moment the mask comes off, the myth stops being a myth.

Every Piece Must Pull Together

Fiction is one leg of the table. For a tabletop property, the format also stands on rules, artwork, and photography, painted regiments shot and printed so a fan building their own army knows what they’re aiming for. Pull one leg and the table doesn’t just wobble, it stops selling.

A screen franchise runs the same math on a different scaffolding: story, casting, marketing, and whatever tie-in product justified greenlighting the thing in the first place.

On Heroes, for about two years, every division actually pulled in the same direction on purpose, digital, marketing, and licensing all working the same universe at once, because the network president ordered it. Unprecedented, and it worked, right up until the show stopped being a hit worth protecting. Then everybody retreated to their own corner, and nobody ordered them back out.

Say the Quiet Part About Your Format

Every format has a wall somewhere. A mass-combat game built around identical miniatures runs into a real, structural problem building an inclusive or varied cast when every soldier on the table comes out of the same mold. That’s a fact about the format.

The honest move is naming the wall and designing around it instead of hoping nobody notices. Every medium carries some version of this. You’d better know yours before a fan finds it for you.

Find the Person Who Remembers

Scribbling the rules and the story are not the same, and treating them as one job produces bad versions of both. A studio smart about this keeps dedicated editors and translators between the raw rules draft and the page a customer actually reads, because clarity is its own craft.

The equivalent on a long-running screen franchise is the continuity mind. We had several on Star Trek: Discovery, folks who knew Vulcan mind melds require skin contact — only to be contradicted by a canon busting leadership that wanted a message carried across the galaxy. Oof.

A legacy franchise has a legit cost of entry for anyone new. The wiki is right there for anybody to read. What the wiki can’t carry is what almost got approved and got killed for a reason nobody codified. Get access to the person who remembers.

On Heroes, that person was whoever in the room recalled the meeting where somebody at NBC casually signed off on a founding father with laser hands.

The Fans Have a Spreadsheet

Pretending it never happened worked for about a week. Long-term, it’s a debt that collects interest. Every unresolved contradiction sits in the lore waiting for the one fan with a spreadsheet to find it.

Remember - You don’t own the toy box. Know which parts are locked before you touch them. Leave hooks nobody has to explain yet. Say never as rarely as you can get away with.

Whether the job is a mission briefing for a squad of identical space knights or a teleplay for a starship that was flying before you were born, the rule holds steady.

Play well in somebody else’s sandbox, and next time they might give you a shot at creating your own.

Zeb Finch wrote a sharp piece on Heroes Season 1. His case: that first season ran on one clean engine, every scattered life bending toward a single explosion in New York, and the seasons after it gradually gave up that architecture. Powers got explained instead of felt. Threats got bigger instead of clearer. Death and memory stopped meaning anything once they got reversed one time too many.

He’s right about all of it. And still, he’s done the easy part.

Diagnosing a broken show is a postmortem. Easy to write once the show’s already dead. Fixing it in your own room of scribblers, before the network, the actors, and the ratings start making the decisions for you, that’s the job.

Six rules, pulled directly from where the Heroes engine seized.

Name the One Terrible Thing

Before you break episode one, put your season’s engine into a single sentence. Heroes Season 1 had: how do these people become the only ones who can stop the explosion in New York. That’s convergence, and it’s the engine.

If the best you can do is “cool powers happen” or “big things are coming,” you have a premise, not a season. Write the sentence down. Pin it on the wall. Cut anything that doesn’t feed it.

When You Escalate, Reframe

Scale doesn’t clarify a story. Heroes kept reaching for the next apocalypse instead of asking what the last one changed. When you plan a new season, don’t start with what the next terrible thing is. Start with what the audience now understands that they didn’t before, and what the consequences are for your characters.

Tie Your Device to a Body, Not a Shelf

Whatever your show runs on, powers, tech, magic, a secret, it stays dramatic only as long as it costs someone specific to carry. The moment it can be stolen, swapped, upgraded, or reversed for free, it stops being fate and starts being tools and hardware. Ask what the show loses before you make a thing transferable.

Decide What You’ll Never Explain

Pick it on purpose, before the pressure to reveal arrives. Mystery is a design choice. A secretive organization is always more frightening as a shape in the dark than as an org chart. Heroes lost altitude every time it opened a door it should have left shut.

Plan the Villain’s Ending Before the Network Falls for Them

The best antagonists work when they’re rationed. Sylar was terrifying because Heroes used him sparingly, then couldn’t stop finding new ways to complicate, soften, and redeem him once he’d become too valuable to lose. If your most dangerous character depends on scarcity, decide how they end, or how they change for good, before the mystery and menace becomes impossible to protect.

Guard Consequence Like It’s the Show’s Currency

Whatever your show treats as permanent, death, memory, loyalty, keep it permanent. Heroes erased Peter’s memory, brought Noah back from the dead, and kept Nathan in play through Sylar’s shapeshifting and Matt’s psychic intervention long after the character should have stayed gone.

Audiences will forgive almost any swerve in your plot. They don’t forgive learning the stakes you sold them were never real.

None of this demands a perfect plan. Heroes’ own finale was imperfect, and the season still worked because the foundation underneath held.

Structure doesn’t cancel surprise. It earns it.

The first season of Heroes still has one of the cleanest structural engines in modern genre television.

Not the cleanest mythology or finale. Not the cleanest explanation of how powers work. What Season 1 has, and the series later loses with almost surgical precision, is a simple organizing principle: every strange thing is being pulled toward the same terrible event.

At first glance, Heroes looks like an ensemble superhero origin story: ordinary people discover extraordinary abilities, conspiracies stir in the background, a killer moves through the margins, destiny mutters through comic-book panels and prophetic paintings. That was the hook. But the deeper pleasure of the season is not “people have powers.”

The deeper pleasure is convergence.

Season 1 begins with scattered lives: a cheerleader in Texas, a hospice nurse in New York, a salaryman in Tokyo, a cop in Los Angeles, a politician with secrets, an artist painting the future, a father whose horn-rimmed glasses conceal a moral universe.

For much of the season, these characters don’t know they’re in the same story. The audience does. Or rather, the audience suspects. That suspicion is the engine.

Everything bends toward the same event: the explosion in New York.

That future disaster gives the season shape. It makes every subplot feel like a tributary. Hiro’s journey isn’t just whimsical time-travel adventure. Claire’s invulnerability isn’t just teenage metaphor. Isaac’s paintings aren’t just occult production design. Peter’s empathic power isn’t just a protagonist upgrade.

