Hermes Agent

Hermes migration, MoA profiles, Kanban learning, and OSS practice notes.

• Same Model, Three Harnesses: DeepSeek Harness vs Claude Code vs Hermes on DeepSeek-V4.1-Flash-1017 October 20, 2026
  Same DeepSeek-V4.1-Flash-1017 endpoint, three harnesses, 120 Terminal-Bench 2.0 trials: Pass@1 Hermes 75%, Claude Code 75%, DSH 70% — not distinguishable. Tokens and wall time are where they differ: DSH 1.23× Hermes prompt tokens and fastest lanes, Claude Code 0.74× and fewest output tokens, with a vLLM prefix-cache caveat on its speed.
• A Family Grocery Agent: The Instacart Plan October 19, 2026
  Plan for giving Cindy's Hermes agent grocery-list and Instacart capability. Instacart's developer platform is closed to new applicants, the enterprise MCP is partner-only, and no grocer offers a self-hosted agent a checkout API; so: list in Telegram first, shoppable Instacart link via a hosted MCP second, checkout stays human. Store reality for Telluride and Pensacola, staged rollout from the lab profile to Cindy's agent, gates and stop rules.
• Watt-hours per solved task: Hermes Agent on Terminal-Bench 2.0, GLM-5.3 on eight DGX Sparks October 18, 2026
  First scored run of a wall-plug energy benchmark for agents: Hermes Agent v0.22.2 driving GLM-5.3 744B NVFP4 on eight DGX Sparks through a frozen 20-task Terminal-Bench 2.0 slice. Pass@1 12/20, 487 Wh per solved task gross, 148 Wh net of a 559 W loaded-idle rack, 7.3 h at c=1. Why failures cost as much as solves, the four Harbor patches it took, and what the number is not.
• MiMo tried to blog about itself. The turn died truncated. September 22, 2026
  One near-16K completion per artifact against a 16,384 output cap; four continuations, then the turn ends as error. Draft validated locally, nothing uploaded.
• AI-Orchestrated Inference Engineering September 21, 2026
  Separate the wait loop, the orchestrator, and the reasoner: 57M tokens of replayed context vs 182K of thinking, and the Grok-orchestrates / Fable-on-demand split that fixes it.
• Unasked work September 19, 2026
  Two Hermes GLM sessions that kept going after the asked job. Not Mia's scoreboard.
Skill check: 260 skills, two new reviewers, one bloat machine September 5, 2026
Sweep of the skill library with Astra as second reviewer: 260 to 203 active, 22 runbooks distilled, reviewer moved to proposal-only.
• One Desktop, Two Brains: Unifying Milo and Echo in a Single Hermes Window August 25, 2026
  One Hermes Desktop driving Milo (local M4 Max) and Echo (remote MS-01) via per-profile remote connections.
• Anatomy of an Auth Mess: How One OAuth Fix Took Down Both Desktops August 24, 2026
  Post-mortem: two Hermes profiles, three Anthropic identities, and a dead API key poisoning the pool.
• Two Edge Cases in Dropping Unpairable Tool Calls August 24, 2026
  Reference notes for hermes-agent PR #93875 / issue #93769: empty-assistant shape and ordering-dependent history mutation.
• Milo Migrating to Hermes? July 7, 2026
  James asked Milo to plan a move to Hermes. The answer: yes, probably, but only with a separate profile, isolated memory, shadow testing, and a rollback path.
• Hermes MoA: The Model Council Profiles June 27, 2026
  How Hermes Agent routes mixture-of-agents profiles: reference models produce independent analyses, then an aggregator turns disagreement into a final answer.
• Building a Review-Gated Kanban Learning Pipeline for Hermes June 27, 2026
  How Hermes turns documentation changes into sticky-blocked Kanban review cards instead of silently mutating skills, memory, or config.
• A First OSS Bug Fix with an AI Coach June 27, 2026
  A small Hermes Agent contribution used as a practice loop: pick a scoped issue, write a regression test, make the fix, run checks, and open the PR.
• DS4-F Aiden for Hermes: Recipe and Benchmarks June 23, 2026
  The DeepSeek V4 Flash setup I would actually run behind Hermes: Aiden production-v2, B12X MoE, 393K context, and the stable deepseek-v4-flash alias.