I remember very early on in my writing career, I was developing a half-hour comedy with a big production company. They were known for premium hour-long dramas; their comedy department was brand new. Mine was one of the first projects on their slate.

I was working with a lovely development producer. On a personal level, we got along great. But she was killing me with rewrites. All told I did seventeen drafts of this one half-hour script…and at some point it occurred to me that this thing wasn’t getting any better.

There was one moment late on in this interminable process that acted as a kind of coming-of-age event for me. Up until this point, I’d been plagued by “good boy” syndrome. I was young and green, and incredibly grateful for the opportunity to develop this thing with this company.

And so I’d nodded along to every note she had, and felt compelled to implement it. I didn’t want to ruffle any feathers and risk derailing the whole project. And crucially, I thought she must know better than me. She’s a professional. This is what she does for a living. Surely she’d read the latest draft, given it all some serious thought, applied all her development wisdom and insights, and crafted a set of revisions that simply must be implemented.

Well on this day, many rounds in, I was exhausted and irked by the whole thing. It was quite clearly getting ridiculous.

Finally, I pushed back. It was on one note (a big note) she had which I just couldn’t square away. It didn’t make any emotional sense. It barely made any logical sense. This was immediately clear to me. For a moment I wondered whether I was missing something. But I opened my mouth anyway and pushed back. I pointed out, a little nervously, the very obvious ways in which her note didn’t really work at all.

She looked at me. Frowned at the script in front of her. Then she shrugged and said “oh yeah.” And crossed out the note.

Then she moved onto the next one.

My blood ran cold.

Had she really not thought it through at all? What about all her wisdom and insights? Did they fail her this one time? Or had they failed her other times as well? How many times had they failed her? How many notes would she have just crossed out if I’d pushed back? How many rounds of revisions could I have saved myself?

Did she have any actual wisdom and insights at all?

There was no coming back from that. The trust (on my side) was broken. We limped on for a while longer. Then the project died a sad, quiet death.

Sixteen or so years later, just recently, I was reminded of that moment again.

*

AI is eating large chunks of the film industry for breakfast. That much is clear. The extent to which it’s going to chow-down on screenwriting is hard to say. At their latest round of collective bargaining in 2023, the WGA went to war with the studios over AI protections. It showed tremendous foresight, and captured some timely saves. Although I can’t shake the feeling the WGA essentially approached a torrent of water rushing down a canyon with a bucket. Then lifted up the full bucket triumphantly and announced “caught it!”

Especially when lots of writers themselves are using AI in the form of Large Language Models, or LLMs.

Although let’s define our terms. “Writers” and “using” are the operative words here.

Seasoned writers who are practiced in the architecture of story structure, understand their process, and have written scripts before tend to use LLMs to make their life and workload easier. And rightly so. If billionaires and heads of industry are profiting from this step-change in technology, so should we. Doing less work to produce the same or similar result is a form of profit.

As a research tool, an inspiration tool, a handy assistant to back-and-forth with, it’s actually not bad. It can unblock things and gin the process along pretty well.

Then there’s the other type of “writers”. The newcomers. The kind of people who would approach you at a party with “oh, you’re a writer? I have a great idea.”

Now they don’t need to. They can sit at their desk and type “Hey ChatGPT, I have a great idea.”

(Their idea, inevitably, is a premise. Maybe twenty minutes of a movie. Forty-five if they’re lucky. But that’s an article for another day).

I strongly suspect these writers who have neither the knowledge nor experience - nor the tenacity, stamina, passion, commitment - to write a full script the old fashion way, are leaning somewhat more heavily on their chosen LLMs to get the job done.

In fact I have no doubt that some of them are cranking out entire scripts off just a few prompts.

Alright. Back to that come-to-Jesus moment all those years ago.

In researching this article, I decided to see what Claude is made of. So I started tooling around with it, testing its capabilities as a writer. And by some metrics I was very impressed. As I said, it makes sense as a tool under certain parameters.

My conclusion was that it’s like a very enthusiastic intern. It’s incredibly diligent. It’s very polite. It doesn’t complain. It doesn’t need to take snack breaks or naps. And on top of all that it’s very, very obliging.

But after a while I started to notice some recurring patterns. If you’ve used an LLM, these might seem familiar.

You’re absolutely right.

You’re thinking is spot-on.

This is brilliant.

You’re right, I made a mistake.

This isn’t just a great story, it’s a completely new narrative framework.

It started to sound like a sycophant. Or more accurately, it started to sound like a person with no real opinion of their own. No vision. No creative principles. No wisdom; no insight. And the more I pushed, the more I was reminded of that development producer sixteen years ago. Every time Claude was challenged on an idea. It would say “oh yeah”, and cross it out.

It knows everything. But it understands nothing. It inuits nothing. Interrogates nothing. Defends nothing. It lacks the quintessential qualities needed in a writer. Most of all, it lacks taste.

