Hacking Hollywood - Creativity & AI in 2026

First, let's set the scene: it's 5:30 AM, and somewhere in Burbank the neon glow from a "For Lease" sign flickers over cracked pavement, a sight you could almost miss if you're busy doomscrolling the trades. Somewhere, someone is whistling "Where Is My Mind?" (Pixies, 1988) and not realising how on the nose that really is. Because if you squint at the ruins of what used to be Hollywood, you won't find a grand finale. No monochrome slow motion. No John Woo doves. Just the chilling hush of deletion. Administrative, bureaucratic, and, worst of all, officially sanctioned by the same executives who promise us "this is just the future, trust us!" (As if I'd trust an exec on anything more consequential than a lunch order.)

Let's not sugarcoat this sausage. It isn't "disruption", that favourite Silicon Valley euphemism. This is an inside job, one part Enron, two parts HAL 9000, where the ladder for aspiring talents has been swapped for an endless conveyor belt of generative scripts, deepfaked actors, and the kind of algorithmic taste-mashing that could make even Cronenberg grimace. Irony poisoning, Hollywood edition: the industry's greatest act of self-annihilation is happening in full view, and instead of pitchforks, people mostly sent emails (unread).

But the real gut punch? The numbers. Picture this: between 2022 and 2025, Los Angeles County lost 41,000 film and TV jobs, about a quarter of the people who once made the city's dreams and delusions manifest. In 2025, shooting days are rare: location production already down 22% from last year, and serial TV? Splat. Down nearly 60% since 2022. (This, by the way, would be the place where someone tries to cheer you up with the "creative destruction" spiel. I'm just not that guy.)

I chatted with a production designer last Thursday at MiniBar. His union, the Art Directors Guild, claims three out of four members are out of work. He says his LinkedIn feed is a haunted house of "open to work" banners and wilting portfolio links. I have no data for the fate of hope in all this, although the Entertainment Community Fund (the old Motion Picture & Television Fund, rebranded for a more anxious era) gave out $5.6 million in the first half of 2024 to people who had run dry. That is nearly six times the norm before COVID, and yes, I checked the spreadsheet, despite my usual allergy to Excel.

This is not just doom and spreadsheets, though. These stats translate to flesh and blood: friends whose last pilot fizzled; a neighbour whose grip kit never left storage; my own barista, once an actor, now struggling to make rent. The erasure is not symbolic, it is the erasure of possibility itself. And still, the demolition crew whirrs silently onward.

So, what are we calling this? Not a collapse, since that would mean a bang. More like a hollowing out, slow, methodical, and (arguably) the strangest magic trick Hollywood has ever pulled.

Is that entertainment? (I'm not so sure).

Here's where the gears really grind: what is so gutting about this era is not the arrival of a new gimmick (welcome, disruptive tech, grab a seat), but the cold precision with which that tech has been wielded to vaporise the launchpads, those unglamorous, foot-in-the-door roles, that used to make upward mobility possible. Hollywood is not being invaded; it is being gutted, plank by plank, as meticulous as a Kubrick tracking shot. And AI? It is busy singling out the rungs everyone else once climbed: writers' room assistants, script wranglers, rookie VFX scribblers, PAs running on five hours of sleep. The bridge from starry-eyed to staffed. Not so much a corporate misfire as a surgical strike, the headline crime of 2025, and maybe the existential cliff's edge for 2026 (assuming anyone's counting years at all by then).

The vanished bits are not on the marquee. They are backstage. Which brings us, circuitously, to the "AI as scapegoat" playbook, a sleight of hand I watched unfurl during the 2023 writers' and performers' strikes (I was there, on Ventura, fearing for my lunch money, another story). Suddenly everything was framed in those high-stakes, Hollywood Reporter-ready terms: Would ChatGPT pen the next prestige season? Would Keanu's likeness outlive us all, endlessly deepfaked into John Wick 28? Big, headline-grabbing questions, tailor-made for strike-era brinkmanship. SAG-AFTRA and the WGA both drew lines in the proverbial sand: the 2023 WGA MBA and the later 2025 SAG-AFTRA deal hard-coded certain terms, no AI ghostwriters in credits, no unwilling digital doubles, and pay rules when your collaborator is a server farm. All real, hard-fought wins. I felt their impact, at least for the lucky ones higher up the call sheet.

But here is the twist nobody admits at the wrap party: these victories did something bigger. Once AI had a contract, everyone could breathe easy. Executives even tossed around phrases like "ethical deployment" with the straightest faces I have seen. And almost overnight, the machinery started humming unimpeded, greenlit and legitimised in legalese. The show could go on. For some.

