“Claude cuts my TikTok”: How this creator is scaling video with AI | Brooke Lacey

I spent years in the trenches of content production, and if there is one thing that has remained consistent, it is the bottleneck of video editing. You record for ten minutes, spend three hours cutting out the dead air, and another two hours fighting with caption plugins. For most founders, this is where the wheels fall off the marketing wagon. You have the ideas, but you don't have the time to be a full-time editor.

We just saw a significant shift in how this works. Brooke Lacey recently demonstrated a workflow using Claude Code that should make every agency owner and creator rethink their stack. She took a raw mobile video and, in about 15 seconds, used an LLM-driven interface to handle jump cuts and captioning. This isn't just another flashy wrapper; it is a fundamental shift in how we manipulate media through code rather than manual timelines.

The End of the Manual Timeline

For decades, the standard for video was the non-linear editor (NLE). Whether you use Premiere, Final Cut, or even CapCut, the logic is the same: you look at a visual representation of sound waves and frames, then you manually snip. It is tedious. It is the definition of low-leverage work for a builder.

What we are seeing now is the move toward intent-based editing. Instead of dragging a cursor to 01:22 to cut a breath, you are giving a set of instructions to a terminal-based agent. Claude Code isn't looking at the video the way a human does; it is processing the data structure of the file and applying programmatic changes based on the transcript and metadata. For a founder, this means the distance between a raw thought recorded on an iPhone and a polished asset is shrinking to near-zero.

Why Claude Code Changes the Math

Most people view Claude as a chatbot. They use it to write emails or debug a snippet of React. But when you move into the CLI (Command Line Interface) environment with Claude Code, you are giving the AI permission to interact with your file system. This is the bridge that was missing.

In the demonstration, the speed was the story. 15 seconds to process a video is faster than a human can even import the footage into a traditional workspace. For those of us building in the AI space, the takeaway isn't that "AI is fast." We know that. The takeaway is that the interface for creative work is shifting from the mouse to the prompt, and eventually, to the autonomous agent.

• Speed over precision: For social media, 90% accuracy in a jump cut is better than 100% accuracy that takes three hours to achieve.
• Developer-centric creative: We are seeing a world where the best video editors might actually be the people who know how to script their workflows.
• Data-driven cuts: Because the LLM understands the transcript, it knows exactly where the punchline is. It isn't just cutting silence; it is cutting for impact.

The Skeptic's Corner: Is it too automated?

I have a rule in my business: never let the machine have the final say without a human sanity check. While 15-second edits are impressive, we have to talk about the "uncanny valley" of AI content. If every creator starts using the exact same Claude-driven jump-cut logic, everything starts to look and feel the same. The rhythm of human speech is slightly messy, and sometimes that messiness is where the personality lives.

However, for the founder who is currently doing zero video because the friction is too high, this is a non-issue. The leap from "nothing" to "AI-edited" is infinite. The leap from "AI-edited" to "Oscar-winning cinematography" is a luxury most startups can't afford and don't need.

The biggest risk to a builder isn't that their video looks a little robotic; it's that they never hit 'publish' because they're stuck in an editing suite.

What This Means for the Content Stack

If you are building a startup right now, your marketing department shouldn't look like it did in 2022. You don't need a bloated team of editors for top-of-funnel awareness. You need one person who understands how to orchestrate these models. The role of the "Content Creator" is evolving into the "Content Architect."

We are looking at a future where your phone doesn't just record video; it communicates with your local LLM to prep that video for every platform before you even finish your coffee. This is the realization of the "founder-as-a-media-company" ideal that has been talked about for years but was always too hard to actually execute.

Tactical Takeaways for Builders

First, get comfortable with the CLI. If you are still only using the web interface for Claude or ChatGPT, you are missing out on the most powerful features. The ability to let an agent work on your local files is the real game-changer.

Second, stop worrying about the "art" of the edit for social platforms. TikTok, Reels, and Shorts are high-volume, low-longevity environments. The goal is to get your ideas out while they are still relevant. Use the time you save on editing to actually talk to your users or refine your product.

Finally, watch the infrastructure. The fact that Claude can handle these tasks signals that the underlying compute and modal understanding are reaching a tipping point. We are very close to a world where we don't "edit" video at all; we simply describe the finished product and let the machine assemble it from the raw assets.

The era of the $100-an-hour editor for basic social clips is effectively over. If you're a founder, that's a massive win for your burn rate and your output. If you're an editor, it's time to start learning how to build these tools rather than just using them.

Read the original at Lenny's Newsletter →