▲Frontier Labs Race on AI Coding Capability

Score history

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

Now 91 · +1 since Oct 11 · ranged 90 to 91

Showing the last few days. Unlock full score history.

Why the conviction moved

• Oct 12
  Strengthened +1

  SWE-Duel adds an adversarial, execution-scored coding benchmark at a time when existing coding tests are contested. It gives labs competing on code a new, contamination-resistant way to compare themselves.

• Oct 7
  Strengthened +2

  Meta's Claude Code user count reportedly fell by about half as Meta and Microsoft move engineers onto in-house and partner models. Coding tools are where big incumbents are actively replacing the current leader, Anthropic, rather than just matching it on benchmarks.

Showing the last 2 days. Unlock the full record.

Source trail

• Supporting · October 12, 2026

  SWE-Duel benchmark pits coding agents against each other to resist training-data contamination

  SWE-Duel adds an adversarial, execution-scored coding benchmark at a time when existing coding tests are contested. It gives labs competing on code a new, contamination-resistant way to compare themselves.

  arXiv cs.CL

• Supporting · October 7, 2026

  Microsoft and Meta cut internal Claude spending weeks before Anthropic's expected listing

  Meta's Claude Code user count reportedly fell by about half as Meta and Microsoft move engineers onto in-house and partner models. Coding tools are where big incumbents are actively replacing the current leader, Anthropic, rather than just matching it on benchmarks.

  AI Post