What they realised, quickly and with spreadsheets, was simpler than swapping out marquee names: don't replace the veterans: erase the apprentices. Why bankroll an assistant when an LLM can spit out meeting summaries, pitch alt lines, scrape comps, and tag half the development slate by topic and urgency? Why keep a bullpen of script interns when ChatGPT can wrangle Final Draft exports, fix formatting gremlins, and run a first-pass conceptual read without complaining about lunch breaks? The first bodies on the floor were the entry-level gigs.

Writers' assistants and script interns were the classic on-ramp (part grad seminar, part scavenger hunt) into actual rooms. You learned the cadences, the shorthand, who actually makes decisions; you got mentorship, a network, and a usable mental model of how pages become payroll (or so I'm told). Those everyday tasks (notes, formatting triage, research pulls, blue-sky spitball scaffolding) are now quietly offloaded to generative tools or deleted from the workflow by automation, like cut scenes you'll never see. Studios, under pressure to make the maths sing, aren't using AI to invent new pipelines; they're using it to dissolve the rungs that once let juniors climb up in the first place. Variety clocked it during the WGA stand-off: a slice of execs treat AI as an "optimise" lever for early-stage work, which, fun twist, shrinks the pool of humans who'd become tomorrow's staff writers.

This isn't a glitch; it's the business case. Do you spend $40,000 a year on a script intern who might grow into a writer, or license a system at a fraction of the cost that produces immediately usable scaffolding? The calculus practically writes itself. And the kicker: the contract language shielding established writers hardens the perimeter while spotlighting what's fair game ("protected" authorship over here, "routine" auxiliaries over there), and the latter is ripe for guilt-free automation. I once spent three days on Stage 12 colour-coding beat notes for a punch-up pass; today, a diarisation model plus a summariser would have chewed that in under an hour (I, of course, never use it... or do I?). It's The Devil Wears Prada without the coats: the job that taught you how the room breathes gets absorbed by a service tier. Who learns the craft if there's no place to stand? The ladder wasn't pulled up; it was laser-cut into confetti.

The distance isn't just growing, it's now a canyon with no rope bridge in sight. What's been lost here isn't just a headcount, it's a living ecology: those smaller, crucial habitats (the job-shadow gigs, the "try not to spill coffee on the showrunner" phase) where talent once got its footing. If you're not already plugged into the industry's power grid (with connections, a forgiving bank account, or the sort of stamina needed for a season or two of invisible, unpaid labour), good luck finding a way in. Film programmes (NYU and USC, I see you) keep touting the virtues of mutual aid and on-the-job mentorship, but out here in the wild, that warm rhetoric often curdles when graduates hit the wall of vanished opportunities. Think: sparkling reels, sharp story sense, and a duffel bag of technical tricks, then the sinking realisation that nobody's buying tickets to your debut. The uncomfortable truth: Hollywood, which once feasted on the unexpected (the voice from nowhere, the left-field perspective), now risks receding into a monoculture of heirs and hustlers. Maybe not fatal (yet), but definitely claustrophobic.

The quiet scramble: what AI in Hollywood actually looks like after 2025

Here's the twist: what you see on an exec's LinkedIn feed isn't where the action is. The real transformation, right now, is a quiet retooling of daily routine. McKinsey (the usual suspects) sat down with execs, creatives, codewrights, even a few sharp-tongued academics, and the takeaway is, we're way past the hackathon stage. Across the big studios, the emphasis is on making the gnarly parts of pre- and postproduction more tractable, and not in the way that just sticks a chatbot on the desk and calls it a day. It turns out those early and end-of-pipeline tasks (think AI-powered previs tools, algorithmic set mock-ups, camera move simulators that play chess with your blocking) are a sweet spot. I overheard one product lead at Cinegear (2024, hot June asphalt, regrettable sunglasses) put it this way: "If your previs isn't charting a straight path from A to B, it's a mess you pay for twice, once in overtime, again in editing.".

A/B-testing your shots before a slate ever claps means you waste less time on set and leave more room for actual invention once filming wraps. That kind of A/B test doesn't sound huge at first, but the changes are real. Preproduction cycles that once took teams of artists weeks can now be finished in days. The budget impact is striking. Adobe's Firefly Foundry setup with commercial-safe, IP-protected models built for specific rights holders is now a core part of studio operations. These are not generic Swiss Army knives. They are custom tools designed for a single company's workflow, locked to enterprise SSO, and out of reach for small independent teams.

