Summer 2026 will be remembered as the moment YouTube creators definitively broke through to Hollywood. Three films — Backrooms ($115M worldwide on a ~$10M budget), Obsession ($104M domestic on a ~$1M budget), and Iron Lung ($51M worldwide on a $3M budget) — dominated the box office, eclipsing legacy franchise entries like The Mandalorian & Grogu.
The most revealing comparison is Ryan's World — a massive kids' YouTube channel whose movie grossed only $624,000 on a ~$10M budget. Simply plugging an influencer into a movie doesn't work. What made the difference this summer was that the creators behind Backrooms, Obsession, and Iron Lung weren't just popular — they were genuinely talented full-stack filmmakers.
Curry Barker ran a YouTube sketch channel called "That's a Bad Idea" where he learned to write, act, and edit with a tight audience feedback loop. Kane Parsons built his Backrooms series in Blender and After Effects — free, open-source tools. Markiplier literally installed server racks in his bathroom to render his own VFX shots, bypassing the turnaround time of a professional studio.
Ben Thompson, who predicted this moment in his 2017 essay "Goodbye Gatekeepers," draws a compelling parallel: YouTube is to Hollywood what AWS was to venture capital. Cloud computing reduced the cost of starting a company to nearly zero, creating an entirely new asset class where VCs evaluate startups on actual products and market signals, not PowerPoint pitch decks. YouTube does the same for filmmaking — a creator's channel is their "audition tape," proving both audience appeal and artistic capability before a studio ever writes a check.
"On the Internet, distribution is free. Anyone can set up a website or start a YouTube account. Content that breaks through has to do so on its own merits. Is it good and compelling or is it not? If it is, it gets more and more views. That's what the algorithm does." — Ben Thompson, Stratechery, June 2026
Lucas's initial claim to fame was American Graffiti, which made over $200 million on a $777,000 budget — a 200x+ return. That success gave him the credibility to make Star Wars. Today's YouTubers are following the exact same playbook: prove yourself with a low-budget hit, then get the blank check. Curry Barker has already signed for the next Texas Chainsaw Massacre with A24.
That all three breakout hits are horror films isn't a coincidence. Horror is the one genre where a low budget and unknown cast are features, not bugs. Fear is universal — it crosses language and culture barriers in a way comedy never can. You don't need Tom Cruise when the audience is supposed to believe anyone could die.
Bernie Su, a three-time Emmy-winning creator (including the first Primetime Emmy wins for YouTube and Twitch shows), put it bluntly: "Look at any horror franchise. Final Destination — seven movies, no household names. Scream is the exception, and Scream built its own stars."
On the horizon: Wesley Wang's "Nothing Except Everything" (TriStar, with Darren Aronofsky producing) and the surreal Skibidi Toilet adaptation (Michael Bay reportedly attached). And eventually, someone will make a film where 90% of the budget is AI compute — a movie that couldn't exist without the technology, not a movie that uses AI to replace human labor.
Bernie Su's current project at Pickford hints at what's coming: a detective series where the AI writes the show as it goes, live-prompting audiences in theaters on their phones. The audience's choices collectively shape the narrative in real time. It's a form of storytelling that is literally impossible without AI — and it doesn't replace human creativity, it extends it into new dimensions. As Su puts it: "AI is not going to Game of Thrones us Game of Thrones. But on customization — making the audience feel consequential — that's where AI's speed and optimization become the killer feature."
Senator Bernie Sanders published a New York Times op-ed proposing the American AI Sovereign Wealth Fund Act: a one-time 50% tax on AI labs, paid directly in stock, creating a government-controlled fund with voting rights, board seats, and dividends distributed to American citizens.
"AI is built on humanity's collective knowledge. The wealth it generates must benefit humanity, not just Elon Musk, Sam Altman, or other AI oligarchs." — Senator Bernie Sanders, New York Times Op-Ed
The proposal raises more questions than it answers. When would these companies actually pay dividends? Tech companies are notorious for hoarding cash for decades — Amazon took over 20 years to become profitable. Which companies count as "AI labs"? Is Salesforce an AI company? Meta? Walmart, with its AI shopping assistant? And perhaps most critically, how does a proposal to own AI companies square with Sanders' simultaneous push to ban AI data centers?
Dean Ball crystallized the contradiction: "Is AI an existential risk that needs to be banned or a public good that should be redistributed? He wants to have it both ways."
