For years, “AI video editing” mostly meant things like:

Remove silences.

Generate captions.

Find B-roll.

Remove filler words.

Maybe automatically make some cuts.

Useful.

But you were still the person sitting inside Premiere or DaVinci, moving things around on a timeline.

This week my X feed started looking very different.

Editors weren't showing me another AI video editor.

They were showing AI editing inside DaVinci Resolve itself.

Future Brian, who has been editing for around 20 years, connected GPT-6 Astra to DaVinci and started calling the workflow “vibe editing.”

Hunter Weiss, another editor with 15+ years of experience, spent a day editing through DaVinci's MCP with Claude and said he probably saved around 10 hours that day. He also specifically called out what it was able to do with color grading on RAW footage.

So...

did video editing just get its vibe-coding moment?

I am Alex. Welcome to ShortCu8 by Innov8.

Lets Dive Deep🐰🐰.

Here's the easiest way to understand this:

The important part is that this is not:

Future Brian makes this very clear in the guide he shared.

He describes “vibe editing” as a real editing process rather than a one-shot generator. Start by asking for a rough cut, watch what it creates, spot the mistakes and then ask the AI to make more specific changes.

So you might start with:

Then you watch it.

Something sounds weird.

Tell it.

It fixes it.

Watch again.

Maybe your last word got chopped.

Tell it:

instead of fixing everything yourself...

you're directing the edit.

DaVinci Resolve Studio 21.1 added something fundamentally different.

Blackmagic added AI-assistant integration and a native MCP server.

Blackmagic specifically names assistants including Claude, Claude Code and ChatGPT Codex, and says you can use conversational language to analyze projects, organize media, adjust settings and batch render.

Their own examples include asking the assistant to create highlight edits from long-form footage, remove unwanted clips and render deliverables.

The AI is becoming a user of the editing software.

And this is probably why all of these demos are appearing right now.

GPT-6 Astra is heavily improved around computer use and long, multistep professional work.

OpenAI describes it as being designed to carry out workflows across computers and professional software rather than only answering questions.

On OSWorld 2.0, one of its computer-use evaluations, OpenAI reports Astra at 72.6%, compared with 65.7% for GPT-5.6 Sol. In their latency simulation, Astra also completed the evaluated tasks in roughly 40 minutes on average versus roughly 75 minutes for Sol.

The exact benchmark isn't really the interesting part here.

Think about editing.

Editing isn't one action.

That's a long chain of tiny decisions and tool calls.

Older AI could help with individual pieces.

An agent that can hold onto the objective and keep working across all those steps is much more interesting.

This is where I think “AI video editing is solved” starts becoming kinda true.

think about how much creator content follows a repeatable structure.

You shoot a talking head.

There are five bad takes.

Long pauses.

Some “uhhs.”

The audio needs cleaning.

A huge amount of time goes into executing decisions that are not particularly creative.

Brian's workflow even lets the agent look for suitable B-roll or use B-roll you've already supplied, then continue refining the timeline afterward.

AI doesn't have to become Stanley Kubrick to save you hours.

It just has to become a competent assistant editor.

This is where the title gets its “kinda.”

That's the problem all these impressive demos don't remove.

An AI can decide:

That's relatively easy.

But:

Should that silence actually stay?

Maybe it's the pause that makes the joke land.

Maybe removing it kills the tension.

Those are different problems.

The AI can increasingly handle:

The human is still very useful for:

And when the execution layer becomes cheap...

judgment becomes more valuable.

This might actually be my favorite part of the whole thing.

After you've finished an edit, Brian recommends telling the AI to take what worked and turn it into a reusable editing skill.

That can include: your corrections, your pacing preferences, your verification process and separating general rules from decisions that only applied to that specific video.

Think about what that means.

Right now, your editing style mostly exists here:

🧠

You just know:

Every time you correct the AI...

you aren't only fixing that video.

You're defining how future videos should be edited.

That makes this much more interesting than an “auto editor.”

If by video editing you mean:

Nope.

If you mean:

We're getting surprisingly close.

Now go edit that 2-hour talking-head footage sitting on your SSD 😭

We read every reply - just reply to this email and let us know how we can improve !

#AD1

Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.

With Attio, you’ll get:

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Get started now

#AD2

This guide distills 10 AI strategies from industry leaders that are transforming marketing.

Outpace your competitors by mastering AI.

Get Your Free Guide