Loading activity…
@matei_zaharia
CTO @Databricks and prof @UCBerkeley. Working on data + AI, @ApacheSpark, @DeltaLakeOSS, @MLflow, @DSPyOSS, @GEPA_ai, @Omnigent_ai.
AI
Databricks says the architecture combines episodic and derived memory through plugins, on-demand tools and offline consolidation.
AI
The paper’s author says coding agents can run for days on one problem but sometimes stop improving early.
AI
A post sharing the announcement says decision models can run at scale in SQL and Spark queries.
AI
A person who says they work at Databricks claims agents using GPT-6 Astra and Opus 5 beat top GPU kernel engineers in a self-hillclimbing loop, with total spending of about $70,000.
AI
Databricks' rollout announcement says Astra outperformed Opus 5 and Sol 5.6 on highly complex tasks in a roughly 200-user pilot. Engineers given Astra increased overall coding spend by around 60% versus baseline.
AI
Arena describes a comparison of Claude Code, Codex and Pi. The researchers say harness choice had little effect on task success rate but can significantly affect cost.
AI
A user quoting the vacation suggestion adds, “I fully support slowing down AI! We need to take things seriously!”
AI
Zaharia says aggregate statistics will be released, following the approach used for an earlier MAP paper.
AI
The post points to harness-design research as a way to improve performance, highlighting Graph-as-Policy (GaP), which it says was accepted to CoRL.
AI
Databricks CTO highlights benefits of open models and inference advances from UC Berkeley work.
AI
Databricks CTO highlights tracing data for AI workload insights via Unity AI Gateway.
AI
The tool targets accurate field extraction from long PDFs and complex enterprise schemas.
AI
Databricks co-founder shares techniques used to lower internal AI costs while scaling use.