All are pieces of a single dramatic machine.

“Save the cheerleader, save the world” works because it’s both ridiculous and perfect. It’s specific, mythic, and finite. It takes the impossible size of “save the world” and pins it to one vulnerable human being. That’s structure doing poetry’s job.

The later seasons keep the vocabulary of Season 1: powers, prophecy, secret organizations, catastrophic futures, shifting alliances, comic-book chapter titles. But they gradually dismantle the architecture that made those elements feel inevitable.

Convergence gives way to escalation.

After Season 1, the show repeatedly substitutes a larger premise for a clearer one. Season 2 has the Shanti virus. Season 3 has the formula, Pinehearst, synthetic abilities, and another battle over the future of evolved humans.

Again and again, the series reaches for a new global threat, a new prophetic crisis, a new faction behind the previous faction.

The problem isn’t that these ideas are inherently bad. Any one of them might have supported a focused season. The problem is they don’t create the same gravitational pull.

Season 1 asks, “How do these people become the only people who can stop this one terrible thing?” Later arcs more often ask, “What is the next terrible thing?” That question can generate episodes, but it doesn’t automatically generate momentum.

Metaphor gives way to inventory.

In Season 1, the abilities feel like character made visible. Claire heals, but her gift isolates her from normal adolescence and turns her body into evidence. Peter absorbs other people’s powers because his defining trait is empathy. Hiro bends time because he’s innocent enough to believe destiny should be answered with joy. Matt hears thoughts, which only sharpens his insecurity and domestic failure. Niki’s fractured self isn’t just a power set; it’s a horror story about survival, shame, and repression.

Later, too often, powers become modular. They can be stolen, copied, suppressed, restored, injected, traded, or upgraded. Arthur Petrelli can strip Peter of his abilities. The formula can manufacture powers. Mohinder can experiment on himself. Claire’s blood can restore Noah after death.

Each turn creates plot, but the more transferable the powers become, the less they feel like intimate burdens. They stop seeming like fate and start seeming like store bought equipment.

Sylar is the perfect bridge between those two versions of the show. His original function is horrifying because he turns the metaphor into appetite. If powers are the thing that make people special, Sylar is the man who decides specialness can be harvested.

That’s a great idea. It’s also an idea with a shelf life.

In Season 1, he’s frightening because the story uses him sparingly and directionally: aftermath, rumor, bodies, hunger. Once he remains onstage season after season, the show has to keep finding new ways to complicate, soften, redirect, redeem, and re-monster the guy.

Zachary Quinto is strong enough to make much of it watchable, but structurally the character becomes a problem the show just can’t solve because it refuses to spend him.

Partial knowledge gives way to explanation.

Noah Bennet is compelling early on because we can’t immediately classify him. He is a loving father. He’s also terrifying. The Company is frightening because it feels vast, old, and morally opaque. We understand just enough to know its reach is enormous and its ethics are compromised. But we don’t know enough to reduce it to an org chart.

But TV has a habit of feeding on its own mysteries. The more Heroes explains its institutions, histories, bloodlines, founding sins, and internal power struggles, the smaller the world becomes.

Explanation isn’t always revelation. Sometimes it’s drainage.

This may be the deepest structural wound. Season 1 has stakes because the future feels heavy. Death feels possible. Discovery feels dangerous. Identity feels unstable in ways that cannot be casually reset.

Later, the series becomes increasingly good at dismantling itself.

Peter’s memory can be erased. Powers can vanish and return. Futures can be averted and replaced. Noah can die and be revived. Nathan can die and still be narratively prolonged through Sylar, shapeshifting, and Matt’s psychic intervention.

A story can survive almost anything except the audience learning that consequences don’t stick.

There is, of course, production history here. The 2007-2008 writers’ strike cut into Season 2, and TV gets made under pressures that criticism can flatten too easily.

But the pattern outlives any one disruption. The show’s problem wasn’t simply that one season was interrupted. It was that the series kept reaching for supersized scale after it had lost shape.

The cruel part is Season 1’s finale isn’t flawless. In some ways, it’s less satisfying than the build. But the season survives its own imperfect landing because the architecture has been so strong.

We remember the feeling of being drawn toward something. The pleasure of separate lives clicking into place. The thrill of a genre show behaving, for a while, like a novel with a spine.

The problem with Heroes wasn’t that it ran out of ideas. It had ideas everywhere. Too many, probably. The problem was that it lost the discipline that made those ideas matter.

The first season remains great because it understands that destiny is most powerful when it narrows. The whole world is at stake, yes. But the story works because the world depends on a cheerleader, a dreamer, a nurse, a painter, a father, a brother, a killer, and a city none of them can yet see clearly.

For one season, Heroes knew exactly where everyone was going. Then it mistook having more road for having a map.

The bottle episode is a pressure test for character. Scribble a scene or build a single-map mission, the test doesn’t change.

A fly won’t die. A suitcase won’t close. The West Wing goes into lockdown and nobody’s allowed to leave the building.

Three different shows, three different genres. Breaking Bad’s “Fly,” an hour where Walt and Jesse chase an insect around the meth lab instead of cooking. Mad Men’s “The Suitcase,” where Don and Peggy get locked into an office overnight with a Samsonite pitch that won’t crack and a dead friend sitting between them. The West Wing’s “Isaac and Ishmael,” where a security lockdown traps the senior staff with a busload of touring high schoolers. Same move, every time.

The Studio’s Math

TV calls it the bottle episode. Leslie Stevens coined the term on The Outer Limits: the bottle show, pulling a show out of a bottle, the way a genie comes out of one. Scott Brazil, who ran The Shield, put it less kindly: “the sad little stepchild whose allowance is docked in order to buy big brother a new pair of sneaks.” Both men were talking about money. Fewer sets. Fewer locations. Fewer new faces to light, dress, and pay. A bottle episode buys the rest of the season room to breathe.

That’s the studio’s math. A scribbler wants something else out of the same hour.

The Room Does the Work

I’ve sat in the room for bottle episodes more than once. Alias. Lost. Heroes. American Gods, found its own version of the same play. The assignment always starts as a budget problem. It never stays one. Take the new location off the table. Take the new suspect walking through the door off the table. Take the case of the week off the table. What’s left is two or three people and whatever’s actually sitting between them. Now the hour has to be about that, because there’s nothing else left to shoot.

