Databricks rolls out Astra to all ~3500 engineers; coding spend up 60%
gdb · x · 2026-09-17
Databricks executive Peter Wendell announced Astra has been deployed to all 3500 engineers at the company, sharing findings from a 200-user pilot:
- Clear gains on hard tasks: Astra unambiguously outperforms their previous highest-end models (Opus 5, Sol 5.6) on highly complex work, especially high-level system design and long-range horizontal tasks.
- Usage surge: Engineers given Astra increased overall coding spend by 60% versus baseline.
- Saturated easy tasks: On medium/low complexity coding, Astra shows no meaningful improvement over earlier models—suggesting those tasks are already nearly perfectly executed by existing models.
The takeaway: frontier coding models now differentiate almost entirely on hard, long-horizon work, while simple tasks offer no headroom.
More from coding & agent
- Auditing my MCP credential broker: the worst hole was the config file — Relevant_Mousse_9117 · 2026-11-02
- AWS's open-source Strands Harness claims 45% cheaper than Claude Code, reigniting harness debate — blaizedsouza · 2026-11-02
- Xiaomi MiMo's CodeSpecWeave trains coding agents on interdependent specs for long-horizon coding — XiaomiMiMo · 2026-11-02
- Sketchy v0.1.6: AI diagrams use 29% fewer output tokens than plain text — Nkgsporty · 2026-11-02
- Avoiding AI brain rot: Miles Cranmer's lecture on learning with AI — _ScottCondron · 2026-11-02
- Scrapling web scraping framework tops 92k GitHub stars, ships MCP server for AI workflows — D4Vinci · 2026-11-02