Amid the narrative that AI would commoditize enterprise software, Salesforce delivered a resounding rebuttal: fiscal year 2027 Q1 revenue of $11.13 billion, up 13% year over year, with $6.7 billion in operating cash flow. Agent Force crossed $1 billion in ARR, and combined with Data Cloud and MuleSoft, the AI+data business reached $3.4 billion.
Marc Benioff was explicit about how AI is changing his workforce: "I'm not hiring more engineers. I'm using coding agents to give me the extra capacity. I didn't hire more service agents — I held it flat and reduced it. But I did hire almost 20% more sales people because we have more demand than ever."
The asymmetry is notable: Benioff cuts engineering and service headcount (areas where his own products don't compete), but hires more salespeople (he sells sales-focused AI agents). It's not hypocrisy — it's a real signal that agentic AI is creating genuine demand that requires human sales capacity to fulfill.
Anthropic has confidentially filed an S-1 registration statement, joining the race with SpaceX (already public) and OpenAI (rumored). Meanwhile, SpaceX's IPO saw index providers waive long-standing rules: the S&P 500's requirement of 12 months of trading and 4 quarters of GAAP profitability — in place since 2002 — were both waived. The seasoning window was cut from 90 days to just 5. The result: over $30 trillion in passive 401(k) and retirement money is structurally forced to buy SpaceX at IPO valuations before the lockup period ends. As Nic Carter observed drily: "So this is what kills index funds."
The debate over AI-driven job displacement reached a surreal peak this week. Apollo's chief economist Torsten Sløk declared there is "zero evidence of AI-related job losses" — ADP employment data remains strong, and the AI spending boom is stoking both employment and inflation. At the other extreme, real estate investor Jason Oppenheim claimed on a podcast that AI will lead to "80% of people out of the workforce" — more than three times the peak unemployment of the Great Depression (24.9%). When asked about his credibility on the topic, he explained: "I've spent probably 500 hours listening and reading a lot of podcasts."
The reality is likely between these extremes. What we're seeing is not mass layoffs but selective headcount freezes: Benioff not hiring engineers, Sirus AI's banking customer not hiring 5 new people, Default's customers using the product to avoid hiring ops teams. Surplus productivity gets channeled into more output, not unemployment. It's Jevons Paradox in action — cheaper technology creates more demand for the output, which creates more demand for complementary human roles (like Benioff's salespeople).
Adam Iscoe, a journalist formerly at The New Yorker and now at The New York Times, spent months reporting on prediction markets — a world where a third of American adult voters have now visited a platform like Kalshi or Polymarket, and where a small group of individual "sharps" routinely out-trade Wall Street institutions.
The market is brutally efficient. Iscoe spoke with a grad student trader operating under the pseudonym "Frozen," who turned $200 into half a million dollars:
"I really am just taking money from people. Every dollar I gain is someone else losing. And there's a lot of people joining and betting, and losing, and leaving. And then there's a group of a couple hundred guys winning, and that's the whole story." — "Frozen," prediction market sharp, to Adam Iscoe
Even institutional players are struggling. Jeff Yass's Susquehanna International Group (SIG) — one of the most respected trading firms in the world — told Iscoe they're "getting taken for a ride by these sharps." The reason: individual traders can web-scrape, build niche models, and operate in ways that institutional compliance departments can't allow. One Rotten Tomatoes trader turned down a job offer from SIG because they couldn't continue their scraping-based strategy under the firm's corporate policies. He made seven figures independently.
The regulatory landscape is fragile. The current CFTC (under the Trump administration) has been "incredibly favorable" to prediction markets, but a change in administration in 2028 or 2029 could shift everything. Meanwhile, national security concerns loom: a special forces operator traded on a "strike on Venezuela" contract — effectively, a classified operation being announced on the internet through betting activity.
After 15 years scaling Meta's infrastructure — data centers, Oculus, Instagram, Facebook AI Research — Mike Schroepfer left to start Gigascale Capital, a $250 million venture fund built on a single thesis: the marginal cost of software went to zero, so the constraint is now physical stuff.
"The marginal cost of software went to zero. The thing that really mattered is how much stuff can we build and how are we gonna build it." — Mike Schroepfer, founder of Gigascale Capital
His portfolio reads like a blueprint for the next industrial revolution:
Schroepfer draws a parallel to the electrification of factories. When electricity first replaced steam, factory owners simply swapped the power source but kept everything else the same. Productivity gains didn't materialize until they rebuilt factories from scratch around the new constraint — putting an electric motor at every workstation. The same thing is happening in robotics: dropping a humanoid robot into a factory designed for humans yields marginal gains. Redesigning the factory with the goal of "keep the robots 80% utilized" unlocks 5-10x improvements.