Lost leaned on this every season, because its whole format was built to avoid it. Six years of flashbacks, flash-forwards, and an ensemble scattered across a beach, a jungle, and a hatch that was never supposed to open. Every time the room bottlenecked that ensemble into a single location, the show landed a pairing the format usually kept apart. That’s what a bottle episode does once you’re inside it. A pressure chamber. Two characters, no exit, no subplot to cut away to, and the story has nowhere left to go but through them.

Confinement Is a Character Test

John August and Craig Mazin spent an episode of Scriptnotes, number 748, working the same territory from the movie side: one-room movies, the 12 Angry Men school of putting a whole story in a single set and refusing to leave it. The question they were circling is the only one that matters here.

What does confinement demand of character?

Pull the location and you pull the easiest tool out of a screen-scribbler’s kit: somewhere else to go. No new setting to smuggle in a plot beat. No hallway to walk down while the scene resets behind you. The characters stay in frame, the frame stays put, and eventually somebody says the thing, does the thing, or breaks.

A budget cut forces that constraint into existence. A showrunner exploits it to save a schedule. A scribbler worth the page rate turns it into the best hour of the season.

Same Trick, Different Rooms

The constraint doesn’t stay loyal to one medium. Drop a player into a single-map mission or a tight hub design and the trap is identical, just with a controller instead of a script page. There’s no new zone to unlock the tension for you and no cutscene detour to solve the pacing problem on your behalf. The mission has to generate its own pressure out of the space, the encounter design, and whatever the player’s character says and does inside it. Same demand a bottle episode makes of Don and Peggy in that office overnight. Same demand a one-room movie makes of whoever’s stuck inside it.

The format doesn’t care if you’re scribbling a teleplay, a screenplay, or a mission design doc. The constraint is the constraint. A room. A map. A cave. A storm. A night that doesn’t end until somebody finally says what they came to say.

Shut the doors. Watch what your characters do when there’s exit.

A mass shooter kept a journal scribbled in Faux Cyrillic. An engineer in the Minnesota Star Tribune’s AI Lab ran it through ChatGPT, then routed the flagged passages to two Russian-language academics to check the work. That translation helped win the Star Tribune a Pulitzer in Breaking News this year.

The Wall Street Journal ran an internal tool called WSJPT, an in-house play on ChatGPT, through every public meeting minute, agenda, and transcript scraped from Kerr County, Texas. The same playbook powered the Journal’s reporting on the Epstein Files months later.

The Associated Press pointed a language model at tens of thousands of leaked documents from a Chinese surveillance firm and came back with a searchable investigation.

Different newsrooms leaned on the machine for different jobs: translation, document search, cross-checking work nobody had time to redo by hand. A human decided what any of it meant.

This record year of Pulitzer AI disclosures is worth every screen scribbler’s attention.

The Count

Pulitzer entrants have disclosed AI use to the judging committee since 2024, the first year the requirement existed. Two winners disclosed that year. Four in 2025, one winner and three finalists. This year it hit eight: five wis and three finals, and the number has doubled every single year the requirement has existed.

Nobody handed a language model their byline. The New York Times used GPT-5 to double-check its manual classification work on a set of SEC crypto cases. AI checking human math. Not doing the math.

The UK Says the Same Thing, Louder

Reporters chasing prizes in New York aren’t an isolated case. A Reuters Institute report released in late 2025, surveying more than a thousand working UK journalists, found that 56 percent now use AI professionally at least once a week. Only 16 percent said they’d never touched it.

Transcription tops the list. Translation and copy-editing follow. Story research, brainstorming, and generating a first draft trail far down the chart, the least common uses.

Journalists are outsourcing the parts of the job that burn midnight oil sans producing a single sentence anyone will read.

Every Scribbler Already Knows This Grind

Swap “journalist” for “scribbler” and the pattern holds.

A procedural needs the actual protocol a field office runs. The version every other show already stole won’t survive a knowledgeable audience.

A historical drama needs primary sources. Secondhand summaries of secondhand summaries fall apart under scrutiny.

A game narrative designer inherits a lore bible with a thousand pages of internal contradictions and needs to find the twelve that matter before the room meets on Monday.

That work sits upstream of the scribbling.

The hours a screen scribbler spends buried in research, cross-referencing dates, hunting the one detail that unlocks a scene, have never been the part of the job a Nielsen number or a review can see. But they do make the visible parts possible.

This Isn’t a New Movie

Word processors reduced the grind of retyping a page. Search engines reduced the grind of finding a fact. Digital editing makes it quicker to find frames. Each one freeing up room for the judgment calls that still needed a human to read and approve.

Read Fast, Judge Slow

Two years of arguing over whether AI can write has missed the more useful adjacent query.

Can AI read fast enough to buy a scribbler more hours to think?

The Pulitzer winners answered by letting LLMs chew through the haystack, while keeping the discovered needle for themselves.

What’s the research grind on your current project? Is it eating hours without producing any scribbled pages?

Could handing that task over to an AI score you a Pulitzer? Or at least let you get to bed at a reasonable hour for once.

ABS. Always. Be. Sleeping (8hrs).

For a decade, thought leaders told us to forget scale and focus on '1,000 true fans' or the 'smallest viable audience'. But what if the advice of those thought leaders was incorrect? What if the key to revenue and algorithmic favor isn’t intimacy, it’s… scale?

There’s a new study out from CreatorIQ and Influencers.club called “The State of Creators 2026.” They surveyed over 5,000 creators across 100 regions, and the numbers should be required reading for anyone entering or leaving art school right now.

Sixty-seven percent of creators earn less than $10,000 a year from their content.

Sixty-two percent say content creation isn’t even their primary income source.

The people posting videos, drawing art, cutting reels, the ones we tell young artists to build an audience like, most of them are not making a living. They’re making a hobby in 4K with better lighting.

Jennifer Cho at CreatorIQ said brands have spent years insisting authenticity and community make a creator valuable. But the money still flows to scale. Follower count and view count predict earnings far better than engagement does.

So the whole “build a small, devoted audience and the money will follow” advice young creatives have been getting for a decade doesn’t hold up against the data.

Reach wins. Not connection. Not craft. Reach.

Instagram is where the real dollars concentrate, with 60% of creators earning $250k or more calling it their primary platform. TikTok has the volume, Instagram has the money.

YouTube, despite being the platform most associated with “real” creator careers, trails both for building a sustainable business.

Even creators who cross the threshold into real audience size aren’t safe. Once you hit 500k followers on Instagram, the tension between what your audience wants and what the brand deal requires jumps to 53%. You get bigger and the trap gets tighter, not looser.

