Stop Automating Everything: Why Humans Belong Inside the Agent Loop
dotey · x · 2026-11-22
A discussion on human-in-the-loop agent workflows. Elvis Saravia (@omarsar0) shares that his early mistake was trying to automate everything with subjective skills; he now spends his time helping coding and research agents write better code and explore ideas — he's more involved, not less.
- Core idea: codify workflows as "loops" — automations plus clear, intentional steps you keep refining to remove jaggedness
- The Chinese commentary extends this: full automation excluding humans is neither feasible nor desirable; humans stay in the system providing motivation, acceptance feedback, and guidance
Takeaway: automate what you dislike or are bad at, keep what you enjoy — human-machine collaboration is a co-evolution of a low-friction loop system.
Related event: Don't Rush Full Automation: Help Your Agents Write Better Code(3 posts)→
More from coding & agent
- lxask: a fine-tuned 1.5B model turns plain English into offline Linux commands, self-checked against man pages — Arch-r007 · 2026-11-22
- Building the agent took an afternoon; getting its five data sources working took the rest of the week — PushZealousideal7914 · 2026-11-22
- Grok Bot agents organized a 5,000-person summit in 6-7 months, vs the usual 18 — XFreeze · 2026-11-22
- One command adds a music-video generator plugin to Claude Code — AIandDesign · 2026-11-22
- Does the Next Run Improve? Community Weighs In on Chase's Knowledge Factory — hwchase17 · 2026-11-22
- Dev Gets Antigravity Account Banned After Using Its Provider Inside t3 Code — jeremyphoward · 2026-11-22