Organisational Intelligence

The Harness

A CEO-model orchestrating disposable frontier-model workers — each isolated in its own git worktree, dispatched through a zero-markup model gateway. Same org, now genuinely multi-model.

OwnerhumanFable(Claude Code)orchestratorworker.pydispatchergit worktreeisolatedOpenCodeCLIsub-harnessVercel AIGateway0% markupFrontier workersKimi K3Grok 4.5Qwen3.7 MaxGrok 4.3Diff reviewFableSquash-mergeto main

Layers

LayerToolWhy
OrchestratorClaude Code (Fable)Plans, dispatches, reviews, merges
Dispatchertools/worker.pyWorktree lifecycle, logs, cost
Sub-harnessOpenCode 1.18.3Headless agentic loop w/ file + bash tools
GatewayVercel AI GatewayOne key, 0% markup, any model
Isolationgit worktreeFiles sandboxed per worker
Guardrailsopencode.jsonDeny-by-default bash allowlist
InstructionsAGENTS.mdPortable, agent-neutral

The honest read

Pros
  • Delegate any task to any model — any Vercel-gateway model is one registry line; any OpenAI-compatible endpoint, including local Ollama, is one config block.
  • Own the harness — not locked to one vendor's UI.
  • 0% gateway markup.
  • Portable instructions — AGENTS.md is read by any harness.
  • Every run logged with real cost.
Cons
  • Claude subagents in-subscription still beat these workers on pure code ROI — this is an R&D + agnostic hedge, not a cheaper dev team.
  • Workers run bash with the user's OS permissions — contained by a deny-by-default allowlist, not a full sandbox; a container/WSL jail is the next step.
  • v1 handles Python / data / content tasks only — no Next.js / npm builds yet.
  • Native-Windows setup had real friction.

Proof, not a pitch

7Worker runs
$1.25Total cost
1.18.3OpenCode
2026-07-17First pilot
4Frontier models