Amazon and Cloudflare accused of cloning Jev decision model in two weeks
A video claims both giants launched near-identical clones of Jev on October 1, replicating its architecture and API format but stripping the language module to output only yes/no verdicts, with far worse performance on complex reasoning and long inputs.
Amazon and Cloudflare just shamelessly ripped off a $200M startup’s tech to seize control of internet infrastructure with a 0.04-second response time. Startup TypeSafe AI spent months launching Jev, its flagship decision model. On October 1st, tech giants dropped near-identical clones, replicating its architecture and mirroring its API format in just two weeks. These systems had their language modules intentionally stripped out: they don't generate text they issue an instant "yes" or "no" verdict in fractions of a second. They’re the silent judges deciding whether a bank approves your transfer, your email hits spam, or a website gets blocked. There is no real breakthrough here: big tech hijacked an open-weights base and cherry-picked the data even Perplexity flaunted an open 27B model, claiming 85.71% vs. Jev 84.51% on a benchmark tailored to suit them. On complex reasoning tasks, these clones perform nearly twice as poorly as the original, and on longer inputs, they lag for twenty seconds instead of hundredths of a second. Yet the startup is getting crushed by sheer corporate scale anyway, why pay the creators when big tech can simply muscle in and take invisible control over global traffic