New: Spacedock is #1 on the Berkeley Data Agent Benchmark

Blog

  1. How Spacedock tops the Data Agent Benchmark (June 2026)

    Spacedock is #1 on DAB at 65.55%, more than 20pp above the published baseline. We explain what makes the benchmark hard, how the three-stage solver workflow works, and how Spacedock ran the experiments that got us there.

  2. The blog is just getting started

    Field notes on building Spacedock are coming: design decisions, the evidence behind them, and release notes when something ships.