We are building the production layer
for software companies
Every software company needs to explain itself — to buyers, to users, to the person who just opened a support ticket. That explanation currently costs a week and an agency, so most of it never gets made. We think that is a tooling problem.
Six opinions that shaped the product
A demo is an argument, not a feature list
The best product demos have a thesis: here is the painful thing, here is the moment it stops being painful. Most demos are a tour of the navigation. We build tooling that makes the argument easier to construct than the tour.
Production quality is a trust signal, unfairly
A buyer who cannot evaluate your architecture will absolutely evaluate whether your demo looks like a real company made it. That is not fair, but it is true, and it means production quality should not be a function of budget.
The second version is where the value is
Anyone can make one good video. The expensive part is the one hundred and fortieth, and the re-record after the UI changes, and the eleven translations. We optimise for the boring middle of the library, not the hero asset.
Automation should be legible
Every decision the pipeline makes is exposed as an editable object. A tool that makes choices you cannot see or change is not a tool, it is a slot machine.
Say where you are the wrong choice
Our comparison pages open by naming the cases where the competitor wins. We would rather lose a deal we would have churned than win one we could not keep.
Recordings are customer data
A screen recording of your product is a recording of your customers. We treat it with the controls that implies: never used for training, masked before render, deletable with a certificate.
How we got here
- 2024
The wrong kind of Tuesday
Our founders were running product marketing at a Series B company and spent a full week producing one four-minute launch video. The feature shipped on the Monday after it went live.
- 2024
First version, one customer
A crude tool that cut filler words and added zooms. The first customer used it to produce eleven videos in a week, which had previously been their annual output.
- 2025
Documentation, by accident
A customer asked whether the interaction data could produce a written guide. It could. Guides are now the most-used output on the platform and are unmetered on every plan.
- 2025
Series A
$22M to build the localization pipeline and the enterprise controls that security teams kept asking us for.
- 2026
Riff Agent
Autonomous demo generation from a changelog entry. Still gated behind human approval, and it always will be.
- Today
9,400 workspaces
2.4 million demos produced, in 34 languages, by teams from two-person startups to companies with four thousand employees.
Four things we hold each other to
Write it down
Decisions live in documents, not in meetings. If it was not written, it did not happen.
Ship to a person
Every feature has a named customer who asked for it. We do not build for a persona.
Be honest in public
Including about limitations, outages and the cases where a competitor is better.
Sustainable pace
No heroics. A company that needs heroics has a planning problem, not a commitment problem.
Companies that trusted us early