Technical blog

What we learned building Nym, for people who build agents

If you build agents, each post shows what we changed in Nym and what changed when we measured it. In the first, handing tool selection to Jev cut Nym’s generative-model calls from 72 to 41 on our 17-task benchmark.