Video editing is following the same path AI coding took in 2025 and 2026, Riley Brown wrote in an X post on 16 November: “The 10x–100x programmers don’t write code anymore. They build software factories.” The people who understand video, brand and storytelling, he says, “will soon be building content factories.” The post had 54,662 views when we read it on 17 November.
Our number is on the agent side of his comparison. On OpenRouter, weekly agent token usage is up 106x since the first week of January and human usage 7x, per OpenRouter token data. Over about 45 weeks, that is roughly 11% growth a week, compounded, for agents and about 4% for humans, by our arithmetic.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on why products built on the same model differ. It teaches the general idea only and does not discuss Riley Brown or video editing.
The key insight: As we read it, Brown’s claim rests on one shift that is already measured on the coding side: on OpenRouter, agent token use grew 106x this year against 7x for human use, and inside Anthropic about 90% of the Claude Code team’s code is written and reviewed by Claude, the company says. For video there is no comparable measure yet.
What Riley Brown Wrote
In a linked X article titled “Content Factories and the Rise of Agentic Editing”, Brown says thousands of creators “have stopped editing by hand and started directing AI to edit for them.” He wrote: “In 18 months, editing a video manually will feel totally insane, the same way writing code by hand already does.”
He describes himself as an early, non-developer user of Claude Code in 2025. His comparison: “I believe Premiere Pro is the IDE for video editors / creators.” Developers still open an IDE “but only when they have to”, he says, and he expects editors to spend most of their time talking to an AI chatbot, “likely in Claude Code or Codex.”
His technical point is that a timeline is already text. “Under the hood, a Premiere Pro timeline isn’t really a video. It’s a text file that references videos”, he wrote. Trimming a clip, in his telling, changes two timestamps; a cross dissolve adds one short line.

What Was Posted on X
Posted by Riley Brown at 20:32 UTC on 16 November 2026, this post says video editing is following the path AI coding took in 2025-26 and that people who understand video, brand and storytelling will build content factories. It links his X article, Content Factories and the Rise of Agentic Editing, which the piece quotes; the OpenRouter, Anthropic and Google figures come from our earlier pieces.
The Agent Path, in Our Data
The software factory Brown describes is one where a developer stops writing code and designs the system that writes it. The closest measured case we have covered is Anthropic’s own: about 90% of the Claude Code team’s code is written and reviewed by Claude, the company says, and on 16 November it published the verify loop the team uses to check that work.
Anthropic has also given the volume. In Anthropic’s monorepo, changes landing per day went from about 2,500 in May to more than 30,000 in September, Anthropic says, 12 times in about four months, by our arithmetic.
The OpenRouter series measures agent use across all its traffic, not coding alone. Agent tokens are counted from API keys that behave like software, using seven signals such as tool-call rate, OpenRouter’s recap says. At about 11% a week for agents against about 4% for humans, by our arithmetic, the agent line compounds almost three times faster.

The Video Side: What Models Cost
On generation, prices are already set per second. Google cut the Gemini Omni video API price by 30%, to $0.07 a second for 720p video, on 5 November. A 10-minute 720p clip would cost $42 at that list price, by our arithmetic, and Google Vids lets anyone with a Google account make up to three 15-second videos a month at no cost, Google said.
Brown’s thesis is about editing real footage rather than generating it. His example is an agent asked to look at the day’s camera footage and make a vlog: Claude will work “for 30 minutes, or maybe even 3 hours”, he wrote, to hand back a coherent video, “(sometimes)”.
He also sets a human limit. “People want to watch other people, so a content factory always has a human at its center”, he wrote, and if the humans at the center bring no value, “it amplifies to zero”.
The Structural Read
As we read his article, the factory Brown describes has two parts: agents that do the cutting, and a person who sets the rules. He wrote that the creative director of the future starts by understanding “the whole story of the creator or company they’re growing”.
As we read the price data, generation is already priced by the second: $0.07 a second for Gemini Omni 1.2 Flash at 720p, Google said, while Brown’s factory is mostly about editing footage a person shot.
Riley Brown, Content Factories and the Rise of Agentic Editing, X article, 16 November 2026
“A software factory is what happens when a developer stops writing code and starts designing the system that writes it.”
Three Implications
FOR CREATORS Brown says a content factory turns one story into long-form videos, shorts, podcasts, newsletters and posts at once, with a human at its center.
FOR EDITING SOFTWARE Brown wrote that editors will open the Premiere Pro or DaVinci Resolve timeline when something goes wrong, the way developers still open an IDE “only when they have to”.
RELATED Our piece on Gemini Omni 1.2 Flash reads Google’s 30% video price cut to $0.07 a second.
The Business Engineer Lens
This story maps onto the Business Engineer framework AI & The Harness Theory.
The framework argues that the next professional jump is not a better prompt: “It is an operating system that wraps your AI — a harness that compresses your thinking, your team’s feedback, and the model’s raw capability into something replicable.”
As we read it, Brown’s content factory is that harness applied to video: the rules a creative director writes down are what make the agents’ edits repeatable. The framework predates his article and does not discuss video editing or Riley Brown.
What Is Not Established
Brown’s timing is his own forecast. He says not much will change in the next three months and that an inflection point comes “sometime in the next 12 months”. He gives no data on how many creators edit with agents today, and the view count he cites for a friend’s AI-edited video is his projection.
The OpenRouter series counts all agent traffic on one router, not coding alone, and not video work. We have no comparable measure of agentic video editing. We read Brown’s post and article in full and did not contact him for this piece.
The Bottom Line
Riley Brown says video editing is on the path AI coding took, from hand work to factories run by agents. On OpenRouter, where agent tokens count all agent traffic, agent token use rose 106x this year, about 11% a week compounded by our arithmetic, against 7x for human use.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Riley Brown’s thesis and quotations are from his X post and the X article it links, both of 16 November 2026, read in full on 17 November. The Anthropic, OpenRouter and Google figures are from our earlier pieces, linked, which read those companies’ own posts and data. The weekly compounded rates, the 45-week span, the 10-minute clip cost and the monorepo multiple are our arithmetic. Riley Brown was not contacted for this piece.
Sources: Riley Brown on X (16 Nov 2026) · Riley Brown: Content Factories and the Rise of Agentic Editing (X article, 16 Nov 2026) · FourWeekMBA: OpenRouter agent tokens 106x (17 Nov 2026) · FourWeekMBA: Claude writes 90% of Claude Code (16 Nov 2026) · FourWeekMBA: Gemini Omni at $0.07 a second (5 Nov 2026)









