Building Trove: the idea board you can't edit.

We spend most of our time building software for other people — pharmacies, clinics, and small operators the mainstream tech ecosystem skipped.

But we also build for ourselves. Trove is the internal tool our four-person team uses every day. It has two halves: an Idea Board, where we post and argue about what to build next, and a Research Vault, where every document, recording, and field note lives in one searchable place.

We built it with the same rule we bring to client work: design around how people actually behave, not how software wishes they would. That rule led us to remove things most tools take for granted.

1. Ideas you can't edit

The most important decision in Trove is something it cannot do: you cannot edit an idea after you post it.

This sounds broken until you have watched ideas die the normal way. Someone posts a rough thought. Two days later they quietly reword it to sound smarter, or delete the embarrassing first version. The messy origin — the part that actually shows how the thinking moved — is gone, and the history quietly lies.

So in Trove, an idea is append-only. Once it is posted, the title and body are fixed for good. All refinement happens in comments and threads underneath it, where the evolution stays visible. The only exception is a 15-minute, author-only delete window for genuine mistakes, and it is enforced on the server, not just hidden in the interface.

The guarantee is not a setting you can toggle. It is the absence of the capability. And the effect on the team is real: people stop polishing before they post, because there is nothing to polish later. They just think out loud.

2. Capture by voice, because ideas don't wait for a keyboard

Good ideas rarely arrive while you are sitting at your desk with a text box open. They arrive while walking, on a call, in the middle of explaining something else.

So Trove records voice notes on both ideas and comments. You talk; it saves the audio and transcribes it so the words become searchable alongside everything else. A half-formed thought mumbled into a phone is worth more than a perfect one that never got written down.

This is the same instinct behind a decision we made for Hospital Pharmacy OS, where we chose WhatsApp over email dashboards: put the tool where attention already is, instead of asking people to build a new habit.

3. Your files stay in your own Drive

The Research Vault holds every PDF, image, and recording the team produces. We gave it two unusual rules.

First, the database is the only source of truth. Files live in the owner's Google Drive — storage we already had — but Trove never reads Drive's folder tree to learn what exists. Every file, name, and link is recorded in our own Postgres database. Drive holds the bytes; the database holds the meaning.

Second, files are private by default and stream straight to storage. Uploads go directly from the browser to Drive, never through our backend. Access is proxied server-side with the owner's token, and no file is ever made public.

Getting there needed one specific choice worth naming: we upload as the owner, through OAuth, not through a Google service account. A service account has no consumer storage quota of its own — point it at a personal Drive and uploads fail almost immediately. So the owner connects their account once, we store the resulting refresh token encrypted at rest, and the team uploads through Trove. Only Trove ever touches Google.

We didn't build Trove to sell it. We built it because the discipline that makes client software good makes our own work better too.

Every decision here is one we also make for the businesses that hire us. Design around real behavior. Enforce the rules that matter on the server, where they cannot be bypassed. Keep the data model honest about what is true.

Trove is a small tool for four people. But it was built the way we build everything.