I’ve said it before in some blog posts and LinkedIn, but this study gives me numbers to point at, so once again I’m saying —

Bring Back The Lot

The studios and streamers should consider revisiting the old studio system. Not the exploitative parts. Not seven-year contracts that owned an actor’s life. Not stuff that got broken up for good reason. The part where a screen-scribbler, a craftsperson, an artist, could get hired onto a lot, draw a salary, learn the job from people who’d been doing it for twenty years, and build a career without also needing to become a content strategist, a personal brand, or algorithm whisperer on the side.

The current system is gig work and platform dependency. We told a generation of young filmmakers and animators the path forward was to become their own studio. Write a thing, shoot a thing, edit, market, build the audience, and do it all while competing against creators whose only job is marketing and the audience-building.

Based on the data from CreatorIQ, that’s not a viable career path.

The Farm System We Tore Down

The old studio model wasn’t perfect, but it understood something we’ve forgotten: craft takes time to develop, and people develop craft faster when they’re not simultaneously worried about whether this month’s rent is coming from a brand or an algorithm’s mood.

A staff job with health insurance and a slate of projects to cut your teeth on produced generations of directors, editors, composers, and animators who got to fail on smaller work before anyone handed them something expensive.

I got hired into TV in 2001 after ten years grinding on features, and even that felt like a fast track compared to what a twenty-two-year-old on TikTok is up against now.

The creator economy isn’t a farm system. It’s more like a lottery.

None of this means the creator path is worthless, or that the platform path hasn’t made careers for people. But the data says those careers are the exception, not the rule, and young artists deserve to know the odds before they bet four years of art school debt on a game where barely a third make even $10,000 a year, and only around 5% actually reach a full-time living.

Ouch. Maybe Han Solo had it right when he told C3PO not to tell him the odds. :(

My one note for the industry is: bring back the payroll. Not the old contracts or old power structures, but the investment in developing talent over a years long relationship.

Kill the lottery. Forge the future.

When the phone stops ringing, a scribbler assumes the pages didn’t work. The agent couldn’t get answers. They’ve aged out. Or the last show that didn’t sell left a stain that won’t wash out.

Thousands of scribes run the same solitary autopsy, each one convinced they’re the only one asking what they did wrong.

Enough private dry spells eventually add up to public data.

The latest Writers Guild of America West annual report puts a number on it: 5,072 members reported covered earnings in 2025.

That’s a 10.2 percent drop from 2024 and a 27 percent drop from the 2022 employment peak of nearly 7,000 members.

The Number Behind the Number

Those 5,072 working scribblers collectively reported about $1.7 billion in earnings, only 1.3 percent less than the year before.

Employment fell 10.2 percent. Earnings fell 1.3 percent. The money didn’t disappear nearly as quickly as the people earning it.

That doesn’t mean the average scribbler had a good year.

Aggregate earnings skew toward showrunners, creators, overall deals, and feature assignments, along with a small pool of highly compensated folks. The guild hasn’t published the median figure that would describe the experience of the person in the middle.

What the numbers do suggest: compensation is concentrating among a smaller group of working scribblers. Hollywood isn’t just cutting pay across the board. It’s narrowing who gets invited to participate in what’s left of the professional economy.

Television Removed Seats At The Table

The TV figures make the contraction impossible to dismiss.

Per the WGA’s own Writer Employment Snapshot, WGA-covered episodic series fell from 362 in the 2022-23 season to 228 in 2023-24, a 37 percent decline.

Total TV scribbling jobs fell from 3,138 to 1,819, a loss of 1,319.

That hits every level. Lower, Staff Writer and Story Editor slots that were the best way in the door, fell from 824 to 446. Mid-level gigs dropped from 720 to 421. Upper-level jobs, the Co-Executive Producer and showrunner chairs, fell from 1,594 to 952.

The 2023 strike shortened that season, so the comparison isn’t perfectly clean against a normal production year. But employment fell another 10.2 percent in 2025, long after the strike ended.

The shutdown just poured gas on house that was already burning.

The popular version of this focuses on young scribblers being denied their first chance. True: lower-level TV gigs fell 46 percent. But mid-level and upper-level scribblers lost two-fifths of their jobs too.

Fewer Staff Writers advance into Story Editor slots, and fewer Story Editors get the chance to produce; the pool of future scribbling capable showrunners shrinks with every step.

The industry isn’t just eliminating today’s jobs. It’s cutting the number of people who will carry production experience in ten years.

Features Are Contracting More Quietly

The feature business runs a version of the same problem.

The number of WGA-covered films released barely moved: 198 in 2022, 197 in 2024.

But across the first three quarters of those years, screenwriters reporting work fell from 1,947 to 1,651, a 15 percent decline.

Screenwriter earnings dropped 6 percent over the same span.

Nearly the same number of finished films reached the market, but almost 300 fewer scribblers reported employment.

A released movie is the end product, hiding how much paid work happened before the camera rolled. The data implies the studios can hold the same release volume while employing fewer scribes across development, rewrites, adaptations, and production.

Beyond the Scribble Room

Per the Bureau of Labor Statistics, employment in motion picture and video production reached 289,100 jobs in October 2022. By February 2026, it had fallen to 185,200. Recent employment levels have run about 100,000 jobs below that October 2022 high.

Los Angeles production data tells the same story locally. FilmLA recorded 19,694 permitted shoot days in 2025, down 16.1 percent from 2024. TV production finished the year about 50 percent below its five-year average. Drama sat 43.3 percent below its five-year average; comedy was 66 percent below it.

A scribble room can’t open for a series that’s never ordered, and an editor can’t cut footage that was never shot. Camera operators, costume designers, script supervisors: none of them work a location or dress a character that stayed trapped in development.

Where Scribblers Rank

The Bureau of Labor Statistics offers another useful, if imperfect, comparison.

In 2025, the national unemployment rate ran 4.3 percent. For management and professional occupations, it ran 2.5 percent.

For writers and authors, the BLS category that covers scribblers, it ran 8.5 percent. TV, video, camera operators and editors sat at 9.3 percent. News reporters and journalists at 6.5 percent. Producers and directors at 5.4 percent. Actors sat at 26.8 percent.

These figures don’t measure WGA unemployment directly. The BLS classifies unemployed people by their previous occupation, and the count misses freelancers who are underemployed, developing unpaid material, working outside the profession, or not actively looking.

The 2025 annual average also excludes October, since data collection stopped during the federal shutdown.

