Skip to content

Show-First Work

When the work is visual, show it.

You can see the bug. You can see the interaction that feels wrong. Then you write four paragraphs, attach a screenshot, and answer the questions that come back.

Show-First Work is the name we use for the other way: record what you mean, hand the recording to the person or the AI agent who will act on it, then look at what comes back.

Free plan · Starts in your browser

The Translation Tax.

Our name for it. We have not measured it.

When what you mean is visual, you turn it into words so someone else can turn the words back into a picture. The layout that is a little off. The click that does nothing. The animation that feels wrong.

Each trip through words loses something, and the reader asks you for it back: which button, which page, what did you expect to happen. We call that the Translation Tax. It has nothing to do with languages. It is the time, the follow-up questions and the rework of describing something you could have shown.

We do not know how large it is for your team and this page does not put a number on it. If you notice yourself writing a long message to describe something you are looking at, that is the moment it applies.

From describing the work to showing the intent.

Same task, different starting point.

Describing itShowing it
Describe the layout in a paragraphRecord yourself pointing at it
Write out the steps to reproduce itDo the steps once, on the recording
Paste screenshots and annotate themShare one link that plays what happened, with a transcript
Answer “what did you mean?”The person or agent starts from what you showed and said

Text still wins for decisions, specifications, code, and anything people will search or skim. Show-First Work is for the moments when what you mean is something you can see.

Show, Act, Prove.

Three steps. Here is what Vertov does at each one today.

01

Show

Record what you mean.

Point at it, click through it, say what you want in your own words.

Today. Vertov records a browser tab, a window or your screen, in the browser or with the Chrome extension, and gives you a share link when you stop.

02

Act

Hand over the context.

The person, or the agent, works from the recording instead of from a description of it.

Today. Every recording with speech has a timestamped transcript. “Copy for an AI agent” turns a recording into a Markdown brief. Agents can read recordings over MCP or with an API key.

03

Prove

Ask for something you can inspect.

A recording of the change working, or a document written from it, rather than a message saying it is done.

Today. An agent can upload a recording it already made, and Vertov gives it a share link you can watch and comment on at a timestamp. Vertov stores the recording; it does not decide whether the work is correct.

What has been demonstrated: recording, the brief, reading a recording over MCP and uploading a recording back have each been checked on their own. A complete run where a real agent client reads a recording, makes a change and returns its own recording has not been recorded yet. Until it has, read “Prove” as the direction and “Show” and “Act” as what works today.

The Visual Handoff.

What changes hands when you pass work over by recording.

A link
Plays in a browser for the people you allow. No account is needed to watch.
A transcript
Timestamped, so “at 0:42, this button” points at the right moment.
A summary and chapters
Written from the transcript; a paid-plan feature on the hosted service.
Clicks and pages
Captured by the Chrome extension, only if you turn click capture on.
A brief for an agent
The above as one Markdown document an agent can read.

A brief does not carry permissions, an expected outcome or checks for the result today. Say those in the recording, or add them in your own words.

Where Vertov fits.

Vertov is a screen recorder and asynchronous video tool. You record in the browser or with the Chrome extension, share a link, and get a transcript. The core is open source under the AGPL, and you can host it yourself.

Show-First Work is not a product category we claim to own, and Vertov is not required to practise it. Any recorder, any note and any agent can be part of it. We built Vertov because we wanted the recording to carry the transcript and the context with it, and to be readable by an agent as well as by a person.

The End of Explaining

You can see the bug. You can see the layout that is wrong. You know exactly what you want it to look like.

Then you type four paragraphs, attach a screenshot, and try again.

The world has enough ways to describe work. It is time to show it.

Show the problem. Show the intent. Show the result.

The practical things

Before you
hit record.

What is Show-First Work?
A way of working: when what you mean is visual, record it instead of describing it. Hand the recording to the person or the AI agent who will act on it, then look at what comes back. The words are the name we use for it, not a standard.
Is Show-First Work an established industry category?
No. It is a name Vertov uses for this way of working. Other teams and tools describe the same idea in their own words, and we do not claim to have invented it or that it is a recognised term.
Does this replace writing things down?
No. Text is still the better tool for decisions, specifications, code and anything people will search or skim. Show-First Work is for the moments when what you mean is something you can see.
What is the Translation Tax?
The time, follow-up questions and rework that come from describing something visual in words that someone else has to turn back into a picture. It has nothing to do with languages. We have not measured how large it is, and this page makes no claim about how much time it costs.
Does Vertov run my AI agent or check its work?
No. Your agent does the work. Vertov holds the recording, the transcript and the brief it reads, and can store a recording the agent made and show it back to you. Whether the work is right is for you, or for checks you set up, to decide.
Which AI agents can read a recording?
Claude, ChatGPT, Codex and other tools that support MCP can connect to Vertov with an OAuth sign-in or an API key and read a recording as a transcript or a Markdown brief. The connection is built and has been tested with API keys and a test client; signing in from the real Claude, ChatGPT and Codex apps has not yet been recorded as tested.
Is it free to try?
Yes. The Free plan keeps 5 hours of video, with no plan limit on how long a single video can be (one recording can run up to six hours). Paid plans add room. People who only watch, react or comment are always free, and so is every member you invite.

Less explaining. More getting somewhere.

Go on. Show us what you mean.

Start recording free

In your browser. No card. No install to get started.

Commands and recent recordings
Actions
Go to
↑ ↓ Select↵ OpenEsc Close