Fifteen seconds. That is how long it takes Brooke Lacey to turn a raw phone video into a captioned, jump-cut TikTok. She does it without an editor, using Claude Code, FFmpeg and Whisper.
A recent episode of How I AI on Lenny's Newsletter walked through her workflow live. The editing trick will get the attention. The part I found more useful is what she does with the time it frees up.
The operation behind the clip
Brooke has been making video since before Facebook had video, and she has more than 25 years in tech. She built a TikTok audience of over 290,000 by learning to cut short-form content herself. Now she also builds her own apps, including Zosha, a free mobile app for learning AI.
Her current setup runs a weekly short-form operation with no editor. Whisper transcribes the footage, Claude Code drives FFmpeg to cut silences and add captions, and the output is ready to post. None of these tools is new. The new part is that one person can chain them without writing a production pipeline from scratch.
This is not an isolated case. Ben's Bites highlighted a designer this month who used coding agents to reformat assets for a global billboard campaign across dozens of panels in New York, Tokyo and London. Production work that used to need a team is becoming a script one person maintains.
When editing is free, ideas become the constraint
Every content team I have worked with blamed production for its low output. Not enough editors, not enough designers, too many revision rounds. That excuse is disappearing fast.
When the cut takes 15 seconds, the real question becomes painfully clear: what are you going to say in the next video? Most brands do not have an answer. Their content calendar is a list of product features and seasonal hooks, written months ahead by people who rarely talk to customers.
That is where Brooke's second habit matters. She mines her comments for new scripts. The audience tells her, in public and for free, what they are confused about, what they disagree with and what they want next.
Comments are cold data with a pulse
I have spent years arguing that marketing should start from cold data, not from opinions in a meeting room. Comments are exactly that, with one advantage over surveys: nobody asked the question, so nobody shaped the answer.
A good comment section contains three kinds of signal. Repeated questions show you where your explanation failed. Objections show you which claim people do not believe yet. Requests show you the next piece of content people will actually watch.
Most companies treat comments as a moderation problem. Delete the spam, reply to the complaints, move on. The data then disappears into a platform you do not own, and the next brief gets written from scratch.
Turning comments into a system
The workflow is not complicated. Export comments from your top-performing posts every week, across TikTok, Instagram, YouTube and LinkedIn. Let Claude or ChatGPT cluster them into questions, objections and requests, with the exact wording people used.
Then rank the clusters by frequency and commercial relevance. A question asked forty times about pricing is worth more than one clever idea from the brainstorm. Each top cluster becomes a script, ideally answering the question in the customer's own words.
Close the loop by replying to the original comments with the new video. That tells the audience their question shaped the content, which brings more questions. Over a few months, you build a content engine that runs on what customers actually want to know.
This connects to a shift I described in YouTube becoming a conversation, not a library. Platforms increasingly let viewers ask follow-up questions inside the video experience. Content built from real questions is the content that survives those follow-ups.
What one person can run now
The structural point is about team size. Brooke runs a weekly operation that, five years ago, would have needed a videographer, an editor and a social media manager. I wrote about one person running the whole stack, and this is a clean example of it in a marketing context.
There is a risk on the other side, too. When production is free, volume becomes tempting, and volume without a point of view is just noise at scale. The comment-driven approach works because it puts a filter in front of the output: you only make what someone already asked for.
That also makes the results easier to measure. A video built to answer a specific question has a clear success test. Did the question stop appearing in the comments, and did the people who asked it move closer to buying?
That does not mean every company should fire its video team. It means the value of that team moves upstream, toward judgment: which questions to answer, which angle to take, what the brand should never say. Tools handle the cut. People handle the call.
It also raises the bar. If a single creator can publish daily, answering real questions in her audience's language, a brand posting twice a month from a calendar written last quarter is not competing on the same field.
Your customers are already writing your content plan. Most companies just never read it.