The area where changes become truly visible is postproduction. Cosmetic fixes, age rollbacks, dialogue swaps, and rotoscoping are now mostly machine-assisted or fully automated. A former studio executive said, "Vanity fixes are a significant share of visual effects, and that's now pretty easy to do with AI. These tasks used to be incredibly manually intensive." Efficiency gains of 80 to 90 percent in VFX and 3D asset creation are being reported. This means smaller crews, faster turnarounds, and lower costs for the repetitive technical work that used to require large staffs. On 12 June 2025, in a Cahuenga edit bay, I watched a colourist remove ten years from a close-up in Resolve with just one node.

Vitrina is now marketing to studios with a pitch to "de-risk the greenlight" through algorithmic development. Their claim: more than 65 percent of global streaming acquisitions are now influenced by predictive models that estimate licensing value before the script is even finalised. It is a Moneyball approach to development, letting spreadsheets inform creative decisions. Whether this makes the industry more efficient or just more uniform remains uncertain.

Why rely on a studio executive’s guess when you can use a neural network? The traditional instinct-driven approach to greenlighting projects is now being replaced by machine-driven forecasting. If your pitch does not align with the algorithm’s priorities, it is unlikely to progress, regardless of its creative originality or cult-classic potential. Entire genres and riskier projects may be rejected before they are properly considered.

This is already happening. Large studios are experimenting with these methods. Ted Sarandos at Netflix recently said the aim is not simply to produce cheaper films, but to make everything "10% better," although the operational meaning of this remains unclear. This suggests a shift in focus: not just saving money, but improving output for the same budget. Amazon MGM Studios has brought in Peter Friedlander from Netflix to lead international television, signalling a commitment to this data-driven model. According to reporting in the New York Times, Amazon’s automation strategy may aim to replace up to three-quarters of its operations staff with robots and automated processes. The actual numbers may differ from these projections, but these ambitions show a significant move towards automation.

Yet, there are uncertainties.

The main event arrives in September 2025 with the launch of Sora 2 and its accompanying Sora social app. Anyone with a smartphone can now create convincing video clips generated entirely by AI. Within hours, servers are flooded with unauthorised videos featuring Disney princesses, Marvel characters, Pixar-inspired animations, and content based on a wide range of intellectual property. None of these have approval from the rights holders. Disney, which had previously remained quiet in the debate about OpenAI, responds with legal action, but the speed and scale of content creation far outstrip traditional enforcement methods.

Attempts to remove infringing videos prove ineffective. Each time a violation is flagged, more appear just as quickly. The pace of new content makes it almost impossible for legal teams to keep up. OpenAI initially suggests that creators should simply "opt out" if they do not wish their work to be used, but this solution is impractical given the sheer number of uploads each day.

Controversial and offensive videos soon attract media attention. In response, OpenAI changes its approach, switching to an "opt-in only" policy for content. However, many users continue to find ways around restrictions by creating new aliases or using altered images.

When asked about the situation, Sam Altman, OpenAI's chief executive, admits he did not anticipate the backlash over deepfakes and the misuse of copyrighted characters. This response highlights the unexpected challenges facing those in charge of major technological shifts.

Things become unstable at this point. The legal framework supporting AI and copyright is still unclear. The New York Times is currently suing OpenAI and Microsoft, claiming both companies used millions of Times articles to train their large language models without permission. The companies respond with a fair use defence, arguing that their model-building is transformative rather than commercial, and benefits society. However, if the Second Circuit rules in favour of the Times, the consequences for Hollywood could be significant.

If OpenAI must secure licences for all the content its models use, many entertainment executives running AI content operations would find they are relying on copyrighted material without proper permission. This is a real issue. Studios using Sora-type models, production companies using generative AI for pre-visualisation, and VFX shops using automated tools are all vulnerable to copyright challenges. The Motion Picture Association has issued a statement urging OpenAI to take immediate and decisive action, though this comes from an industry that has previously allowed AI models to use Disney content. At the same time, talent representatives at WME and CAA have announced their determination to protect their clients' likenesses from Sora. Even so, the technology operates regardless of permissions, only reacting if there is a clear legal ruling. If courts decide to restrict AI use, studios relying on these tools may face major liabilities. What happens in that scenario remains uncertain.

Studios are playing calendar chicken with the courts: either the docket crawls long enough to amortise the risk, or they cut hush-money deals with rights holders before the damages meter starts whirring like a Vegas slot. Federal litigation often runs five to ten years (Jarndyce v. Jarndyce energy, minus the fog and wigs), so the cold maths is a risk-adjusted NPV that favours delay. That calculation works only if copyright enforcement stays jammed and underfunded. Building the next era of entertainment on that assumption is like pouring a foundation on wet drywall. It stands up until it doesn't.