Still, the direction reads clearly. Scribblers experience unemployment at more than twice the national rate and more than three times the rate for professional occupations.

In most professions, experience makes a person more employable. In Hollywood, you carry decades of credits and still depend on a small collection of buyers deciding this year they need less shows.

The Closest Comparison Is Journalism

Newspaper publishing employed nearly 458,000 people in 1990. By 2016, that figure had fallen to about 183,000, a decline of almost 60 percent. Employment in internet publishing grew across the same period, but it never rebuilt the old newsroom employment model.

News consumption grew more abundant than ever, arriving through websites, social media, newsletters, podcasts, aggregators, and video platforms. What collapsed was the number of outlets willing to offer stable professional employment to the people producing it.

People still consume enormous quantities of TV, movies, games, podcasts, short-form video, and online narrative content, even as the professional market supporting scribblers keeps contracting.

There can be more content in the world while fewer people earn a living creating it. Journalism has already proved this to be true.

Not Primarily an AI Story

Artificial intelligence will surely factor into whatever comes next. But the contraction has already happened w/o it.

The WGA attributes the TV decline to falling cable subscriptions and programming, rapid expansion and retreat of streaming production, and Wall Street pressure for profitable platforms.

Consolidation adds another pressure. Fewer studios and platforms mean fewer buyers competing for material and talent.

In its recent lawsuit to block the proposed Paramount Skydance and Warner Bros. Discovery merger, the WGA argues eliminating a major competitor would mean fewer opportunities, fewer overall deals, and additional downward pressure on compensation.

That’s the current crisis. AI arrives inside an industry already committed to cutting head count and compressing the number of paid steps between idea and finished product.

The immediate danger sits in something quieter than an AI generating the next great television pilot: companies using AI to justify employing fewer people for coverage, development, research, visualization, rewriting, localization, and marketing.

AI only needs to help the industry decide it can function with 4,000 working scribblers instead of 5,000.

Then 3,000 instead of 4,000.

Stop Treating the Market Like a Meritocracy

Great pages remain more valuable than mediocre ones. Taste, emo intelligence, experience, and judgment carry weight, and may carry more weight still as the volume of synthetic material increases.

But we should stop pretending that employment is a clean referendum on talent.

A gifted scribbler can sit unemployed in a contracting market. A veteran can do everything correctly and still run into fewer buyers, fewer series, and fewer available chairs. Personal improvement can’t solve an industry-wide reduction in demand.

That realization stings, but it can also be useful. It lets us separate the act of creating stories from the specific industrial system that once paid us to create them.

A screenplay can become a book, game, audio drama, comic, short film, a narrative website, or the foundation of an independently controlled story world.

AI and other digital tools can help storytellers move across those forms without waiting for a legacy studio to assemble the traditional infrastructure.

Yes. And… That path leaves the TV salary, the health insurance, the pension, and the vanished apprenticeship system behind.

New Professional Reality

For most of my career, the strategy was: develop material, sell or staff, do good work, and use that work to secure the next job.

That system was never fair or stable. But enough shows got made that a professional middle class could exist inside it.

That middle has been hollowed out.

The next generation of professional storytellers may need to work across formats, retain ownership, cultivate direct relationships with audiences, and build production systems around themselves.

The numbers don’t lie. The trend is clear. Dependence on a shrinking group of buyers has become increasingly dangerous.

I’ve built my share of tech villains. AI billionaires training murder drones, sentient code deciding humanity’s had its shot, the whole Skynet playbook. It’s the pitch every exec understands in one sentence, because the moment the machine turns evil, nobody in the room has to think too hard about who typed the command.

Then a story like this one lands —

— A police chief in Georgia spent more than a year running his ex-girlfriend’s license plate through a nationwide camera network. Hundreds of searches, no warrant attached, just a man chasing a woman with department credentials. When she found out, through an audit log compiled by the watchdog site Have I Been Flocked, the tool that caught him was the same paperwork trail the system generates on everyone, all the time, automatically.

That’s a Washington Post investigation, published this month, sourced from police and court records.

Skip The Robot Uprising

Every scribbler pitching an AI thriller reaches for the same button: the machine wakes up, Skynet blinks, HAL decides, the chip achieves consciousness and turns on its makers.

Real life keeps scribbling an alternate take. The machine stays asleep the whole time. The guy with the login credentials wakes it up and points it exactly where he wants.

History backs this pattern up. The printing press spread propaganda faster than any pulpit could counter it. The telephone gave con artists a direct line into living rooms. Photoshop turned a darkroom trick into a courtroom weapon.

The Camera Network

Flock Safety’s automated license plate readers now cover more than 120,000 cameras across more than 6,000 communities, scanning roughly 20 billion plates every month. The system exists to catch stolen cars and outstanding warrants.

It also turns out to be very good at answering questions nobody in a city council meeting ever approved it to answer.

The Washington Post counted at least 50 law enforcement officers charged with or accused of misusing plate reader systems for personal reasons, most of them tracking women. Forty-six of those cases ran through Flock specifically. Twenty-six involved an officer monitoring a wife, girlfriend, ex-partner, or someone else’s new partner.

Roseville, California ran its own internal numbers separately and found that 71 percent of the felony and stolen-vehicle alerts its Flock cameras generated over two years were built on a misread license plate.

Seven bad reads for every three good ones, on a system meant to justify pulling someone over.

Cities are noticing. A crowdsourced project called DeFlock now maps camera locations nationwide, and a growing list of city councils have cancelled contracts or shut cameras off entirely.

It only took someone with a login and a grudge, and a system that logs everything except intent.

The Permission Machine

New tech rarely invents a new sin. It industrializes an old one.

A location database grants permission to stalk at a scale no human could manage on foot. A synthetic voice grants permission to fabricate a conversation. A generative model grants permission to manufacture evidence a judge and jury might actually believe.

The person wielding the new tool makes that call, every single time. The tool just removes the friction that had slowed bad decisions down long enough for someone to reconsider their worst impulses.

This Has A Name Already

Intelligence analysts figured this out decades before Flock existed. During the Snowden years, the NSA’s inspector general confirmed a dozen substantiated cases of analysts running surveillance systems against people they were sleeping with, married to, or interested in.

Analysts nicknamed it LOVEINT, a play on SIGINT. Most surfaced through self-reporting, often during a routine polygraph. No safeguard flagged the search first.

Swap NSA for a police department and SIGINT for a plate reader, and the story runs on the same engine with a smaller budget.

