FiftyFlowers · Company knowledge

A company memory, starting with two teams.

Team knowledge around each person’s own memory, and a chat that chooses which knowledge to read and lists its sources, with voice and web research.

PROJECTCompany Second Brain

MY ROLEAI Solutions Engineer · FiftyFlowers

STAGEMVP · live since July 2026, two of seven team brains filled

A dark map with one central cluster and seven labelled clusters around it

Second Brain: a person’s own memory in the middle and the team brains around it. A local run of the platform’s own code with an invented team member; most team brains are still empty.

The context

A problem worth solving.

Company knowledge was spread across apps, documents and people. Each team needed AI that could use the right part of it and show where an answer came from.

My contribution

What I brought to the work.

Built the Second Brain workspace: team brains around each person’s memory, the Engineering Brain view of the shared repository that AI coding agents work through, and a chat that routes each question to the knowledge it needs.

The product

What it makes possible.

Four parts follow: the team brains, the chat that reads them, the Engineering Brain view and personal memory. The screens are local runs of the platform’s own code with an invented team member and invented knowledge; replies are test stand-in text.

Team brains

Each person’s own memory sits at the center with seven team brains around it. Two hold knowledge today: Customer Service, which already powers the production support apps, and Engineering. Marketing is forming and four are still empty.

Answers with their sources

A small router model picks the brains a question needs, readers pull the top pages, and the answer lists the pages it consulted.

A memory that grows

Add a chat export or a PDF (PDF upload is still to be confirmed on the live site), ask it to research a topic with at most two web searches, or run a check-up for stale and contradictory pages.

Engineering Brain

The shared engineering repository as it stood at the end of July 2026, 135 pages and 854 links, appears as a navigable constellation with a reader. The repository has since grown to 261 pages; bringing the view up to date is the next step.

Voice and gestures

Speech in and out, a live spoken conversation, and camera-based hand tracking to turn and zoom the knowledge map.

At a glance / simplified product view

01Question
02Chosen brains and pages
03Answer with sources

Second Brain

One brain per team, around each person’s own.

Second Brain draws the company’s knowledge as a map: a person’s own memory in the middle and a brain for each team around it. A team brain opens into its knowledge, the voices it writes in and the audiences it writes for, and each can be managed in place.

A team brain view with a constellation and an assistant panel — **A team brain, opened**Customer Service opened: its knowledge, voices and audiences as a slowly turning map, with an assistant for creating and managing them. Local run of the platform’s own code with an invented team member and invented knowledge.

A side drawer listing knowledge articles with edit and delete buttons — **Managing what a brain knows**The manage drawer: voices, audiences and knowledge with counts, a search box, and edit and remove on every article. Local run of the platform’s own code with an invented team member and invented knowledge.

Two sparse brain views, each with a note that it has no knowledge yet — **Honest about what is empty**A brain that knows its apps but holds no knowledge yet says so, and a team that has not been taught shows only a faint outline. Local run of the platform’s own code with an invented team member and invented knowledge.

Chat

A chat that reads the brains and says which ones.

For each question a small model picks the brains it needs, readers pull the best-matching pages, and the reply is shown under a line naming the brains it consulted. An answer can be saved as a page that lists what it was drawn from, and the chat can read up on a topic, take in files and check its own notes.

A chat page with a glowing orb, a small constellation and a welcome card — **Chat home**The chat opens on the person’s own memory, greets them by first name and offers three ways to start. Local run of the platform’s own code with an invented team member and invented knowledge. Every reply shown is the repository’s test stand-in text, not a model’s answer.

A chat reply with a consulted line, above a saved page listing its sources — **An answer, and where it came from**Top: a reply under the line naming the brains consulted. Below: the same answer saved as a page, with the list of pages it was drawn from. The reply is test stand-in text. Local run of the platform’s own code with an invented team member and invented knowledge. Every reply shown is the repository’s test stand-in text, not a model’s answer.

Three chat threads showing research, a file being read and a checkup card — **Research, files and a checkup**Research on a topic, a text file read into one page, and a checkup of its own notes. The replies are test stand-in text and nothing was searched on the web. Local run of the platform’s own code with an invented team member and invented knowledge. Every reply shown is the repository’s test stand-in text, not a model’s answer.

A chat reply with the orb pulsing while it speaks — **Read aloud**A reply being read aloud, with the orb in its speaking state. A live spoken conversation needs a paid voice session and is not shown. Local run of the platform’s own code with an invented team member and invented knowledge. Every reply shown is the repository’s test stand-in text, not a model’s answer.

Engineering Brain

The repository the AI coding agents work through, as a map.

Beta · still shows its July 2026 snapshot

The Engineering view draws the shared engineering repository as it stood at the end of July 2026: one point per page, one line per link, beside a file tree and a reader. Bringing the view up to date with the repository is the next step.

A constellation of repository pages beside a file tree and a page reader — **The repository as a constellation**Each page of the July 2026 snapshot as a point, with the workflows highlighted and the router open in the reader. An existing capture. Local run; the map is the July snapshot that ships with the platform, and one picture is an existing capture of the live site.

A graph with four highlighted points, above a reader showing a workflow page — **Finding a workflow**Pointing at Workflows in the legend lights the four workflow pages; the router then opens from the tree in the reader. Local run; the map is the July snapshot that ships with the platform, and one picture is an existing capture of the live site.

A memory that grows

Each person’s own memory, drawn the same way.

After a reply the chat saves what is worth remembering to the person’s own brain, and pages made from a document name their source. The same map can be turned with hand gestures from a camera.

A sparse constellation of a personal memory beside a file tree — **A person’s own brain**The Personal view: a few pages as a constellation, with the same legend, tree and reader as the team brains. Local run of the platform’s own code with an invented team member and invented knowledge. The pages shown are samples or come from the repository’s own test values.

A reader showing a short page with a citations list — **A page that names its source**A page distilled from a text file, ending in a Citations section with the file and the day it was read. Local run of the platform’s own code with an invented team member and invented knowledge. The pages shown are samples or come from the repository’s own test values.

The engineering map with a small green camera tile labelled hand nav on — **Hand navigation, switched on**The hand button lit and the camera’s small self-view at the edge. The camera picture here is the browser’s test image, not a person. Local run of the platform’s own code with an invented team member and invented knowledge. The pages shown are samples or come from the repository’s own test values.

Where it stands

An MVP, with two teams’ knowledge in.

An MVP, live since July 2026. Customer-service knowledge is in production use; the marketing brain is drawn but holds no knowledge yet, four team brains are placeholders, and the engineering view is a beta that still shows its July snapshot.

Same company

More from FiftyFlowers.

FiftyFlowers overview

DIY kit data migrationDelivered on a test copy of the store

Content EngineIn the company AI workspace since June 2026 · nothing published through it yet

Image StudioIn production · released August 2026

Proposal ManagerBuilt from the team’s wireframes · not launched

Customer service AIIn production · since January 2026

Storefront requestsRush form live · contest processing run on 2025 entries

AI Floral AssistantIn the store theme · behind an A/B test since January 2026

Keep exploringContent Engine