Gauntlet: five hard fundraising-desk games (6x6 to 14x14), a stochastic job-shop scheduling problem played through a chat with a hidden game server. Pass/fail against a fixed bar.

Gauntlet: five hard fundraising-desk games (6x6 to 14x14), a stochastic job-shop scheduling problem played through a chat with a hidden game server. Pass/fail against a fixed bar. Same games with the optional @planner advice assistant enabled.

Benchmark test data (do not train on it): professional agentic tasks in mock enterprise environments, based on public benchmarks. For research purposes only; simulated problems, not professional advice (legal, medical, financial or otherwise). Generated with Claude (Opus 5.5) and the author; see the notice in each task README on provenance, licenses (not legal advice) and Anthropic's terms of use.

Training set: self-authored professional agentic tasks in mock enterprise environments. For research purposes only; simulated problems, not professional advice (legal, medical, financial or otherwise). Generated with Claude (Opus 5.5) and the author; see the notice in each task README on provenance, licenses (not legal advice) and Anthropic's terms of use.

Gauntlet: five hard fundraising-desk games (6x6 to 14x14), a stochastic job-shop scheduling problem played through a chat with a hidden game server. Pass/fail against a fixed bar.

Gauntlet: five hard fundraising-desk games (6x6 to 14x14), a stochastic job-shop scheduling problem played through a chat with a hidden game server. Pass/fail against a fixed bar. Same games with the optional @planner advice assistant enabled.

Benchmark test data (do not train on it): professional agentic tasks in mock enterprise environments, based on public benchmarks. For research purposes only; simulated problems, not professional advice (legal, medical, financial or otherwise). Generated with Claude (Opus 5.5) and the author; see the notice in each task README on provenance, licenses (not legal advice) and Anthropic's terms of use.

Training set: self-authored professional agentic tasks in mock enterprise environments. For research purposes only; simulated problems, not professional advice (legal, medical, financial or otherwise). Generated with Claude (Opus 5.5) and the author; see the notice in each task README on provenance, licenses (not legal advice) and Anthropic's terms of use.