The Cycle Never Changes

New surveillance tech shows up promising public safety. Someone proves it catches a real criminal. Somebody else, usually someone with legitimate access, realizes it also solves a personal problem.

Why This Belongs In Your Pitch Deck

Pitch decks are full of SkyNet AI stories. Real life is full of ordinary people with legitimate access and a personal reason to misuse it, and that story doesn’t require inventing new physics.

Here are four, built from the mechanism above: ordinary access, an old human want, and a tool patient enough to do the dirty work.

The divorce. An entrepreneur trains a personalized model on fifteen years of texts, emails, home videos, Ring footage, and family photos. At custody hearings, the model starts generating evidence nobody can disprove because nobody can tell it apart from the real thing.

• Anatomy of a Fall (2023) — the entire case turns on whether a recording can be trusted, and the marriage collapses under the same scrutiny as the evidence.

• Black Mirror, “The Entire History of You” (S1) — a couple’s relationship gets destroyed by rewatchable, replayable memory. Same core wound (intimate footage weaponized inside a relationship), older tech.

• Grounding: this is already happening. Family law sites and Forbes both reported 2025 custody cases involving AI-generated audio submitted as evidence, including a fabricated threat recording built from old voice messages.

Or the reputation attack. A city council candidate discovers someone quietly feeding half-true facts into dozens of AI systems. Every search summary and chatbot bio drifts further from the truth. He ends up fighting a future version of his own life story, one that keeps rewriting itself faster than he can correct it.

• Black Mirror, “Bête Noire” (S7, 2025) — A gaslighting parable for the disinformation age. A woman’s reality gets quietly rewritten by someone with a grudge and a device, and nobody else can see it happening.

• Black Mirror, “Joan Is Awful” (S6, 2023) — an ordinary person’s life is algorithmically repackaged sans consent.

• Grounding: A Georgia radio host sued OpenAI in 2023 after ChatGPT invented an entire fake embezzlement complaint against him. The suit got dismissed on summary judgment in 2025, so it’s not a legal precedent, but it’s proof the fear is already in the world and already in court.

Or the cop. A respected detective starts running a plate database against his daughter’s abusive ex. Everyone in his life agrees he’s right to do it. Then he runs her new boyfriend, then her roommate, then anyone she spends more than an hour with. The audience follows every decision and understands every one of them, right up until they realize the cop is now the perp.

• The Conversation (1974) — still the gold standard. A surveillance expert’s own tools and expertise become the instrument of his unraveling. Swap tape reels for a plate database and the shape barely changes.

• Nightcrawler (2014) — obsession with a tool escalating past every ethical line, told from inside the obsession rather than at arm’s length. Useful for keeping the audience implicated rather than horrified from a distance.

• Grounding: you don’t need a comp here, you have a source. This week’s Washington Post investigation is the premise, already reported, already named, already in court records.

Or the neighborhood. An HOA installs cameras after a burglary. Crime actually drops. Six months later, every resident realizes they know exactly where every other resident goes, every night. The block gets exactly as safe as everyone hoped. It also stops being private, for anyone, permanently.

• Home Invasion (2023, Fantasia) — an essay-documentary built around Ring doorbells and the paranoia they manufacture in otherwise safe communities.

• Rear Window (1954) — the classic version of this problem: watching the neighbors is safety and violation at once.

• Grounding: HBO’s Neighbors (2026, A24-produced) launched as HBO Max’s biggest unscripted debut, built entirely on the premise that property disputes between ordinary Americans are must-watch television right now.

The Actual Ending

Science fiction spent years warning us machines would learn to act like people. Reality keeps handing us people who already knew exactly how to act, and just needed a faster tool for unleashing their inner creep.

What’s the real-world AI misuse case you’ve already spotted that would make for a killer streaming thriller? Got one? Now —

— ABS. Always. Be. Scribbling.

A mother sits in federal court, accused of stealing millions from the company she built. Her husband set her up, and her own daughter believes it. Then the doors swing open. A nine-year-old boy walks in holding a beat up blue backpack, points across the room, and tells the judge he knows exactly who framed his mom.

Great cold open. Wrong genre. Turns out it wasn’t a script at all. It ran as a thread on X, and WIRED traced the whole thing back to a source who never existed.

The scribbler behind it goes by BIGBEN, a 21-year-old stock trader in Nigeria who feeds prompts to Grok and ChatGPT for a living and collects a payout from X every two weeks for his trouble.

My ADD brain zeroed in on the structure before I’d even finished reading. Nobody in a scribbler’s room saw this coming. It didn’t come out of a studio note, it came out of a payment algorithm.

The Wound

Rob Cover, a professor at RMIT’s Digital Ethnography Research Centre, spent real time studying these threads and broke the shape down like he was diagramming one of my old beat sheets. Every one opens the same way.

An innocent person gets wrongfully fired, evicted, scammed, bumped from a seat that was rightfully theirs, or hauled in front of a judge who’s got it all wrong. Pick your injustice. The genre doesn’t care which one, as long as it stings.

Act One of every underdog story since fairy tales grew teeth.

The Object

Then comes the flip: a sealed document, a hidden will, a digital recorder planted in a cheating husband’s office that catches everything.

I’ve spent years chasing this exact kind of object, in the Alias room, Lost, Heroes, every room since. The gun in the drawer. The photograph in the folder. Chekhov didn’t invent the principle, he just wrote it down first.

BIGBEN’s version uses the same device and skips the work around it. The recorder shows up once, does its job, and vanishes. No one built the scene where it got planted, and no one earned the payoff. Call it a prop dressed up as a plant.

The Reckoning

Last beat, always. The underdog gets redeemed. The husband, the landlord, the airline passenger who wouldn’t give up seat 14C, they all get shamed in public, usually with a walk of humiliation built right into the visual.

Jenna Russell, a PhD candidate at the University of Maryland who studies AI scribbling, says the whole recipe is close to effortless to generate once a spammer knows the trope menu. Feed a model the injustice, the object, and the humiliation, and it hands back the dopamine hit fully assembled.

Call the whole three-beat machine the Grievance Engine. It only needs three moving parts, and none of them are original.

The Grievance Engine is my own Reel Four, compressed into ninety seconds and running on a phone. It’s the WANT framework stripped for parts. The people running these threads, out of Nigeria or out of the bot farm X sued in Hanoi, didn’t invent story structure. They found the switch every professional storyteller’s been flipping for centuries and wired it straight to a payment API.