The producer's dilemma: building with uncertainty

If this ambiguity feels like a mild head cold for studios, it is pneumonia for small shops and indies. The cheery pitch that AI would "democratise" production, letting outsiders cut costs and compete, has not materialised. The gains are going to players with big budgets and real infrastructure (SRE/MLOps teams and a bench of lawyers who can quote Rule 26 from memory). You can now conjure near-blockbuster VFX for thousands instead of millions, but only if you can deploy the stack correctly, train and retrain your team, wrangle licensing, carry errors and omissions insurance, and sit calmly through the possibility of an injunction landing mid-shoot (or mid-stream, which is worse).

A studio with a $10 billion annual content spend and a legal department the size of a small airport can treat those variables as line items. An independent filmmaker fronting costs on two credit cards cannot. I keep hearing the counter: "But a five-person crew can now do what took fifty!" Possibly true, in the same way Mario could skip half the level with a perfect warp, great if you've practised for 200 hours and know every pixel-perfect frame (I, of course, never try unstable checkpoints on weekend shorts... or do I?). In practice, the supposed levelling effect looks like enterprise software with a glossy creative UI, dazzling if you've got the GPUs and the pipeline, plus indemnities and the stomach for discovery disputes, paralysing if you don't. This raises an awkward question: if "democratisation" requires a corporate legal squad, who exactly gets to vote?.

The subtler rot is upstream: AI in development has split the field in two. The majors can now spin up hundreds of high-fidelity concepts (paper pilots, lookbooks, teaser comps) via algorithmic pipelines for roughly what one human-heavy development cycle used to cost. Smaller outfits, still jogging through the traditional gauntlet, are not losing to superior taste; they are getting steamrolled by throughput. If a studio can bandit-test 500 ideas for the budget that previously bought 50 hand-built ones, that is a structural edge talent or vision cannot outlift (I tried a weekend sprint of GPT-assisted loglines in a Burbank coffee shop, the spreadsheet ate me). Think StarCraft APM, but for greenlights: volume wins long before art shows up.

You can already see the defensive formation being drilled. Warner Bros. Discovery just unveiled Discovery Global's 2026 upfront, combining the domestic ad inventory of Warner Bros. and Discovery into one sell sheet. That is not a creative duet; it is a shield wall. Pool the reach, pool the first-party data, and sell certainty. The subtext for 2026 is blunt: scale plus consolidation beats novelty. Helm's Deep rules apply.

Meanwhile, the one patch where independents still had oxygen, YouTube and short form, has been rerouted by the same currents. Channels leaning hard on generative tools are rocketing. Translation: audiences do not flinch at machine-made if the watch-time curve behaves (CTR up, retention smooth). But here is the catch: open the gates and the noise floor goes supersonic. The barrier drops; the Library of Babel moves in. I would never queue a hundred auto-voiced shorts at 2 a.m. to see what the recommender gods prefer (or would I?). Democratisation, yes. Also, a collapse in signal-to-noise that makes "finding the good stuff" feel like No Man's Sky without a map.

When prompt-to-video is good enough to pass as "professional", discovery stops rewarding inspiration and starts rewarding ops. The game isn't paint-by-soul; it's click-through engineering, retention curve smoothing, upload cadence, tags, and thumbnail A/Bs until your eyes glaze. Think Moneyball for the For You Page: whoever min-maxes the funnel wins long before the muse shows up. I certainly don't schedule six alternate cuts with three thumbnail variants at 1:17 a.m. to tickle the recommender (or do I?). The truth is, creativity now audits as optimisation.

Now for the real siren in the shaft: the screenwriter, specifically the not-yet-in-the-guild one. Not the veteran with reps and a produced sample; the kid trying to get a first draft read without an uncle at CAA. Nick Geisler mapped the mechanism with painful clarity. In the mid-2010s, he was cranking how-to copy for a tutorial farm, actual craft on miserable rates with no benefits. That rung is gone. ChatGPT can spit the same baseline competent explainer in seconds, so the pennies-per-word market evaporates and the on-ramp with it. I wrote a "how to reset a router" piece from a WeWork in 2016 for rent-and-ramen money; today the prompt would do it faster and, annoyingly, cleaner.

