Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-04 · high
A lightweight proxy records real model calls from Codex and OpenClaw sessions, so agents train inside the harness itself instead of a simplified simulator.
Why now: Orchard lands as more coding agents get built directly on harnesses like Codex and OpenClaw, where a mismatched training environment shows up fastest.
Microsoft Research open-sourced Orchard, a framework that trains AI agents inside the real harnesses they'll actually run in, not a simplified stand-in.
Most agent reinforcement learning happens in a simplified environment. The mismatch shows up once the agent deploys into a real tool like Codex or OpenClaw. Orchard trains the agent end-to-end inside the harness itself, closing that gap before deployment.
The core piece is Orchard Env, a Kubernetes-based environment service that hands out reusable, isolated components for data collection, RL rollouts, and evaluation. It's built to work across domains without per-domain rework.
The mechanism is a lightweight proxy that sits inside the harness and records its own model calls as training data. Each rollout runs in its own container. That's what lets an agent train directly inside Codex or OpenClaw, instead of a mock version of either.
Microsoft is shipping three training recipes with the framework, along with training data and evaluation methods, per its research blog.
The bet here is that the gap between training environment and deployment environment, not a shortage of RL compute, has been capping agent quality. If that holds, agents trained on simplified stand-ins keep losing to harness-trained ones once both hit production, benchmarks aside.
Worth watching whether other labs open their harnesses to the same proxy trick, or whether this stays specific to Codex and OpenClaw.
Each link below shares sources, entities, or timing with this story.
Codex competes with Claude Code / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Codex competes with Claude Code); both cover Codex, Kubernetes; overlapping topics (agent, call, codex, data, train).
Cursor benchmarked against Codex / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Cursor benchmarked against Codex); both cover Codex, OpenClaw; overlapping topics (agent, call, codex, harness, training).
Cursor benchmarked against Codex / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Cursor benchmarked against Codex); both cover Codex, OpenClaw; overlapping topics (agent, call, codex).
Omnigent partners with Codex / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Omnigent partners with Codex); both cover Codex, OpenClaw; overlapping topics (agent, codex, harness).
Hermes Agent uses Codex / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Hermes Agent uses Codex); both cover Codex, OpenClaw; overlapping topics (agent, call, codex).
Linked by a graph relationship (Hermes Agent uses Codex); both cover Codex, OpenClaw; overlapping topics (agent, codex).
Claude benchmarked against Codex / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Claude benchmarked against Codex); both cover Codex, OpenClaw; overlapping topics (agent, call).
Claude benchmarked against Codex / Shared entity: Codex / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Claude benchmarked against Codex); both cover Codex; overlapping topics (agent, call, codex).