Solar panels have gotten 99% cheaper, but they're still deployed the old way: heavy steel frames, manual installation, treating the panel like a precious mineral. "The panel is now the cheapest thing on the field," Schroepfer notes. "The steel frame costs more." The opportunity isn't making cheaper panels — China already owns that. It's rethinking deployment: form factor, automation, installation. Assume the glass is free and redesign everything else.
Danial Jameel's startup builds AI workflow agents for the back offices of America's ~8,000 banks and credit unions — the institutions serving the "lumber mill in Mississippi, the young family in Arkansas getting their first mortgage, the farmer in Illinois." His Series A: $28.8 million, led by a16z.
The impact is tangible. One customer had a team spending 5pm to 9:30pm, five days a week, clearing a backlog of 300 loan documents — only to face a new backlog at 9am. Three weeks after deploying Sirus, the same team goes home at 6:20pm. They didn't fire anyone. They didn't even avoid hiring — they had planned to hire 5 people, realized they didn't need to, and gave the existing team raises instead. Jameel calls this "the dignity of work."
"They're going home at 6:20. They can be with their families. That's dignity of work." — Danial Jameel, founder & CEO of Sirus AI
Nico Ferreyra's Default ($10M Series A) attacks a problem nearly every B2B company faces: the "Frankenstack" of tools that don't talk to each other. His diagnosis is simple: "All growth problems are data problems, at least in go-to-market." Default provides a real-time data layer that plugs into CRMs and translates all the messy signals — website visits, form submissions, meeting notes — into a language AI agents can understand.
Ferreyra's one-liner resonates universally: "I've never met someone who didn't hate their CRM." The best companies — Ramp, Rippling, Owner.com — treat distribution as a product, building their own internal platforms early. Default aims to give everyone else that capability out of the box.
Sue Khim, co-founder and CEO of Brilliant, launched an AI tutor named "Koji" — a graphical tutor that can see how you interact with problems and point, sketch, and annotate on screen alongside you. The design philosophy is deliberately counter to the chatbot paradigm:
"We designed this tutor for that moment of understanding, not for the moment of explanation." — Sue Khim, co-founder & CEO of Brilliant
Research on human tutoring shows that when tutors just explain, students learn less. Better outcomes come when the student does the work and the tutor facilitates. Chatbots are optimized for explanation — ask a question, get an answer, leave. They receive no signal on whether learning actually happened. Brilliant's approach is the opposite: dense real-time interaction data, a purpose-built learning graph, and a design goal that mirrors dating apps. "If we're successful," Khim says, "the tutor becomes unnecessary. You scaffold the learner until they're asking the questions the tutor used to ask."
In retrospect, this episode of TBPN was about one thing: gatekeepers losing their power, across multiple domains simultaneously.
Hollywood's gatekeepers — studio executives who believed they could "shove whatever supply they wanted through the tubes" — are watching three YouTube-native films outperform their franchise entries. The lesson isn't that influencers can sell tickets. It's that YouTube's algorithm is a more efficient talent filter than any human gatekeeper. The people who break through on a platform with zero quality bar face a far higher bar than any studio exec ever set.
AI's gatekeepers — the question of who gets to own and control artificial intelligence — is being contested on multiple fronts. Sanders wants the public to own 50%. The labs want to IPO. Index providers are rewriting rules to funnel passive money into the biggest names. The ownership structure of the technology that may define the century is being decided right now, largely through op-eds and S-1 filings.
Financial market gatekeepers — the traders and institutions who once had an information edge — are being outcompeted by individuals on $600 laptops who can scrape websites and build niche models that compliance departments won't allow. A third of American voters now participate in markets that were illegal in most states just a few years ago.
Physical world gatekeepers — the assumption that manufacturing, energy, and construction were "solved" industries with no room for disruption — is being challenged by a wave of talent from SpaceX and Tesla who are applying software-era thinking to century-old industrial problems. The marginal cost of software went to zero, and the stuff we build with it is about to get a lot better.
The old gatekeepers didn't disappear this week. But 2026 is the year they started looking over their shoulders.