Why AI is a nothing burger — it’s the writing stupid
Stop me if you’ve heard this one before — a new technology that will enable thousands of hours of quality entertainment to be on demand at all times. In January 2007 Netflix launched “Watch Now,” its first internet video streaming feature.
Now almost 20 years later “streaming” of premier shows has mostly settled on a weekly release format:
Data from entertainment research firm Luminate (covering the first half of 2026 across major U.S. streaming platforms like Apple TV+, Disney+, HBO Max, Hulu, Paramount+, Peacock, and Prime Video) shows the overall split for TV show releases:
Weekly Releases: 61% of TV series releases.
Binge (All-at-Once) Releases: 38% of TV series releases.
And even when binge is available the number of quality shows that a “streaming platform” (was called a “network” back in the day) produces is inline with the same seasonal cadence we’ve always had.
In short “streaming” is a nothing burger. Lots of elements of the entertainment business were able to scale up — the platform, the special effects, the editing, the directing, and even the acting. On the acting the number of quality actors available is so staggering that sometimes one is just reduced to saying, “Oh look another academy award level performance from someone I’ve never heard of before.”
What wasn’t able to scale up is the writing.
Netflix is scaling back on traditional original movies and scripted series output (with 2026 seeing some of its lowest movie output since 2017) while pivoting heavily into sports, live events, reality TV, and gaming.
There is no factory or school that can output quality writing at scale. Even if you are willing to throw a lot of money at aspiring writers, which Netflix is vehemently not willing to do, you are not guaranteed good writing.
Now AI has reduced coding, science, and mathematics to the problem of writing. Literally so, here is just one example of the markdown that I am maintaining to tell AI what to do.
Not only that but this must be good writing. If the writing is not concise enough it burns tokens on every model call (all of your money would be wasted on AI processing bad writing over and over, less caching). If the writing is not innovative enough neither will the result.
If you ask AI why agents are not good at writing it will tell you (in a paragraph that I almost missed because it was buried in drivel):
Until agent frameworks move away from additive committee loops and incorporate specialized subtractive pruning, taste-based selection, and tolerance for sharp viewpoints, AI agents will remain stuck generating generic drafts that human writers have to rewrite from scratch.
Science runs on good grant proposals and an AI that can’t write one will leave that bottleneck in place. The creators of AI, just like the creators of streaming, could invest in making AI capable of good writing but there are no guarantees that investment will pay.
Like Netflix, LLM companies best bet is to become an essential component in providing services that are mostly identical to what already exists. That’s true of non LLM AI like Waymo also. They aim to replace Uber which replaced taxis, but in the end you’ve still just hired a vehicle to drive you point A to point B through heavy traffic.
This strategy is so common it has its own term, tech-enabled. Tech-enabled is how Fast Company ranked Adidas №7 in the World’s Top 50 Most Innovative Companies. Even if Adidas had some incredible innovation in sports shoes, which I really doubt, everyone you play against would be wearing the same shoes and so nothing burger.
Pit successive generations of these tech-enabled nothing burger innovations against each other and you have a classic red queen’s race.
Of course the possibility of a red queen’s race already occurred to the Nation Institutes of Health and so on July 17, 2025:
NIH will not consider applications that are either substantially developed by AI, or contain sections substantially developed by AI, to be original ideas of applicants. If the detection of AI is identified post award, NIH may refer the matter to the Office of Research Integrity to determine whether there is research misconduct while simultaneously taking enforcement actions including but not limited to disallowing costs, withholding future awards, wholly or in part suspending the grant, and possible termination.
Still even if AI is used mostly for tech enabled services, some tech enabled services, like AirBnB, reduced the prices we pay to where some people were able to afford things they previously could not. For instance two years ago in October of 2024, Gartner predicted:
Through 2026, 20% of organizations will use AI to flatten their organizational structure, eliminating more than half of current middle management positions.
The problem here is eliminating half of current middle management positions has been possible for a long time. Just like remote working the technical barriers to flatter organizations have long been solved — already most middle managers are meeting jockeys of useless meetings.
You can say well AI will summarize that meeting for you or AI will evaluate worker productivity for you, but most of those meetings and evaluations were already fake.
And so the great flattening in reality is barely meriting a Wikipedia entry:
Between May 2022 and May 2025, manager headcount at public companies fell 6.1%, while executive roles fell 4.6% (CNBC, 2025)
It might not be exactly writing that keeps middle managers in place, but its something AI can’t do well like maybe networking or bullying.
However that’s with the current one size fits all, mostly generic, models we have today. The possibility of an AI that can do as well as a middle manager at networking and bullying is very real.
Everyone is already talking to an AI agent so the networking is automatic and an LLM trained on the internet must have some talent for bullying. Plus an AI boss doesn’t try to be part of an all boy’s club or ship jobs to a cheaper country; its only agenda is replacing humans with more AI.
That’s the case for AI replacing layers of a hierarchy. The non nothing burger version gets rid of hierarchy altogether.
At Block, we’re questioning the underlying assumption: that organizations have to be hierarchically organized with humans as the coordination mechanism. Instead, we intend to replace what the hierarchy does. Most companies using AI today are giving everyone a copilot, which makes the existing structure work slightly better without changing it. We’re after something different: a company built as an intelligence (or mini-AGI).
That copilot point is pretty sharp and sounds like Jack wrote this himself and does not have access to an AI that can write. Can AI serve as a coordination mechanism without being a good writer? Unclear.