Cherry Pi proposes layered harness architecture: agent loops compose with governed workspaces
sull · x · 2026-11-03
redstarsystems released Cherry Pi, a multi-layered harness architecture for Governed Workspace Environments (GWE), laying out its definition of "harness."
Core claim: harnesses compose.
- Around the model (agent harness): a model alone only writes text; to act it runs a loop that assembles context, dispatches tool calls (reading files, running commands), feeds results back, and decides when to stop or ask a human. Tool definitions and execution rules shape what the agent can accomplish;
- Around the work (workspace harness): manages the project copy the agent runs in, access boundaries, processes, and records that persist after the agent exits, with humans inspecting changes before they enter a shared project.
Both layers legitimately deserve the name, and their boundaries and composition are a central concern of agent engineering.
More from coding & agent
- DeepSeek publishes prompts for all 7 Harness Creator Mode plugin demos — aigclink · 2026-11-03
- DeepSeek Harness's Creator Mode generates plugins from plain language, threatening micro-SaaS tools — aigclink · 2026-11-03
- A one-night AI-built project nets 333 points and ~22k visitors — Aizkmusic · 2026-11-03
- Five manual checks before trusting an agent actually finished its job — Kodenstudio · 2026-11-03
- Open-source AI pit crew for iRacing runs 5 agents on Kimi, checks every number against telemetry — JeremyCMorgan · 2026-11-03
- Agent control plane benchmark: Smalltalk-80 cascades hit 100% compliance vs 2.5% for Gherkin — RandalSchwartz · 2026-11-03