▲The Harness Becomes the Capability Layer

Score history

Daily conviction score, 0 to 100. Higher means the thesis is more strongly corroborated.

Now 37 · +5 since Sep 13 · ranged 32 to 37

Showing the last few days. Unlock full score history.

Why the conviction moved

• Sep 14
  Strengthened +5

  A controlled study of paired runs on 80 private tasks put Anthropic's own agent software within 1.25 percentage points of an open-source harness on the same model, with the gap falling in both directions. If the vendor's proprietary scaffolding carries no measurable advantage on its own weights, the scaffolding layer is commoditized and capability attribution shifts back to the checkpoint plus freely copyable runtime conventions.

• Sep 11
  Strengthened +5

  OpenAI released the Codex harness as a managed Agents API in public beta with no platform fee beyond tokens and tools, and lets developers run the agent loop in OpenAI sandboxes or on Cloudflare, DigitalOcean and Oracle infrastructure. Pricing the scaffolding at zero is an admission that the harness is not the revenue layer — OpenAI is commoditizing the runtime to keep token demand attached to its weights.

• Sep 11
  Strengthened +4

  The Cornell/Microsoft Research result attributes long-horizon reliability gains to subagent dispatch versus context injection — a scaffolding choice made entirely outside the model weights. Measured capability moving on runtime design with frozen checkpoints is the thesis's core claim.

Showing the last 2 days. Unlock the full record.

Source trail

• Supporting · September 14, 2026

  A Controlled Test Finds No Advantage for Vendor-Native Coding Harnesses

  A controlled study of paired runs on 80 private tasks put Anthropic's own agent software within 1.25 percentage points of an open-source harness on the same model, with the gap falling in both directions. If the vendor's proprietary scaffolding carries no measurable advantage on its own weights, the scaffolding layer is commoditized and capability attribution shifts back to the checkpoint plus freely copyable runtime conventions.

  arXiv

• Supporting · September 11, 2026

  OpenAI Opens the Codex Harness as a Managed Agents API

  OpenAI released the Codex harness as a managed Agents API in public beta with no platform fee beyond tokens and tools, and lets developers run the agent loop in OpenAI sandboxes or on Cloudflare, DigitalOcean and Oracle infrastructure. Pricing the scaffolding at zero is an admission that the harness is not the revenue layer — OpenAI is commoditizing the runtime to keep token demand attached to its weights.

  OpenAI News