Agent engineering splits into three tiers: scripts, System One decision models, and reasoning models

arpit_bhayani · x · 2026-09-19

Developer Arpit Bhayani argues agent engineering now spans a spectrum: deterministic scripts on one end, reasoning models on the other, with System One decision models like Jev in between. He suggests splitting skills/instructions into three tiers — deterministic logic goes to code, judgment calls go to system-one decision models, and only true reasoning goes to reasoning models — with new architectures and frameworks forming around this division.

Related event: Agents Split Into Generation Layer Plus Small Decision Models(2 posts)→

Original post →

More from coding & agent

coding & agent channel →