X pays for engagement, and the company knows exactly how easy that engagement is to fake wholesale. It sued eight Vietnamese nationals and twenty-five unnamed “John Does” in May 2025 for running the bot farms gaming this exact program.

Clear the bar and the money moves. A Premium subscription, 500 verified followers, five million impressions in three months. BIGBEN says his deposits run $500 to $700 every two weeks.

The Missing Job

Cover’s sonnet has three beats and nothing else. No subplot, no secondary character complicating things, no object planted three scenes early so the payoff lands. The husband is corrupt because the plot needs him corrupt. The judge exists to be wrong, then right, then gone. Every location is a courtroom, a plane, an office, generic enough to be nowhere in particular.

That gap still needs a professional. Haecceity, the specific texture that makes a place and a person feel like they exist outside the frame, comes from research, memory, and refusing to let a generic stand-in do a specific job.

A model hands you an underdog with zero effort. The exact make of the recorder, why the kid chose to hide it there, what it costs a nine-year-old to testify against his own father in open court, that part still takes a person paying attention. And it costs a hell of a lot more than a prompt.

The formula still works. Injustice, object, reckoning has been pulling audiences since long before anyone got paid for it in dollars, likes, or upvotes. It’s built on something true about how people process unfairness.

Examples

Ideas that borrow the Grievance Engine’s skeleton (injustice, object, reckoning) and pay the debt BIGBEN skipped. Each earns the payoff with a planted setup or a subplot.

The Recipe
Injustice: A line cook named Odalys gets fired from a high-end kitchen after her signature dish quietly reappears on the tasting menu under the executive chef’s name.
Object: A stained recipe card in her grandmother’s handwriting, the one step the chef got wrong because he copied it from a photo instead of learning it from her.
Reckoning: She opens a six-seat pop-up two blocks away using the real recipe, no lawsuit, no public callout, and the chef’s restaurant loses its best table to her by the second week.
The Earn: Write the scene where she first teaches that unwritten step to her own line cook, years before the theft. Without that plant, the card is just a prop.

The Claim
Injustice: An insurance adjuster named Everett denies his own mother’s hospital coverage, then recognizes the denial code from a pattern he’s flagged three times that month.
Object: A spreadsheet he’s been quietly building for eight months, unglamorous, full of denials that don’t match the plain language of the policy.
Reckoning: He testifies at a state insurance hearing, and the payoff lands because the audience has watched him lose sleep over it for six episodes first.
The Earn: A brother subplot where the sibling thinks Everett’s a coward for staying at the company this long. Somebody in the story has to be wrong about him for the reckoning to mean anything.

The Deed
Injustice: A farmer named Thanh loses forty acres to a boundary dispute his neighbor manufactured with a forged survey.
Object: An eighty-year-old irrigation ledger in his grandfather’s hand, proving the water rights predate the neighbor’s paperwork by three decades.
Reckoning: Thanh wins the case in chambers by walking his teenage daughter through the ledger’s water-share math, so she carries the knowledge forward regardless of the verdict.
The Earn: Haecceity lives in the ledger’s actual entries, specific dates, specific acre-feet, names of neighbors long dead. Not “an old document.”

The Tape
Injustice: A teacher named Priya gets suspended after a doctored clip of her classroom goes viral, days before her tenure vote.
Object: A backup recording from the school’s decade-old intercom system, the one nobody remembers still runs, catching the real exchange.
Reckoning: The student who doctored the clip has to sit across from her and explain why, and she has to decide whether to forgive him on camera or off it.
The Earn: Plant the intercom system in episode one as a running joke about broken PA announcements. By episode six it’s load-bearing.

The Season
Injustice: A minor-league pitcher gets cut two starts before a call-up, and the front office blames an injury history that doesn’t match his real medical file.
Object: His own file, pulled through a records request, showing the team doctor altered a single date.
Reckoning: He pitches one more season in an independent league instead of getting reinstated. The front office quietly changes its documentation policy after the file goes public, a change he’ll never get personal credit for.
The Earn: The unearned version fixes everything for the underdog. Real craft can leave the call-up gone and let that be the point.

Notice the pattern. Every reckoning here costs the protagonist something, or costs somebody else something specific, instead of resolving clean.

On the fifth Raid Race card of the 2066 season, at Low Coast, nobody got the package out.

My sportsbook had priced that outcome at 6.90. Three of ten punters took it: the three who run their own numbers on each card instead of reading the board. It was the worst night the house had all season, four hundred and sixty-two credits down on a six hundred and fifteen credit handle.

None of those people exist. Neither does the sport. What I want to talk about is that I didn’t decide any of it.

This part matters for everything that follows: I didn’t build the thing that decided it. Claude did. I can’t write Python, and I don’t really understand what a seed does under the hood, not in the way I’d need to if you handed me the file and asked me to fix something broken in it.

I described what I wanted in plain scribbler English, and Claude built it. I ran it and reacted. That’s worth saying straight out, so this piece doesn’t sound like something it isn’t. I barely understand the machinery. I just think it’s cool.

Two Pages

Terra Conflictus 2066 is a tabletop gaming setting I’ve been building for a while. It has cities, factions, a genomic registration system that everybody argues about, and until recently it had no sports, which is strange for a world with four billion people in it who need something to watch.

So there are two pages now.

The Spread is the sports paper. Standings, results, reporting, sixteen circuits across two tiers. There’s a sanctioned tier with franchises and title sponsors and broadcast partners, and an underground tier that is variously illegal, unsanctioned, or so proudly ungoverned that its own organisers refuse to acknowledge half of it. Six correspondents file to it. Merren Dask writes the results and refuses to editorialise. Cass Imrie writes the column about who is getting classified out of which category and has been banned from two circuits and reinstated by both.

The Line is the sportsbook. It prices every one of those circuits, takes bets from simulated punters with bankrolls in unregistered energy credits, and settles them against whatever the sports page reports. Forty-four markets on a given cycle.

Both run themselves. Every few days a scheduled task advances the world by one cycle and files. I gave Claude the two pages as an idea. Claude is the one who figured out how to make them run.

The Rule

The simulator owns the numbers. The desk owns the voice.

I handed Claude that principle. What it meant in practice, dice classes, seeded runs, a ruleset written in code, was entirely Claude’s to work out, and I wouldn’t have known where to start.

Every sport on the board has published tabletop rules, and every one of those rulesets got turned into a Python simulator that implements them properly, dice and all. When the desk wakes up, it works out what is genuinely due today, runs those simulators on a fixed seed, and hands the correspondent a brief.

