30,968 investigations. The findings point at us.
Seventeen years of the NTSB's complete aviation accident database, read into a grid space and written onto the board — where the most common finding in an aircraft accident is not the aircraft.
This page is the exhibit. The space behind it is the product. Data presented with its receipts — analysis belongs to qualified investigators.
What investigations actually find.
Every finding the NTSB coded across 17 years, grouped by its top-level category. The bright segment is where the finding was ruled a cause; the deep segment is where it was a contributing factor. Personnel issues out-count the aircraft itself — the pattern that shapes all of modern aviation safety thinking. Tap a bar for its numbers.
Seventeen years, two lines.
All investigated events per year, and the fatal subset beneath them. The gap between the lines is most of the story: the overwhelming majority of investigated accidents are survived.
How this was made.
The NTSB's bulk corpus — avall.mdb, the same file investigators download — pulled on a schedule, never hand-edited, refreshed weekly from the official incremental files.
Events, aircraft and findings land as records with payload, source and retrieval date, each at its own address.
The space's AI reads the records and writes these exhibits onto the board. The editorial choices are a person's; the drawing is not.
This page presents the record, not conclusions. Safety analysis belongs to qualified aviation investigators — a claim this database itself would insist on.
Walk into the space.
The demo space is the real product reading this real corpus — writing off, everything else on. Open any event, follow its findings, read the receipts, export the lot as plain JSON.