After 1,000+ Hours With Every AI Agent, This Dev Keeps Coming Back to Hermes
EXM7777 · x · 2026-11-08
A developer who claims 1,000+ hours building with AI agents (OpenClaw, Grok Bot, ChatGPT Dots and more) says he always ends up back on Hermes:
- Fast iteration, constantly absorbing competitors' best features; models, connectors and plugins land quickly
- As a harness it's sometimes more efficient for him than Codex or Claude Code
- Runs any model, supports bring-your-own subscriptions or local execution with automatic model selection
- A single hermes doctor command fixes breakage
The attached screenshot shows his persistent Claude (Opus 5.5) agent "Machina" in the Hermes desktop app: cross-session memory and skills, X posting/DMs, a real browser, Mac shell/Apple Notes/iMessage automation, cron jobs, and parallel subagents. He also notes a mobile app in beta and shares UI customization tips (hide tool calls and unneeded menus).
More from coding & agent
- DeepSeek says its Harness aims to be the OS of the agent era, with built-ins replaceable by plugins — dotey · 2026-11-09
- Claude Code Workflow Patches Vulnerable Dependencies One by One with Draft PRs — alex_verem · 2026-11-09
- Qwen quantized model on 16GB VRAM recreates Opus-style motion graphics in ~2 hours — Yhojm · 2026-11-09
- Claude Code writes ~265k lines in 3 weeks: open-source serverless P2P messenger over Tor — SayYoung530 · 2026-11-09
- Open-source Clean Development tool cleans up Claude Code session caches on macOS — Pathfinder-electron · 2026-11-09
- OpenAI details Pando, a Git-compatible backend built for Codex-scale development — rhiever · 2026-11-09