A showrunner is unlikely to ask an intern for their opinions on a script or creative project. But let’s say they did, and the intern stammered an answer, “well, I thought this, and that, and I guess the ending was kinda…”

And the showrunner said “no, but hang on…because…”

The intern would say “no, of course. You’re absolutely right.” And if they continued down that road, the intern would ultimately forsake every note they had offered on the thing in the first place. An exercise in futility.

This is the level of collaboration you’re dealing with when you write with an LLM.

This complete inability to make its own distinct creative choices creates a cascade effect as you move through the writing process.

It’s halfway good at the fundamental building blocks. But as you get increasingly specific (and specificity is EVERYTHING in screenwriting), things start to degrade. Until the point where you ask it to actually write dialogue, and then the quality really falls of a cliff.

But you would only recognise this downward trend if you had vision, creative principles, wisdom, insight, intuition, and taste of your own. If you are lacking in those faculties, you risk creating a vicious cycle of the blind leading the blind.

*

So what does all this mean?

Gun to my head, here’s where I think we’ll land in the next few years.

A development producer will think “I have a great idea.” Or more likely, “I have a great piece of intellectual property.”

Per their updated job description, they will tap a prompt into ChatGPT, and it will grind out a script. Then, per the stipulations of the WGA, the development producer will hire a writer - because screenplay credit cannot be assigned to AI (that stipulation landed in the WGA’s outstretched bucket).

This writer will be a professional with many years in the industry under their belt. They’ll have vision, creative principles, taste, the full toolbox.

The offer will be this: We give you a little bit of money; you make this script production-ready, and you get to take the credit for the resulting film. That means you get the production fee if it gets made. You get the residuals and royalties if it gets a theatrical release.

The writer accepts the job, reads the script, and lets out a deep sigh. Because they’ll realise the script is so dysfunctional they have no choice but to start from scratch. 100% of the work for maybe 20% of their usual writer’s fee. Plus any back-end in the highly unlikely event the film gets made.

And so a new model will be born. A small contingent of high-end screenwriters taking bad work for bad pay, and getting bonuses in success. The ladder underneath them getting yanked up all the higher, if not straight-up set on fire.

That will account for a certain chunk of the work opportunity out there. Maybe 30 or 40%. In other words, the percentage of work that’s always been out of reach for all but the top contingent of professional writers.

As for the rest? Their efforts, however valiantly handmade, will be swept up in the tsunami of AI slop flooding the market. They’ll be just another note in the cacophony of noise.

But that’s quite not as apocalyptic as it sounds.

Because when a situation get so dire, it can actually foster a certain clarity of thinking.

What’s signal in the noise of screenwriting? It’s always a bold, unique voice.

If you have a truly distinctive outlier of a concept, you want readers to be overwhelmed by homogenous crap.

*

Even before the creep of AI cinema was getting too generic. When the money drains out of any sector, that sector gets more risk averse. It’s a fundamental truth of economics. In Hollywood, minimising risk means making films that are precision-engineered to capture the largest conceivable audience. In other words, the opposite of a bold, singular voice.

This clammy scramble to the median of mainstream tastes created a firehose of boring banal “content”, but it was also making it much harder for newer writers to break in.

To get to the level where you might be invited to write one of these units of content films, you had to show you could do it. But writing an action comedy or a superhero movie on spec wouldn’t get you very far. Because you’d be aiming for the median. The middle of the road. You’d be aiming for the exact type of film that gets made, and then dumped on Netflix. Which as we’ve established is itself a capitulation to mainstream tastes. And so what’s to differentiate you from the thousands of other perfectly competent writers doing the same thing?

How could an agent or exec possibly decide which of these solid, well-constructed identikit samples to elevate? Out on the spec market, as in the IP market, it’s almost all noise.

And now that pretty much anyone can scotch together a semi-coherent script using ChatGPT, the noise is about to become total.

But therein lies the beauty. Because agents and execs and everyone in between are now keening desperately for the signal. They are leaning in, cupping their ear, searching for that one voice that will cut right through everything like lemon juice through heavy cream. That’s how stunning deals like this come together. It’s how Kane Parsons (AKA Kane Pixels) gets his first major writing-directing gig at 19 years old.

So rejoice. Your weirdest, loopiest, most out-there ideas? Those ideas that you’re sure appeal to you and you alone in this world? Now is the time to make them happen. Forget mainstream entertainment. AI is cooking that to a crisp. The time has come to make your art. That’s always been a good idea. Now it might be the only option you have left.

The future looks a little bit scary. But the future always looks a bit scary. Writers might well take a bit of a beating. But writers are always taking a bit of a beating. They have throughout all of history. In this round though, they’ve also been given a tremendous gift…the opportunity to be artists again.