Here's the twist that stings. The WGA's hard-won safeguards (they ban replacing human authorship with AI and prohibit forcing writers to use AI or using AI output to interfere with separated rights) fortify the castle walls for credited writers. But (there's always a but) the unintended consequence is a classic water-finds-a-crack problem: studios route around the protected zone and automate everything outside it. Entry-level writing work (those cheap, repetitive assignments that teach you basics like format and deadline discipline, including notes-call etiquette) gets Hoovered by models. So the ladder snaps at the first rung while the drawbridge stays up. It's Les Mis, but the barricade is made of policy language.

Once, a junior writer could be brought on to handle outlines, beat sheets, and continuity fixes, gaining experience on small tasks before moving up. Now, it is cheaper to use a model for coverage, synopses, and script cleanups, then have an experienced writer handle the main pages. The opportunities for learning and advancement shrink.

The 2025 SAG-AFTRA agreements address this, but only for a specific group. Digital doubles and synthetic performers must trigger pay if an AI performance replaces what a human would have done on set. The details remain unclear, such as how to audit usage, who verifies it, and what qualifies as a replacement. Technology advances quickly, but contracts lag behind.

These protections only apply within the union. Outside the guild, no such rules exist. An independent producer or streamer using a non-union team faces no restrictions; digital replicas and AI-generated performances can be created with no payment for the people whose previous work trained the systems.

There is little evidence that working screenwriters are using AI to produce meaningful sections of scripts. Reviews of 2022 to 2024 Black List material found almost no signs of model-generated dialogue or action. This is not because professionals avoid AI or because the WGA prohibits it—neither is accurate. At the level where structure, subtext, and rhythm are crucial, AI output still falls short. When the writing itself is what matters, autocomplete is not enough. Yet.

ChatGPT can produce functional, generic dialogue. However, it cannot create the distinct spark of character or reshape a scene until it works. Despite claims to the contrary, experienced writers are not being completely replaced by algorithms; instead, they are moved down the priority list as entry-level opportunities disappear.

The shift in Hollywood’s business model did not start with a surge of creativity or fan demand. It happened when the numbers became unsustainable. Streaming companies grew quickly in the 2010s, but eventually faced limits: expansion could not continue indefinitely, and producing TV still costs money regardless of the platform. Netflix, for example, faced setbacks before successfully launching its ad-supported tier. Others—such as HBO Max and Paramount+—spent heavily but did not reach profitability.

As traditional cable declines, Standard & Poor’s forecasts that 2025 will mark the point when linear TV enters terminal decline. Revenues are down, audiences are shrinking, and major companies are restructuring their businesses. Comcast is spinning off its cable channels, Warner Bros. Discovery is being broken up, Netflix is interested in film and streaming assets, and Skydance may purchase Paramount. These are not bold reinventions but urgent responses to a faltering industry model. The established approach is being dismantled piece by piece.

What's emerging from this merger mania is not just a slimmer version of the old studio system, it is a new operating system for culture. The stack is now end to end, with production, distribution and advertising fused into one machine that feeds on first-party data and makes decisions in real time (yes, slightly Borg-like). By 2026, the winners are the platforms that treat content like a fully instrumented supply chain, with AI smoothing kinks from green-light to programmatic ad slot.

Netflix wrote the playbook and kept the home advantage because it welded creative judgement to algorithmic optimisation. I sat in a Netflix pitch room in 2019 and watched someone pull up a completion-curve cohort chart like it was an ECG for story beats (I, of course, never peek at those dashboards in my own work... or do I?). When they bet on Squid Game, it was not a vibe check, it was collaborative filtering and engagement telemetry mapped to global taste. The result managed an odd mix: artistically coherent yet tuned for the platform's discovery engine. Others copied the chassis with uneven results, the same metrics, different organs.

The power shift is the real plot twist. For the first time, the companies that own the pipes also want to write what flows through them. When the distributor becomes the studio, who actually "owns" the audience? Apple is spending heavily and treating the shows as glossy billboards for devices (a loss leader for the ecosystem). Amazon treats Prime Video as anti-churn epoxy for retail (great for LTV, not required to stand alone on a P&L). Google's YouTube took the other path entirely: skip prestige subscriptions, dominate creators and short-form at planetary scale. Different rings, same Sauron energy.

But here's the thing: once the recommendation engine is also the green-light committee, the line between "what people want" and "what the system can efficiently deliver" gets blurry. Functional? Absolutely. Freeing? I'm not so sure.

