Scuba asks teams to save their work in one store that AI agents can also read

The longest audio note Scuba will transcribe runs 90 minutes ^[6]. Picture someone on a marketing team recording one after a customer meeting, saving it, and moving on to the next call. Scuba sorts uploaded files into thematic Collections ^[7] and says they need no manual editing ^[8], so the transcript lands in the campaign's Collection without anyone filing it.

The pitch has one human step on the user's side. Saving through Capture is the only way in that the launch report describes ^[5]. We'd expect a rollout plan to say people will save as they go. Whether they do is a retention question that the funding announcement leaves open.

Noah Desai Weiss, who was Slack's chief product officer, started the company last year with Dolapo Falola, formerly head of engineering at The Browser Company ^[4]. He states the case in terms of the models. "The frontier models aren't intelligence bound. They're context constrained," said Desai Weiss. "That's why we're building Scuba. So people never lose a great idea again." ^[14]

The first sentence is about models. The last is about people keeping their ideas, a note-taking promise. Scuba says the service helps people find information faster and improves what AI agents produce ^[15]. SiliconANGLE's launch report did not disclose pricing, customer numbers, retention or any measured gain in agent output.

The integration list points at the intended buyer, a team that already works inside AI tools. Scuba says the link removes the need to track down files and paste them into prompts ^[11].

Six named venture firms backed the company ^[13], and the proceeds go to engineering and go-to-market work ^[12].

Two checks sort the decision. The first is how often someone on the team pasted a file into an AI prompt by hand last week. The second is whether people already keep a saving habit, such as clipping pages or recording notes, in any tool. High on both: pilot. Frequent pasting but no saving habit: test the habit first, because the integration would have nothing to read. A saving habit but little pasting: wait, since the integration has little to remove. Low on both: skip.

We'd pilot with one team in the first cell, on one Collection, for a month. The tradeoff is that one team shows whether saving holds up and says little about whether a second team will read what the first saved. The comparison to make is saves per person in week four against week one.