Who the modern creator actually is

The job changed. Most advice did not.

For about fifteen years, being good at video meant being good at software. The gap between an idea and a watchable video ran through a timeline, and the timeline was hard. It rewarded people who would spend a weekend learning keyframes, and it quietly filtered out everyone who would not. That filter is where a lot of received wisdom comes from: learn the tool, then earn the right to have an opinion.

That filter is gone. A competent cut, clean audio, accurate captions and a vertical crop are no longer the product of skill and time; they are the product of asking. Anyone can now clear the bar that used to separate amateur from professional output. Which means the bar moved — and it moved to somewhere much less comfortable.

When execution is free, the only remaining scarcity is judgement: knowing what is worth making, who it is for, and why anyone should care. There is no tool that supplies that, ours included. What a tool can do is stop taking the hours you need in order to develop it.

Our definition: video creation is a sequence of decisions, not a sequence of operations

Here is the definition we build against, and it is deliberately narrow.

Video creation is the act of making a series of decisions — about what to say, what to keep, what to cut, and who it is for — and then having those decisions rendered faithfully. Everything that is not a decision is an operation, and operations should be free.

"Trim 400 milliseconds off the head of clip three" is an operation. "This opening is slow" is a decision. "Set the font to 72px at y minus 220" is an operation. "The title has to land before they scroll" is a decision. The traditional timeline makes you express every decision as a long chain of operations, and the translation cost is so high that it eats the time you would otherwise spend deciding.

This is why we do not describe Vidmoat as a faster editor. A faster editor still asks you to think in operations. We are trying to move the interface up a level, so that you state the decision and the operations happen underneath — and so that when the machine gets it wrong, you correct it in the same language you used to ask.

It also explains what we refuse to build. We will not generate a point of view for you. The agent can cut your dead air, find your best thirty seconds and write your captions; it cannot know what you actually think, and a tool that pretends otherwise produces the thing the internet already has too much of.

The four decisions that are actually yours

If operations are free, it is worth being precise about what is left. In our experience there are four, and they are the whole job.

Anything outside these four is a candidate to hand over. Anything inside them is where your time should go, and the reason to hand the rest over at all.

1. 1What is this about? Not the topic — the claim. "Three lens tips" is a topic. "Most lens advice is sold to people who would be better served by moving closer" is a claim. Claims get watched and argued with; topics get scrolled past.
2. 2Who is it for? Specific enough that someone outside it would not click. "Creators" is not an audience. "People who edit podcasts alone at midnight" is.
3. 3What stays? The hardest one, and the one most new creators get wrong by keeping everything they shot because they shot it. Length is not value.
4. 4Where does it go, and in what shape? A sixteen-by-nine explainer and a vertical hook are not the same video with different edges. Deciding what each platform gets is an editorial decision, not an export setting.

A roadmap for the first year

This is the sequence we would give someone starting now. It is deliberately slow at the start and deliberately unglamorous — the failure mode for new creators is not moving too slowly, it is buying equipment and building a workflow for an audience that does not exist yet.

Each stage has an exit condition. Do not move on because time has passed; move on because the condition is met.

1. 1Stage 0 — Publish ten things badly (weeks 1-4). One camera, whatever you own. No intro, no branding, no schedule. The purpose is not growth; it is to find out what you actually have to say when nobody is watching, and to make editing normal rather than special. Exit condition: ten videos published. Not ten filmed. Ten published.
2. 2Stage 1 — Find the claim (weeks 4-10). Look at your ten. One or two will have a real argument in them and you will know which. Make five more that are only that argument, from different angles. Let the agent handle the cut so you are spending your hours on what you are saying, not on the timeline. Exit condition: you can state, in one sentence, what your channel is arguing.
3. 3Stage 2 — Get repeatable (months 3-5). Now build the workflow, because now you know what it is for. Fix a format and a cadence you can actually hold on a bad week — weekly beats daily that lasts a month. This is where automation earns its place: the same silence pass, the same caption style, the same reframe, every time, without deciding again. Exit condition: four consecutive weeks shipped on schedule.
4. 4Stage 3 — Multiply (months 5-8). One recording is several videos. The long version, the vertical cut, the single argument pulled out as a short. This is where Repurpose does the work: same source, different shapes, deliberately chosen per platform rather than exported four ways and hoped for. Exit condition: your best video exists in at least three forms and you know which form performed.
5. 5Stage 4 — Build the machine (months 8-12). By now you know your format well enough to describe it to something else. This is where the agent, the MCP integration or the API stop being novelties: you are no longer editing, you are specifying. Some creators reach this and hire a person instead, which is a perfectly good answer. The point is that the editing stops being the constraint. Exit condition: you can be away for two weeks and still publish.

What to hand over, and what never to

The useful rule: hand over anything where there is a correct answer and you would be annoyed to do it by hand. Keep anything where the answer is a matter of taste.

Hand over: removing silences and filler words, transcribing, generating captions and styling them consistently, reframing to vertical, matching colour across clips, ducking music under speech, cutting the ten-minute recording down to the parts where you were actually making the point, producing the four platform variants, normalising loudness. None of these have an opinion in them. All of them take hours.

Never hand over: the claim, the opening ten seconds, what you cut for length versus what you cut because it was weak, and the moments you know are good even though nothing in the data says so. The agent optimises for what it can measure. Some of what makes a video worth watching is not measurable, and if you hand that over you will produce something competent and forgettable — which is a worse outcome than something rough with a point.

A practical version of this: let the machine make the first pass and then watch it once, all the way through, without touching anything. Every time you want to reach for the timeline, note why. Those notes are your taste, and they are the only part of this that compounds.

Where Vidmoat fits

Concretely, for someone at Stage 0 or 1: record on whatever you have, drop the file in, and say what you want. "Cut the dead air, remove the filler words, add captions, make it vertical." That is one sentence and it replaces the evening you were going to lose. Then spend that evening on the next video instead of on this one.

For Stage 2 and 3, the value moves from any single edit to consistency: the same treatment every time without re-deciding it, and one recording turned into the set of things each platform actually wants.

For Stage 4, there is an API and an MCP server, so the workflow you have proven by hand can be driven by an agent or by your own code. That is genuinely a minority of creators and we would rather say so than pretend everyone needs it.

And if you are earlier than all of this — if you have not published anything yet — the honest advice is to close this page and go publish something badly. No tool fixes not having started. Ours included.