Meta (née Facebook), having quietly tossed its scripted originals into the fires of Mount Doom (or maybe just the recycling bin at Menlo Park), is now shovelling billions into AI tools, the kind that fuel creator output everywhere else but its own platform. It is a reverse play: rather than hiring dozens of showrunners or building its own Westworld, Meta wants to be the arms dealer for the creator economy's widening war. Remember when Netflix staked its claim by staffing up with script editors and producers? Or when YouTube, in a flex worthy of Lex Luthor, decided it needed zero traditional studio infrastructure because its secret sauce was not content production, but industrial-scale curation? Now, in the 2026 power rankings, YouTube sits atop the pyramid by orchestrating. It aggregates a surge of outside content and lets its neural nets sort and recommend it on the fly. What are the implications for jobs? Once the recommendation gods run the city, most of the old trades vanish overnight, and what replaces them is anyone's guess. There has never been a model like this before (unless you count the Library of Alexandria with sponsored content and no one ever losing a scroll).

The union victory that became a trap

Ah yes, the 2023 WGA stand-off: memory lane for some, trauma for others. I still remember watching colleagues tally strike mileage like RPG hit points. It ended, to hear the rhetoric, in a double victory lap: writers and actors securing, on paper, major safeguards against the synthetic onslaught. The guilds wrung real concessions: authorship locked to flesh-and-blood humans with contracts requiring consent for digital twins, and new minimums pegged to AI-assisted labour. There were celebrations, and for a month the air was thick with "historic" this and "landmark" that. But a peculiar feedback loop set in: as soon as protections for established talent hardened, it became almost algorithmically sensible to prune everything outside that garden. By rendering the boundaries bitterly clear (union labour, sacred; everything else, grist for automation), the agreements handed the studios a highly efficient knife. Ask yourself: if entry-level work can be rendered unprotected, why keep it around at all?.

With shrinking budgets and relentless ROI spreadsheets, producers had a stark choice: keep paying for entry-level jobs, the kind that create tomorrow's union talent, or feed that work to automation. If you're a line producer doing the maths, which do you think wins? Junior writers became a rounding error nobody was incentivised to preserve. Much of this traces back to the "augment, not replace" phrasing in the latest Writers Guild contracts, a clause that sounds reasonable when you picture a seasoned screenwriter, union card in wallet and three pilots under their belt, fielding AI-generated alt lines (and probably mocking most of them over a double espresso). But give that same latitude to the studio's spreadsheet chiefs, and you get a different outcome. It does not "help" junior writers. It removes them, with their tasks outsourced to the algorithm and no protest, because there is nobody left to protest.

The guild, to its credit (after many sleepless negotiating sessions at the Beverly Hilton), tried to close the obvious loopholes. It prohibited companies from using AI-generated copy to deny separated rights, and it blocked the move of padding credits with LLM output. Those clauses matter if you're already inside the velvet rope of union membership. But they cannot restore ladders that have been cut off at ground level. There is no clause for protecting jobs that now exist only as a memory, and no path for aspirants to gain the credentials needed for the protections to apply. You get a sort of regulatory ouroboros, only those inside the fortress get the shield, while everyone else peers through the gate, wondering if the next rewrite will be handled by a human or a silicon ghost. (Spoiler: increasingly, it is the latter).

Here is the gamble: as of Q1 2026, studios are betting that AI-assisted pipelines will produce enough watchable content with enough cultural relevance to justify the investment. The expectation is not just faster production, but improved subscriber retention, healthier ARPU, and satisfied advertisers. This is a significant, but uncertain, claim.

A Tencent executive recently predicted that within two years, up to a third of long-form film and animation will be "dominated by or deeply involving AI". This implies audiences will accept large-scale synthetic workflows, and the content will keep viewers subscribed and attract advertisers. Some evidence suggests this is happening; certain YouTube channels using AI tools have gained large audiences. I have watched these on my phone at the Burbank airport between delayed flights. These channels, however, focus on formats where output speed is more important than quality—listicles, lore explainers, and "cozy" background loops—where volume matters more than substance, and viewers understand that trade-off.

Will this acceptance extend to prestige drama or stories driven by complex characters? For shows like The Bear, where subtle storytelling is central, or the slow-burn narrative of Better Call Saul, it remains doubtful. The issues go beyond surface-level oddities to deeper narrative elements, such as subtext and emotional nuance. If systems focus on retention data, they may remove the quirks and specifics that make stories memorable.

A major concern is that studios are changing their production processes before fully understanding the technology’s limits. There is a risk tied to both platform stability and content ownership. If AI-generated content underperforms with audiences or if copyright issues force a rapid change in approach, the workforce and mentorship structures needed for recovery may already be gone. This situation resembles the Sorcerer’s Apprentice problem, but with financial statements and residuals on the line.

This outcome is not hypothetical. It is a probable result of current industry trends, and serves as a clear warning.

