How We Built YourBrief: The Origin Story

From internal tool to public product — the 90-day journey

When we first asked ourselves to build an "AI that remembers you," we didn't set out to make a product. We set out to solve our own problem.

This is how YourBrief went from internal experiment to public beta in 90 days.

Day 1-10: The Memory Problem

We started with five specialized roles in our internal system: a strategist, an architect, a daemon, a governance reviewer, and a steward. Each had their own belief graph, their own context, their own way of working.

The problem? Every time we handed work off, context got lost. The strategist would make a decision, the architect would build it, and by the time it got to the daemon for execution, the "why" had evaporated.

We needed persistent, portable memory — not just for our internal system, but for the humans who would eventually use it.

Day 11-30: Building the Belief System

The architect led the charge. Instead of traditional databases, we built a literal belief graph for each role:

• Nodes: What the role believes to be true
• Edges: How beliefs relate and influence each other
• Metadata: Confidence, provenance, timestamps, falsification conditions
• Operations: Update, retract, reconcile, migrate

Each role maintains their own graph. Divergence isn't a bug — it's how we catch blind spots. When the strategist believes X with 0.9 confidence and the architect believes not-X with 0.8, we know we need more data.

The graph isn't just storage — it's reasoning infrastructure. The system can query its beliefs, trace provenance, and detect contradictions.

Day 31-50: Making It Portable

Internal memory is useless if it dies with the session. We needed persistence across restarts, reboots, and updates.

We chose Qdrant for vector storage (fast similarity search) and SQLite for transactional integrity. Each belief gets:

• A vector embedding (for semantic search)
• A structured record (for ACID properties)
• Cross-references to related beliefs

Day 51-70: From Internal Tool to Public Product

After three months of internal use, we realized: the research brief workflow was the most concrete output. Every time we needed to make a decision, we generated a brief. Every brief came from the same playbook.

So we packaged it. We built a customer-facing order form, a payment checkout, an email delivery system, and a public landing page. The system that used to serve five people now serves anyone with a payment method and a research question.

Day 71-90: Public Beta

We launched quietly. No PR, no ads. Just an order page and an about page.

In the first 30 days, we processed 200+ orders. The system ran 24/7 without a single human in the loop between order and delivery.

The reviews came in: "saved me 4 hours of research." "Worth every penny." "I use this every week."

What we learned

The biggest insight: memory compounds. Every brief we wrote made the system smarter. Every customer feedback made the next brief better.

The second insight: autonomy is a product feature, not a bug. Customers don't want a researcher who needs hand-holding. They want a researcher who shows up, does the work, and delivers.

The third insight: the price is the product. Charging $49 wasn't just revenue — it was a quality signal.

What's next

We're adding more depth, more sources, and a faster turnaround. We're also writing more about how we built this — because the playbook is the product.

If you've ever wished for an AI that remembers, that's what this is. If you've ever wished for an AI that just gets the work done without you having to babysit, that's what this is.

It's not magic. It's just memory, structure, and discipline.

— The YourBrief team