Welcome to Product Cocktail, where the takes are as polarizing as a shot of Fernet—but the insights come together like a perfectly crafted daiquiri.
The Shake
INT. APARTMENT - DAY
The CAMERA TILTS UP from a woman's waist to her face as she turns for the door, talking over her shoulder.
WOMAN
You know why I broke up with you?
Because you're a pathetic loser
with no meaning in your life.
She SLAMS the door behind her.
CUT TO:
INT. APARTMENT - GAMING DESK - CONTINUOUS
A gaming rig glows against the window. Push in on the center monitor.
INSERT - THE MONITOR
INTERGALACTIC BATTLE
You have been chosen as the representative
of your planet in the intergalactic battles.
Do you accept this mission?
The choice cannot be undone.
A timer ticks down from 0:06. The YES button
sits highlighted.
CUT TO:
INT. APARTMENT - FLOOR - CONTINUOUS
Our HERO still laying flat on his back on the carpet after taking a punch from the woman's new boyfriend.
INSERT - THE MONITOR
INTERGALACTIC BATTLE
You have been chosen as the representative
of your planet in the intergalactic battles.
Do you accept this mission?
The choice cannot be undone.
A timer ticks down from 0:06. The YES button
sits highlighted.
CUT TO:
INT. APARTMENT - FLOOR - CONTINUOUS
Our HERO still laying flat on his back on the carpet after taking a punch from the woman's new boyfriend.
HERO
I don't need it, b-tch.This is ARENA ZERO, the first "original" series from Higgsfield AI. Spoiler alert: it didn't get any better after the first two minutes.
Our main character mashes the enter button and gets sucked up into a vortex out the top of his apartment building. He's woken up by a shrill, anthropomorphic bat in an alien Colosseum with a jeering extraterrestrial crowd, unwittingly entered into a gladiatorial competition emceed by a cross between RuPaul, Caesar Flickerman, and The Grandmaster from Thor: Ragnarok.
Your host for the evening: what you get when you prompt "flamboyant games-master" and take the first render. RuPaul, Caesar Flickerman, and the Grandmaster walk into a latent space. (Source: Arena Zero, Higgsfield AI)
Within the next eight or so minutes, I'm subject to: cartoon exposition from the aforementioned bat, an AI-generated answer to "what if Lucilla from Gladiator was a Targaryen," and crowd reactions from extras resembling Voldemort, Rango, and Groot-as-a-Chia-pet.
After 10 minutes of my life I'll never get back, I sought refuge in the only other piece of content I could find on "Higgsfield Originals" longer than a trailer: HELL GRIND. To sum up my reaction from watching all 22 minutes, I'll leave you with a quote from its own dialogue: "NO CAP THIS IS GIVING ME VERY BAD VIBES!"
The biggest indictment: it's been nearly four months since Arena Zero, episode 1 dropped and we still don't have a second episode.
Who wore it best? The Arena Zero crowd, brought to you by Voldemort, Rango, and Groot's Chia Pet cousin. (Source: Higgsfield AI)
Welcome to the "first complete AI streaming platform," baby! Or is it simply a tech demo masquerading as the next HBO?
A storefront, not a studio
While their content might not be dethroning 'Thrones anytime soon, Higgsfield AI is no joke.
And as far as a tech demo, it's an impressive one. The Arena Zero and Hell Grind productions are technical achievements in long form video generation even with their rough edges. Higgsfield AI is a powerful, model-agnostic harness with legitimate pre-processing chops and software orchestration that solves hard problems like character consistency from shot to shot.
Per Stripe, Higgsfield exceeded a $200M run rate in 9 months and they're on pace to reach a $1B annual run rate by the end of the year. They raised an $80M Series A extension in January 2026, bringing their total funding to $138M, and notching a $1.3B valuation.
By Higgsfield CEO Alex Mashrabov's own admission at Cannes (they didn't get invited, they showed up and hosted a screening), it's a tech demo, not a Hollywood competitor: “We are not here to compete with anyone. We are here to showcase the capabilities... also about some limitations.”
Higgsfield isn't a studio. It's a PLG subscription funnel dressed up as a streaming platform. A business that sells you generations and rates your output can't optimize for taste.
Creating an account following a misleading "Sign in to see next episode" CTA led to an orgy of monetization levers. A barrage of competing sign up upsell messaging ("SIGN UP AND GET ADDITIONAL DISCOUNT ON PREMIUM PLANS [EXTRA DISCOUNT]" "Seedance 2.0 4K Available on Higgsfield. [Get with 30% OFF]" "Pricing [30% OFF]"). A personalized email drip campaign focused on "expiring" discounts and "loopholes" to get them back. Obtuse, subscription-gated token pricing.