Netflix didn't reach the top of streaming by worshipping the dashboard. It studied what people actually watched, long before it tried to contort production around predictive models, and built from there. The advantage was the recommender plus the hiring and the staying power of the creatives behind the tiles. People first, then the maths. As the field crowds with big tech and social video, Netflix is leaning harder into that advantage: keep backing talent while others swap out the writers' room for an inference cluster (or so the slide decks promise).

What 2026 actually looks like: as I type this (early January 2026, at a table in Los Feliz that wobbles like a bad steadicam), the smart money says neither full robo-cinema nor a soothing return to "normal". Expect a split: a few giants with the scale and cash, and the tolerance for model risk, to ride the AI transition, and a long tail of smaller shops quietly squeezed out. Jobs? Still falling. The messaging will be tidy: cost-focused executives will cite "automation" as the main reason for headcount cuts even when the real causes are strategic misfires and margin pressure. Some will use AI as the perfect alibi, "inevitability" as cover for yesterday's bets gone sideways, because shareholder gravity demands a story with clean lines. Think Succession, but with more finely tuned CPM targets and fewer F-bombs. I, of course, never refresh the Top 10 row before bed to see which version of the future is winning... (or do I?).

The much-discussed union firewall from 2023 to 2025 will still have some meaning, but only for those who can secure and hold onto a union job. These positions are becoming scarce as more of the workforce is pushed out. Legally, disputes over copyright are about to intensify. The ongoing arguments over AI training and fair use are moving towards major legal battles, and companies that relied on unlicensed data may soon face serious consequences. Disney will remain highly protective of its intellectual property. The company will be tough in court but might still make occasional, tightly controlled licensing agreements with AI vendors, always favouring Disney. Independent creators now face a difficult choice: accept whatever terms the AI platforms demand or risk falling behind competitors who can produce new work quickly using AI tools. Consolidation is now the norm. Without the scale of Netflix or a strong connection to a big player like Amazon, mid-sized media companies are searching for partners, mergers, or investors. The result is a market dominated by a few large players, with Netflix and others battling for streaming, YouTube leading in short-form and independent content, technology companies using media as part of larger business strategies, and traditional firms combining forces as their numbers dwindle. The rhetoric from executives will become more common, using "automation is destiny" as a justification for cutting staff, making these decisions seem inevitable and avoiding discussion of their real motivations. This approach lets leadership avoid difficult questions while the workers most affected by these changes have the least influence over what happens next.

Here's the question everyone keeps tiptoeing around while pretending to admire the wallpaper: what happens when the marginal cost of making a show rounds to zero? If (when?) a small squad, or one unreasonably caffeinated person, can spin up an entire season with commodity models and off-the-shelf tools, the old Hollywood maths implodes. The whole edifice was built on a single, sturdy premise: high-quality work is inherently expensive, with costly crews and stages, plus heavy distribution capex. Break that premise and the structure caves in on itself. Not with a bang, more like a quiet, embarrassing crunch.

And here's the trick Netflix, YouTube, et al. already worked out: the game is not "who manufactures the content most elegantly", it is "who gets you to the right thing, right now, with minimal regret". Discovery and personalisation beat steel and concrete infrastructure. Collaborative filtering beats soundstage footprint. Their edge is not that they shoot better, it is that they know what you will watch and deliver it with unnerving reliability. If somebody else attains that audience understanding by different means (say, AI-accelerated development plus ruthless A/B pipelines), they inherit Netflix's advantage at a fraction of the cost basis. Which is exactly why Netflix hugs its talent while others sprint towards headcount guillotines and workflow bots.

Because talent, real taste, the knack for spotting worthwhile stories and the social skill to assemble a team, with the fluency to map narrative to audience segments, remains maddeningly resistant to automation. You can generate infinite options, sure; deciding which one deserves oxygen is the hard part (ask any dungeon master or product manager). As I write this in early 2026, my own late-night experiments cutting shorts with auto-assembly inside Resolve have taught me a humbling truth: the software never gets tired; my judgement absolutely does. I, of course, never lean on a recommender to decide rough-cut order... or do I?

The truth is, if quality becomes the gating variable again (if we loop back from sheer volume to "does this actually work on a human?"), the studio that nuked its development brain and flattened its middle layer (where so much tacit knowledge actually lives), then rebuilt the pipeline around generative widgets, may discover it cannot compete at all. Recommendation engines and latent embeddings can surface taste; they cannot substitute for it. Think of it like the Star Trek replicator: once matter is cheap, culture reverts to scarcity in curation and trust. And if you have automated away the people who knew how to say no at the right moment, good luck replicating that.

