Treg ships a Jev-powered social API for competitor ads
One endpoint for social data, mention tracking and Jev-filtered trending posts, with a price nobody outside Treg has checked yet.
Treg has released a Jev-powered social API, an open source service that pulls social data through a single endpoint, tracks mentions of you and your competitors, and uses Jev to filter trending posts in a niche. His headline number is under $0.05 to pull 1,000 competitor ads. He says the project ships with 0% markup, meaning you pay what the underlying calls cost.
Under $0.05 to pull 1,000 competitor ads
The Jev part is the filtering step. Rather than returning everything the upstream sources hand back, the API runs a model pass over trending posts to decide what belongs in a given niche. That is the same shape as the cheap-classifier work Treg has shipped before, including his earlier open source people search, which he priced at $0.0089 a lead.
What the number is and is not
Under $0.05 per 1,000 ads is Treg's own figure for his own pipeline, posted alongside the launch. There is no eval attached, no precision or recall on the filtering, and no second person who has run the API and reported what it got right or wrong. If you are costing out a competitor monitoring job, treat the price as a starting estimate from the author rather than a measured benchmark, and budget for the fact that filtering quality is the part nobody has graded.
That gap matters more than the cost. A filter that is cheap and wrong is a filter you will end up reviewing by hand, which puts the real price somewhere else entirely. The useful test is not whether a thousand ads cost five cents, it is how many of the thousand the filter dropped that you wanted, and that number does not exist in public yet.
Why people care about the cheap pass
Diogo Almeida, posting as @completeskeptic, framed the launch as an early case of applying models to data that organisations already hold but never process. His argument is that the two historical blockers were cost and reliability in map-reduce style passes over large collections, and that a cheap classifier changes the arithmetic on both. That is a thesis, not a result, and it comes from a reader of the launch rather than someone who ran the API.
For people building on it, the open source and 0% markup claims mean your bill is whatever the upstream social data providers and the model calls actually charge, so read the repo before you assume five cents is the all-in cost at your volume. And if the filtering step is going to sit in front of anything you act on, write your own small labelled set first. Treg has published a price. Nobody has published an accuracy.
why don't we throw intelligence at the big data we already have? too expensive + too unreliable in map-reduces
Know what Jev is good at before you bet on it
One email a day. Under three minutes. What builders found out about Jev yesterday.

