Databox vs OpenHouse

Databox's newer work sits in an undated block of the feed and is where the direction actually shows: Genie, an AI analyst answering performance questions in plain language; artifacts that package a Genie conversation into a shareable interactive document, now saved, searchable and directly editable; Databox MCP rendering interactive charts inside a Claude conversation; and a semantic layer where datasets, columns and metrics are defined and verified so people and Genie read the same numbers. The dated entries are older platform work — a push API, cloud warehouse connections, 350-plus integrations via Dataddo, OKRs and forecasting.

The through-line is that the dashboard is no longer the destination. Analysis starts as a question, ends as an artifact someone else can read, and increasingly happens inside another tool entirely through MCP. That only holds if the numbers are trustworthy, which explains the parallel investment in definitions and verification — marking which metric is official is what keeps an AI analyst from confidently answering from the wrong one. The connectivity work underneath, from the open API to custom API integrations, keeps widening what Genie can be asked about.

Expect verification and semantic definitions to become prerequisites Genie enforces rather than metadata users optionally fill in, and the artifact to keep absorbing what dashboards did. Note that these entries reach this feed with truncated bodies and missing dates, so scope is often unreadable even where direction is clear.