Databricks rolled Astra out to all 3,500 engineers: beats Opus 5 on complex tasks, +60% coding spend
gdb · x · 2026-09-17
Databricks exec Peter Wendell shared first-hand notes after deploying Astra to all 3,500 engineers, amplified by gdb.
- On highly complex tasks (high-level system design, long-range horizontal work), Astra unambiguously outperforms their previous top models, Opus 5 and Sol 5.6.
- Engineers given Astra increased overall coding spend by 60% vs baseline.
- No meaningful gains on medium/low-complexity tasks — they suspect those are already saturated by earlier models.
- Findings came from piloting with 200 users to gather quality and cost signal.
A rare large-scale enterprise data point: frontier model value is concentrating on the hardest long-horizon tasks.
More from coding & agent
- Tencent unveils Hyra 2.0, a research agent that builds reusable scientific tools — TencentHunyuan · 2026-11-02
- General Reasoning explains how to train agent swarms with RL and latency pressure — rosstaylor90 · 2026-11-02
- "Coding isn't the bottleneck anymore — usage windows are," says founder Kyle Gawley — kylegawley · 2026-11-02
- Demo passed ≠ production-ready: 5 questions to answer before letting agents run your business — _kellhus · 2026-11-02
- An AI agent passed 100 test chats—still not ready to run the business — djgoel123 · 2026-11-02
- Agent given total creative freedom produces a 1:44 code-drawn animated short film — ivegotnoidea1 · 2026-11-02