But here is where the time machine stalls. If you remove all the entry points, there is no one left to move forward. Studios that eliminate entry-level jobs and block opportunities for newcomers eventually notice, too late, that they have no talent pipeline. No junior writers, no new showrunners, no one to take over when the current team moves on or starts a new venture. The next Lena Waithe or Donald Glover could be sending scripts into a void.

So, how do you avoid this slow decline? Here are some suggestions, with the more unlikely ones at the end.

First, recognise that entry-level jobs are not acts of charity; they are how you develop future talent. People who handle basic tasks today approve big projects in ten years. Cutting these jobs is not efficient—it is like destroying designs to save drawing tables. Studios that value mentorship, development, and keep assistants and interns will have more choices in 2030, while others may end up restarting shows using only recruitment algorithms.

Second, legal standards need to be clarified. Depending on courts to approve your AI data collection is purely hopeful. Clear rules about what data can be used for training, what is fair use, who gets paid, and when, would allow proper business planning. Without this, the industry risks relying only on issuing copyright takedowns.

One more idea: unions that represent everyone. Imagine assistants, mid-level creatives, and stars all bargaining together, without exceptions or loopholes for AI. This would change the financial calculations. Human staff would become cost-effective again, or at least the labour market would reflect real conditions rather than betting entirely on automation.

Will any of this happen before studios resort to unusual strategies to make use of empty facilities? It seems unlikely. But if a future where newcomers are welcome is the goal, something must change.

At present, every financial incentive points toward automation, since entry-level, non-union jobs are easy to cut. If those roles fell under union jurisdiction, suddenly it is not so simple: either studios keep actual humans at minimum guild rates, or pay a hefty premium to automate them out of existence. (A rare case where "just hire people" might even be cheaper.) Not that any of this is happening, at least not in this timeline. The prevailing logic is to double down on bots and flood the market with AI-generated filler while hoping the copyright asteroid never makes landfall. The gains won by striking unions, real but narrow, do not reach far enough to stop the boardroom's planned obsolescence for newcomers. Would-be writers and assistant producers are atomised, set against one another in a freelance arena with no leverage to bargain collectively, or even to secure a foothold long enough to protest as the way in dissolves beneath their feet.

Which brings us to the bleak midwinter: January 2026. By now, show business has moulted into something new and, maybe, unrecognisable, an ecosystem pared to the bone, with the damage locked in. Jobs are a receding tide, unlikely to return except as a ghostly rumour. The circus of media consolidation only gathers speed, with fewer bosses and more gatekeepers. Streaming still burns cash, so budgets are cut and tactics get stranger (note to self: learn to pitch quiz shows for TikTok). Meanwhile, the copyright powder keg, after years of quietly sizzling, is set to blow, either studios cave to lawsuits or ditch the tools they bet the house on, and neither option is easy. Who is left standing? The handful of giants (Netflix, Amazon, YouTube, Apple) whose war chests let them treat media as a rounding error in vast, metagalactic business empires. Everyone else clings to the wreckage and looks for a way out that probably is not coming. (Not that I'm giving up yet... but ask again in 2027.).

The legacy conglomerates that manage to stay together will hold on to defensible areas such as live sports and breaking news, where their existing advantages still matter. ESPN in Bristol, Fox Sports in Century City, the big network newsrooms at 30 Rock and Hudson Yards: these franchises keep a foothold because immediacy and distribution still give them a clear advantage. Everyone else—the indie set without powerful hardware and tailored recommendation systems—gets pushed to the sidelines.

What gets lost are the middle layers: the small development teams, the apprenticeship steps, the junior jobs where you learned to handle feedback and write proper script coverage. My first month as a PA in Burbank was mostly lunch runs and script breakdowns on a worn folding table. In 2026, these tasks are formal routines handled by automated tools and summarising software. Think Office Space reports, except now the printer works and has taken your job.

You can call it cruel, but it isn’t accidental. This isn’t an error caused by AI; it’s how the system was designed. The aim was clear: reduce costs by automating repeatable tasks. Entry-level creators, by nature, do straightforward work—this is exactly what an optimisation tool can focus on with clear metrics. It is efficient headcount management.

The ending isn’t sudden destruction. It is a slow, deliberate process carried out by executives making logical economic choices in a system that encourages individual gain over teamwork. The people making these decisions aren’t likely to stop—their incentives drive them to continue.

The cost shows up later. A few years on, when teams can’t find mid-level staff because none were trained, when the talent pool is empty, someone will argue for rebuilding the system.

By then, the ladder is gone, and the factory that built ladders has been turned into flats.