700 Leads Scored for $0.09

The author says concept preview only, not actual product footage, in the follow-up post. So the interface in the clip is not what a customer sees, and the 700-lead run is the only part being claimed as something that actually happened.

The post never defines predict. It could be a fit between lead attributes and message, or a learned response predictor, and without that definition a confidence score on an unnamed quantity cannot be checked. No accuracy figure against later reply or conversion rates is given.

No. Both numbers are vendor self-reported with no dataset, no baseline, and no outside run. Nine cents over 700 leads works out to about $0.00013 per lead, which lines up with the published input-token price, but a price that cheap does not verify the quality of the scores.

Early access. The post invites comments for access and the follow-up points to a free trial at gojiberry.ai to get ahead of the JEV integration, with the MCP part described as coming soon. There is nothing published to install or benchmark yourself today.

Romàn runs GojiberryAI, which he describes as an AI go-to-market brain with more than two thousand daily customers and a YC P26 batch label. The 700-lead run is a Gojiberry demonstration, so it is a vendor showing its own pipeline rather than a neutral test of Jev on lead data.

Beyond the 700-lead run it lists lead scoring, buying-signal analysis, matching each prospect to the best message, and identifying the campaigns most likely to perform. All of that is described as coming soon to Gojiberry, including an MCP integration, so none of it is presented as shipped.

Skeptically. A reply says the whole thing feels a bit orchestrated, and the author himself follows up to say the video is a concept preview rather than real product footage. So the thread's temper is enthusiasm about the price, doubt about the demonstration, and no third party reproducing the result.