Jev in 34 seconds

Watch out

The captured page holds the post text and the video's runtime marker, not a transcript, so what the 34 seconds actually covers is not documented here: whether it explains the decision loop, contrasts Jev with an LLM classifier, or shows an API call is unknown, and any claim about the video's content would be inference rather than something this page states. Not documented: any repository, demo link, code, model configuration, prompt, or cost and latency figure. The post itself contains no example of anything built, only the author's reaction to understanding the model. The replies supply the substance and belong to other accounts: one asks for an example and describes an AI text classification use case, compares it to Pangram as roughly 6000x cheaper and 5-10x faster, says it was tested on a single essay, and links a detector site; another asks the author to open their DMs. Those figures are the replier's claim, unreproduced and unmeasured in the capture, and the linked detector is not part of this post. Copy reviewed against the linked page on 2026-10-04.