Short essays on archives, provenance, and putting AI to work on real libraries.
Teams buy a DAM, then quietly route around it. The reason is not discipline, it is direction: getting media in is manual, so the good stuff stays on drives, tape and cloud.
Joining ShotGrid, FileMaker and Airtable into one governed graph on the MovieLabs OMC model, so people and agents can ask one question that spans them all.
One query, checksums and metadata back, across every store you own. Read-only by design, so a question can never change your data.
Byte-identical copies sprawl across every tier and quietly pay rent. How to find what you are storing twice, priced on your own rates.
Camera RAW, EXR, DPX and legacy 3D, judged in a browser tab. No restore, no specialist software, and video previews a fraction of the bytes.
Generative tools are entering the pipeline whether it is ready or not. The line between chaos and control is a watermark and a credential on every output.
A shooting script becomes a 12-category breakdown before principal photography, deterministically and with no LLM in the parse.
Put an agent like Claude on your estate over MCP: join a breakdown with ShotGrid, ask across the archive, orchestrate ingest and generation, all audited.
LTO is the cheapest storage in the building and the least loved. With an index and previews on top, it stops being the place files go to be forgotten.
The four-week pilot: a script or ShotGrid export becomes a brain, the archive gets indexed read-only, a test pipeline runs, agents go on top. Your numbers.