Advertising can raise prices infinitely without user churn; subscriptions cannot
Ad pricing rises with algorithmic precision without affecting retention; subscription hikes trigger churn, and once they do, it's too late. This is what Netflix should really fear as engagement slides and ad revenue misses targets.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Netflix's admission that engagement is the problem signals worse news than churn rates
Netflix laid off 6% of employees this quarter. Content head Bella Bajaria and Netflix Films chair both were removed. In past years, when asked about engagement, Netflix executives would deflect and say everything was on track. This time they directly stated "engagement is a problem." On the same call, ad business growth target was cut from 100% to 80%, and actual revenue also missed targets—engagement and ads both red, and Netflix disclosed it themselves.
— BenFlat churn rates don't mean Netflix's model is working
Netflix's recent price hikes (16% in 2025, 12% in 2026) worked because churn didn't spike visibly—Netflix's moat for a decade. But the host notes churn is a lagging indicator: once it rises, it's too late to fix. More worrisome: facing engagement questions, Netflix's only defense is "no churn yet," but can't produce other positive evidence. That alone is damning.
— BenAd revenue can scale infinitely; subscription revenue has a limit
Subscription price hikes have a ceiling—users do the math and leave at some point, Netflix's real current pressure. Ads work differently: as targeting improves, ad value rises continuously, barely affecting user experience or retention. Meta's user count peaked years ago but ad revenue still climbs. That's why Netflix, if it wants growth, must lean on ads long-term, not higher subscription fees. Yet this quarter, ad revenue missed targets.
— BenNetflix bought Warner to end the streaming wars, though not soon enough
The host thinks Hollywood should have picked two rational paths: Sony sells content to the highest bidder; Fox bets on sports and ad-supported streaming. Most others won't accept that they should sell, but can't run streaming either, so Netflix simply buys Warner Brothers—the logic being "you should all sell to us, so we'll buy you and force that outcome." This ends the chaos, but slower than Netflix can wait. It still has to show growth.
— BenInstagram Reels, not rival studios, is Netflix's real competition
When Ben talks about his wife's unshakeable 'Reels time' every night, and Andrew's daughter is a heavy Instagram user—that's deadlier than Disney or Warner competition. It's not losing one show's audience but an entire evening window of attention. Netflix's sliding engagement may not stem from content quality but from short video eating the time people used to spend on long-form. Netflix can't solve that by spending more on content.
— BenAn Indian certificate authority accidentally reduced 256-bit encryption to 4-bit
An Indian certificate authority was supposed to generate 256-bit keys but a code defect meant only a handful of key types ever got generated—effectively none. This root cert is buried in countless IoT devices that will never be patched; damage is permanent. Same week, NSA urgently warned of critical PHP flaw that lets attackers almost directly seize server rights. Only version 8.2+ (released 2022) has the patch; vast older websites still exposed.
— BenThe open internet of the past can never return
Ben says the old open internet where anyone ran their own WordPress and mail server is dead—self-hosted mail can't send out today. He borrows a metaphor from "Snow Crash": future internet will be fortified city-states, each wrapped by zero-trust networks like Tailscale, guarded by AI agents full-time, with dangerous wasteland between. He stresses nostalgia is a misreading—today's internet is already history's worst state. No going back. Only forward.
— BenAnthropic teaches AI to be good instead of locking it down
Ben distinguishes two alignment philosophies: Anthropic doesn't try to control Claude but teaches it to be "good" enough to self-regulate—a narrative that treats AI almost as conscious. Ben's own view is AI should be controlled like a tool; immense power doesn't mean uncontrollable, any more than a nuclear bomb's yield exceeds human muscle. He cites an interpretability paper proving large models aren't as opaque as feared, but the safety camp resents it because it weakens the "AI is uncontrollable" argument.
— BenIn their own words · checked verbatim
Make no mistake. Like Netflix's low churn is the key to their business. It's the superpower that they've had for 10 years.
Ben10:18
With ads, you get to raise prices infinitely with no effect on your customer base.
Ben18:27
everyone should be selling to us fine we'll just go and buy you and then you have to sell to us because we own you and that's why i called it the hollywood end game
Ben22:35
it's devouring everything and like how many people are sitting down and watching meta reels every night
Ben25:44
The internet seeded AI. It gave all the content for AI.
Ben46:28
their old view of alignment is not that they can control quad it's that they teach quad to be good
Ben1:09:08
my reading comprehension and speed, which I think often go hand in hand, are now the limit on how deep the work can go.
CK1:18:26
Figures
| Netflix staff reduction | 6% | 5:02 |
| Netflix subscription price increases | 16% in 2025, 12% in 2026 | 6:05 |
| Oracle stock YTD decline | 53% | 7:11 |
| Skydance stock YTD decline | 59% | 7:11 |
| Netflix ad growth target | 100% target, ~80% actual | 10:18 |
| Google NFL Sunday Ticket rights | $2 billion, under 1% of 2027 capex | 1:30:40 |
Glossary
- Elliptical curve cryptography
- Encryption based on mathematical properties of elliptic curves and random seeds for key generation
- Zero trust
- Security model assuming no implicit trust; every access requires re-verification
- Generative UI
- User interfaces that AI generates dynamically for each task, not static pages
How to listen
Founders and investors tracking streaming and ad monetization models; engineers handling cryptographic and AI agent security risks.
Meta VR glasses review and commentary on the film You Can See Everything—lightweight content.