Thomas - thanks for the follow-up, and Gustavo, your M-Files example is a great comparison point.
What I've been running is part of our research and experimentation into AI-native practice tools. The first piece is called FlatFile Prep, currently in beta and tested locally on my actual project files. It shows how an AI-driven layer can eliminate folders and let you call up meeting minutes, schedules, window quotes - anything really - instantly by meaning.
Alongside that, we've released Jellyfish Lite - essentially a throttled version of an internal firm GPT-5. It can do what FlatFile Prep does, but in the wild it's currently limited to about 10 files at a time. Even at that scale, it connects with everyday tools and supports live practice conditions.
Unlike M-Files, this isn't about configuring metadata schemas. Instead, the AI backbone (Bailow.ai / Studio Stack™) auto-tags and links files by meaning, while building a cross-project memory so lessons from one job carry into the next.
This is still very much research and experimentation, but if anyone wants to try it, Jellyfish Lite is free here: Jellyfish Lite Architect.
------------------------------
Christopher Bailow
Principal, Bailow Architects / Bailow.ai
Boston, MA
------------------------------
Original Message:
Sent: 09-09-2025 05:49 PM
From: Thomas F. Fallon AIA
Subject: One Canonical Drive Instead of Folders?
Any chance you can name the system you are using so people can compare it with the M-File Gustavo uses in his response?
------------------------------
Thomas Fallon AIA
Benner Stange Associates Architects, Inc.
Portland OR
Original Message:
Sent: 09-07-2025 10:25 PM
From: Christopher M. Bailow
Subject: One Canonical Drive Instead of Folders?
In my practice, I've been testing an AI-native system that eliminates folders and file trees altogether. Instead, everything lives in one canonical drive that's tagged and retrievable by meaning.
The idea is simple: no matter where a drawing, submittal, or record originated, you can surface it instantly - without hunting through nested directories. Over time, the system builds a cross-project memory that makes old knowledge easier to reuse.
Has anyone else tried or seen approaches like this in architectural practice?
------------------------------
Christopher Bailow
Principal, Bailow Architects / Bailow.ai
Boston, MA
------------------------------