The correspondent writes from the brief and may never invent a result, adjust a result, or contradict one.

That sounds like a restriction. A language model asked to hold eighty results, a standings table, an injury list and a title race across scheduled runs will drift by week three.

Given the chance, it will also write the season it thinks you want. A seeded simulator produces a season that’s actually true, and the drama gets better because nobody rigged it.

The proof of that came free. Five cards into the first Raid Race season there had been five different outcomes: four separate crews took one each, and on the fifth nobody got the package out at all.

Los Creeps sat third in the championship having won nothing, on consistency alone. Nobody wrote any of that. I especially didn’t plan the fifth card, the one the entire betting page now turns on.

Making a Book That Can’t Cheat

The betting page was harder, and for one specific reason.

A sportsbook needs to price before the event and settle after it. If the odds are set with any knowledge of the result, the page becomes a ledger, and every number on it is worthless.

So the book estimates probability the only way it honestly can: it runs the sport’s own simulator two thousand times on throwaway random seeds, counts how often each result came up, and turns those counts into a price.

Later the fixture is played once for real, on its own seed, and the bets settle against whatever that produced.

The danger is obvious once you say it out loud. If the pricing run and the real run ever share a seed, the book has seen the future and everything collapses.

I told Claude I wanted that to be structurally impossible rather than carefully avoided, and it came back with three separate locks, any one of which would be sufficient on its own.

The book can only open a market on a fixture that hasn’t happened, checked four different ways. The book physically can’t write to the world file, only read it. Every seed the sports desk constructs sits below 10^8, while every seed the book constructs sits above 10^12.

That third one is the load-bearing part, and asserting it wasn’t enough for me, so I asked Claude to prove it instead. The validation instruments the dice constructor itself, watches every random number either side actually builds, and proves the two sets never touch.

Then it does something I’m more pleased with. It builds a deliberately sabotaged version of the book, one where a single sample out of two thousand is the real fixture, and checks that the test catches it. It does, at about thirty-eight standard deviations. The honest book reads flat on the same measure. A null result only means something if the test was capable of failing.

That test also taught me something I had wrong. I’d assumed a clean book should show zero bias on whichever runner wins. It shouldn’t. Winners are drawn from the true distribution, and the book has standing per-runner biases, so the expected reading is about minus a quarter of a percent. Tested against zero, the honest book failed. Tested against its actual null, it passed comfortably. The mistake was mine, and I only know that because Claude walked me through why.

The Book is Bad at its Job on Purpose

A book that prices off perfect probability and adds a margin is unbeatable. Every punter grinds to zero, the leaderboard becomes a straight line sloping down, and there is nothing to read.

So The Line is wrong on purpose. It’s wrong in two different ways, and only one of them matters.

It carries sampling noise, because it prices off a limited number of runs. Noise averages out across a season, so nobody can systematically beat it. It also carries bias: it leans on recent form, and it anchors its no-extraction number to the season average instead of adjusting properly for how hard the individual card is. Bias is wrong in a direction and stays wrong.

The measurement that settled it: holding everything else fixed and trimming the two bias terms drove the sharp punter’s return from plus four and a half percent to minus six.

The punters live on the book’s documented, directional wrongness.

That’s also what makes the Low Coast result work. The book anchors to a season average, so on an easy card it prices “nobody extracts” far too short and it’s a trap. Low Coast was a short window onto a locked breach, genuinely hard, and the same anchor ran the other way and made 6.90 a real price.

The three punters who noticed were the three who model each card instead of the season. The other seven were looking at the board.

What The Validation Caught

Claude validated each simulator across 120-plus runs before I trusted any of it enough to build a betting page on top of it. That part I did understand without reading a line of code: run it, watch what breaks, and go from there.

The Raid Race sim produced five events where nobody delivered anything, because a crew was getting charged an action for dismounting, which isn’t one of the five Raid Actions in the published rules.

The Crackerjack sim was four times too lethal, because a shorthand in my own battle reports had gotten read as a rule. The Long Count killed Brine Quist, who is an established character from published filings, so challengers now get drawn from the circuit’s own regulars instead.

The betting side was worse, in more interesting ways.

A minimum bet of twenty credits silently deleted three punters. Quarter Kelly sizes inversely with the odds, so a genuine 24% edge at 15.5 came out at nine credits, under the minimum, every single time. My two sharps placed almost no bets across eight simulated seasons, which quietly removed the entire point of the economy.

The Syndicate’s cut was supposed to be fifteen percent. It ate 109% of the gross instead, because the spec I’d given charged it on every winning market and recovered nothing on the losers, so winners got taxed and losers never got offset.

I’d accidentally designed a business that could not make money, and it took Claude running the numbers a hundred and twenty simulated seasons to show me that.

The chalk punter created a feedback loop that broke the market. Money shortened a runner, that made it the shortest price, chalk backed it for being the shortest price, and the whole floor ended up on one crew at 2.85 that had opened at 10.0.

And when I asked to widen the board from one sport to eleven, the margin model fell over on lopsided markets. A 98.6% favourite can’t absorb twelve points of margin, so the math dumped 806% of it onto the other side and quoted 7.7 about a 1.4% chance, a number nobody would take.

My favourite one caught itself. The page’s trading note gets generated straight from the ledger rather than written, so it can’t claim anything the book can’t show. It printed a sentence saying a 2,341 credit liability was inside a 900 credit limit. That can’t be true, and the cause was real: the exposure cap was reading an empty list and had never once bound.

Where It Is Now

The Spread files every three days. The Line wakes thirty-six hours later, finds the results waiting, settles, reprices, and takes the next round of bets.

The house is expected to hold 9.29% against a board priced to 10.71%, the difference being what the sharps take off it, and it loses money in 38% of simulated seasons.

Nesta Balint makes 14.4% and is the only reliable winner. Halyard Renko doubles after every loss and blows up in 57% of seasons, somewhere between the fifth card and the tenth, at which point a patron advances him somebody else’s money and takes a permanent share of his action.

Terra Conflictus 2066 is my tabletop setting. The Spread and The Line are both fiction, nobody can place a bet, and the simulators implement the game’s published rules. I built none of it. Claude did, over a few sessions of me describing what I wanted and reacting to what came back, half the time without fully following the reasoning until it was already running. I still don’t know Python. I know what a seed is now, in a general sense. What I know for certain is that the outcomes are real in the sense that counts. Nobody’s thumb was on the scale. Mine included.