실무자의 시선.

How to connect Claude Code, Claude, Cursor and ChatGPT to the GA Genie One MCP server, and what to change before the Beta endpoint sunsets on 31 October.

Databricks serverless micro apps: since September 23, 2026, Genie App Builder apps scale to zero in Beta. Code-first apps can't. The name hasn't shipped.

16 live credentials, one $200 exam fee, a 14-day retake wait, two-year validity, and which Databricks certification to take first, verified on databricks.com.

Genie One vs Genie Agents vs Genie Code: Databricks now ships three different tools under one name. Here's what each does, what it costs, and who needs it.

An Electron shell around the Genie One web app, an Alt+Space launcher the docs never mention, and an updater that could not fetch build 0.2.2 on the day we tested it. What you get on macOS, what it costs in memory, what an admin can and cannot switch off, and who should wait.

The direct engine became the workspace default in July, automatic migration started shipping on 26 August, and the Terraform engine goes away in new CLI releases in September. The nine failure modes worth checking, each traceable to the release that fixed it.

A skill is now a catalog object you grant like a table. What the Beta does, the download-versus-live decision it forces, and why the docs cannot decide what to call it.

Full-text indexes with no embeddings, two SQL functions that do different jobs, and a scaling switch that landed separately. What changed on 1 June 2026 and what you now have to decide.

Databricks 환경 전체의 데이터 상태, AI 사용량, 비용, 태그를 계정 수준의 화면 하나로 보여 줍니다. 각 페이지가 무엇을 보여 주는지, 어떻게 켜는지, 다음에 무엇이 나오는지 정리했습니다.

프로그램 가이드는 단계를 나열합니다. 이 글은 그 단계를 실제로 통과하며 겪은 기록이고, 제 경우 무엇이 가장 크게 작용했는지에 대한 이야기입니다.

절반은 사람이 의도적으로 정의합니다. 나머지 절반은 이미 있는 대시보드와 쿼리에서 스스로 조립됩니다. 둘이 충돌하면, 감사할 수 없는 점수가 승자를 고릅니다.

Everything the Databricks Certified Associate Developer for Apache Spark exam tests, on one page, in the order the official guide lists the domains.

Databricks gives you a strong engine for enforcing data quality. Writing and maintaining the rules is still manual toil. That gap is where AI earns its place.