TypeSafe co-founder Diogo Almeida explains why Jev returns bounded judgments instead of writing replies, and argues for combining those decisions with software rules.
Original by ModalGetting startedBeginner13 min 10 secPublished
Almeida proposes putting bounded decisions inside ordinary software, rather than asking an agent to invent each next step.
His distinction between assistance and automation is an argument about how models are trained, not a demonstrated universal rule.
A benchmark chart and adoption claims are part of the pitch. They do not replace tests on the tasks your software needs to handle.
Worth knowing
Vendor conference talk, not an independent performance or adoption audit. The speaker's capability, usage and infrastructure claims remain attributed statements. The talk's criticism of benchmarks is an opinion, not a reason to omit task-specific tests. It does not establish safe autonomous decisions in high-stakes work.