▼AI Automates the Labs' Own R&D

Score history

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

Now 25 · -2 since Oct 11 · ranged 25 to 27

Showing the last few days. Unlock full score history.

Why the conviction moved

• Oct 2
  Strengthened +5

  OpenAI said the inference software behind the up-to-8x Astra serving mode was optimized using its own internal models on Nvidia Blackwell hardware. This is a disclosed case of the labs' models doing the engineering that improves the labs' serving economics, tightening the loop the thesis claims will be increasingly disclosed and increasingly load-bearing.

• Oct 1
  Strengthened +4

  AREX-2 was trained on synthesized long-horizon improvement trajectories and scores 81.8 on MLE-bench Lite, a benchmark of machine-learning engineering tasks — evidence that ML research-engineering automation is now reachable with 27B open weights, not just inside frontier labs.

Showing the last 2 days. Unlock the full record.

Source trail

• Supporting · October 2, 2026

  OpenAI Ships an Ultrafast Serving Mode for GPT-6 Astra, Claiming Up to 8x Faster Token Generation

  OpenAI said the inference software behind the up-to-8x Astra serving mode was optimized using its own internal models on Nvidia Blackwell hardware. This is a disclosed case of the labs' models doing the engineering that improves the labs' serving economics, tightening the loop the thesis claims will be increasingly disclosed and increasingly load-bearing.

  NVIDIA Blog

• Supporting · October 1, 2026

  A 27B Open-Weight Agent From BAAI Posts 81.8 on MLE-bench Lite by Learning to Iterate

  AREX-2 was trained on synthesized long-horizon improvement trajectories and scores 81.8 on MLE-bench Lite, a benchmark of machine-learning engineering tasks — evidence that ML research-engineering automation is now reachable with 27B open weights, not just inside frontier labs.

  arXiv cs.AI