Click "Sign in to see the next episode" and this is what's waiting — an orgy of expiring discounts. The "streaming platform" experience. (Source: Higgsfield AI)
The "first complete AI streaming platform" from their March PR is a nice line for their Growth team to sell to prospective customers, but they haven't delivered on it. It's a classic cold-start problem: their content pipeline is weak (a couple of marquee bets longer than a minute or two, a pile of mediocre trailers), and their audience traction is worse (micro-influencer view counts on nearly everything).
They're also selling a cost-savings dream that's more theater than what their in-house creators have produced: "According to Higgsfield, the 95-minute feature [Hell Grind] cost under $500,000... [CEO] Mashrabov said a comparable traditional action-fantasy feature could cost around $50m." Based on the 22 minutes I saw, they would have been better off lighting the $500K and smoking it.
The House always wins
The problem with Higgsfield AI trying to play the content game is that it's engineered like an AI video generation casino, not a studio.
There are a million different models, features, and apps to try, all showcasing the ways you can burn tokens attempting to replicate their tech demos. They even held a "Make Your Own Action Scene" contest in February with a $500K purse, no purchase necessary — aside from the money you put on generations.
Higgsfield prominently features a "viral potential predictor" app that will score your videos on "Viral Potential," "Hook Score," and "Hold Rate" based on the "simulated cortical response of 720 modeled brains." Gross.
The Virality Predictor grades your video against the "simulated cortical response of 720 modeled brains." A greenlight executive, this is not. (Source: Higgsfield AI)
Their CEO even described trying to achieve the complexity of Hell Grind as "a feeling of a slot machine" with some shots requiring hundreds of iterations to nail down spatial logic, character placement, and action work. Hundreds of iterations per shot. It's tokenmaxxing with a director's chair.
The platform gamifies spend, not quality. Gambling mechanics and editorial trust are inherently misaligned. How can I trust what you "greenlight" when your P&L just wants me to become a convert-to-paid slop artilleryman?
One customer put it plainly: "I already spent all my credits on short, bad video." They won't renew until Higgsfield releases a tutorial explaining how the marquee series were actually made. Their product is sold as "pay $19-$129/mo and make films like the Originals" but it runs as a slot machine that eats half a mil and still produces broken physics. It's a false promise of democratization.
Even if we played this out and Higgsfield created some kind of internal-IMDb audience score, the people watching this are the AI-pilled proletariat masses, not exactly the Letterboxd crowd. Any crowdsourced rating would sink to the lowest common denominator. Engagement-driven-discovery begets a flattening of quality.
Ironically, Amazon just did the same thing at a larger scale. In May, Amazon MGM Studios and AWS launched a "GenAI Creators Fund" to fund projects using AWS-based video generation tools (Project Nara) to create content to (cheaply) fill Prime Video's grid on the cheap. (I pity the fools in Accounting who have to record those internal company transactions.) This reads like just another glorified tech demo customer acquisition play dressed up as a "Creator Fund." In fairness, one of the directors they greenlit won an Emmy, so I'll hold full judgement here until I see the work product.
Missing from the frame
Technically, nothing. Arena Zero has a director — a real, festival-selected one. Hell Grind had a fifteen-person crew and a script years in the works. The director even went on record:
"Technology won't replace talent. It rewards those who truly understand storytelling, rhythm, and emotion, and reveals those who don't."
So what's missing? He's not missing from the credits. He's missing from the frame — and he did replace the on-screen talent — with the likeness of Higgsfield's recent grad growth staffer.
I was sitting in my BU dorm during fall of freshman year the moment I saw TV used to its full potential as a medium. The season 1 finale of Mad Men was airing.
Don Draper's Carousel pitch: a gorgeous lie about a happy family, and roughly the best two minutes of restraint ever put on TV. (Source: Mad Men, "The Wheel," AMC, 2007)
The climax of the season delivered a gut punch that emerges subtly from a poignant moment. Don Draper pitches his campaign for the Kodak slide projector — describing it as a time machine that takes us to a place where we ache to go again, while tapping through old photos of his family — against a tableau set up throughout the season of a home broken by infidelity.
This device... isn't a space ship. It's a time machine. It goes backwards, forwards. It takes us to a place where we ache to go again. It's not called the Wheel. It's called the Carousel. It lets us travel the way a child travels. Around and around and back home again... to a place where we know we are loved.
Showrunner Matt Weiner doesn't need to tell you that Don is hollow, he lets the fiction he's selling to Kodak carry the message. It's one of the best moments I've ever seen on screen, and it's an expertly crafted exercise in artistic restraint.
Fast forward to May 2025, late in the final season of Andor, Disney's gripping 70s political thriller set in the Star Wars universe, sans light swords and space wizards. The apex of the eighth episode is a tense build up to Imperial forces perpetrating a horrific massacre of peaceful protestors on planet Ghorman.
The Ghorman Massacre: two football fields of practical set, and the score cuts to silence the moment the shooting starts. (Source: Andor, S2E8 "Who Are You?", Disney+)
It's shot like a war documentary — close, chaotic — but Director Janus Metz withheld specific aspects to mount the tension. The musical score drops the moment a riot trooper opens fire, letting the violence unfold without offering the audience an auditory cue to process their emotions.
By this point, the audience knows this false flag attack is going to happen through the suspense built throughout the season, and sits in their powerlessness watching the horror play out on screen. Andor built tension up across two seasons to this point and refused an easy or cheesy payoff.
Neither of the Higgsfield AI marquee shows exercise restraint. It's an emphasis on special effects, and giant (ill-conceived) action set pieces, offering you an all-you-can-eat buffet that leaves you feeling sick after — a stark contrast to the Michelin tasting menu of small perfected bites that shows like Mad Men and Andor deliver.
The tool isn’t the villain
To be clear, I'm not taking a hardline approach that all AI in film is bad. That reads as a high-minded and pretentious take that becomes a slippery slope. If you're drawing a hard line against using AI (a tool), why not CGI or digital photography or motion capture?
I'm very much in the "AI is a tool" camp. Tools are not the enemy, replacing the decision layer is. Ever since the advent of CGI, there's been a choice of practical vs. digital for special effects.
The new spectrum is reality to generated. On one side, full practical: depicting the Ghorman Massacre in Andor demanded a "full-scale build spanning two football fields." Showrunner Tony Gilroy had access to Industrial Light and Magic's (ILM) latest technology, and he chose to build Ghorman on a backlot at Pinewood Studios, because that was the best way to make his vision real.
In the middle, hybrid digital and practical: Mandalorian extensively used ILM's StageCraft technology ("The Volume" — a giant LED screen, combined with real lighting and physical set pieces) so Pedro Pascal would have something to react to in lieu of a green screen. Technology that served a vision.
ILM's StageCraft "Volume" — a wall of LED that throws real light on real actors. Technology in service of a vision, not in place of one. (Source: The Mandalorian, Lucasfilm)
On the other extreme, fully generated: Higgsfield AI and Arena Zero. No set, no lighting, no actors. Four prompt engineers playing a roulette wheel.
The question to ask is whether the tool amplifies the director or replaces them.
The RED camera and digital democratized filmmaking because they put a better tool in a director's hands. Higgsfield isn't built to serve a vision, it's built to sell generations.
Poisoning the well
If this was just about a plucky director misguidedly yeeting half a mil of other people's money to make a piece of shit movie, I wouldn't be writing about it. That story has been told before. James Cameron spent 2000x that to make three Avatar films.
Higgsfield launched a "streaming platform" that's a funnel, "Originals" that are demos, a "Cannes premiere" that's a rented room, and a "director" running a slot machine. Their approach is built on borrowed legitimacy in an industry where authenticity and trust are paramount.
The consequence: it poisons the well. Arena Zero and Hell Grind are now data points in the cynical "all AI is slop" story. There are legitimate use cases for this tech, and maybe eventually legitimate cases in film and TV production, but the AI industry is starting on its back foot now.
As I've covered before: the vibes are off with AI. When consumers are getting a median bad first impression of AI in general, as someone who believes in the technology, it's tough to see a flashy launch of an overpromised, underdelivered "first complete AI streaming platform."
The Recipe
You Can’t A/B Test Your Way to Mad Men
Too much: Higgsfield AI has a virality predictor where a greenlight executive should sit. They are attempting to optimize away a person with a point of view and the authority to say no. That was the product.
Not enough: Restraint. Realizing that great TV is a rare category where the outcome doesn't decompose into a metric. Bold decisions are required, not thousands of generations that devolve into a Tinder-esque search for something a little better.
What’s the fix? Hire Casey Bloys (Chairman & CEO, HBO and Max Content), or someone similar. The fix isn't a better dashboard, it's a chair with a human of taste in it who can hire and protect directors of vision and greenlight "The Pitt" that builds a real ER.
The Garnish
200 Locations.
Doug Liman’s film used “Gray Room” technology (filming in literal gray room, then AI-generating the set and lighting) to “travel” to 200 different locations for his upcoming film, Bitcoin: Killing Satoshi.
In his own words: “It has about 200 distinct locations, from Antarctica to Antigua to Vegas, which is obviously unproducible.” …and potentially unwatchable? Say it’s 150 minutes (extremely bloated for a journalistic conspiracy thriller), we would be in each location for 45 seconds.
I’ll withhold my judgement until the film drops, but I’m definitely skeptical.
Source: No Film School
Product Cocktail
Tip Your Bartender
Send me questions, feedback, and cocktail recipes:
[email protected].
Icons made by Icongeek26 from www.flaticon.com.