Understanding the human perspective behind political disagreement.
Before We Speak is a tool for the conversations that matter, the ones with the family member, the coworker, the old friend who sees politics nothing like you do.
It doesn't tell you what to think. It shows you what their world actually looks like today: the stories getting heavy coverage in their feed, the figures dominating their cycle, the emotional tenor of it all. Then it gives you a way in: topics to engage, framings that open dialogue instead of closing it.
What makes it different is where the understanding comes from: real human perspective, not AI guesswork.
Most AI tools collapse people into labels. "Lean-left" becomes a stereotype. "Evangelical" becomes a caricature. A 62-year-old Cuban-American Catholic in Miami gets the same flat treatment as a 28-year-old white Catholic in Boston. They see the world through radically different eyes.
So I built a taxonomy. Country, leaning, region, culture, religion, generation, plus the intersections between them, because a Jewish Gen Z progressive isn't just "Jewish + Gen Z + progressive" in a blender. Identity is load-bearing in specific combinations.
Every node in that taxonomy is fed human-contributed perspectives from real people who actually live inside those identities. The AI doesn't invent the scaffolding. It synthesizes what contributors have said, and that synthesis shapes what the model pays attention to when it builds your briefing. The fault lines that matter to someone like the person you're calling. Not a generic bucket of talking points a language model confabulated from its training data.
The taxonomy is sparse by design. Only nodes with real contributions show up. If you see something thin, or think your perspective is missing, get in touch. That's how this gets better.
I'm Peter Surrena, a product designer based in Brooklyn. My career has sat at the intersection of editorial, product, and technology: leading design at Digiday Media (Digiday, Glossy, Modern Retail), then building out IDC's knowledge platform, and now at Bank of America.
Before We Speak started as a side project trying to reconcile two things I'd been thinking about for years. The first: watching media bubbles form and harden from inside the publishing world. The same event framed five different ways depending on which feed you land in. The second: the everyday experience of trying to talk politics with people I love who see things differently, and how quickly those conversations collapse even when everyone's acting in good faith.
The taxonomy approach came out of a designer's instinct for patterns. People aren't labels. They're the specific intersection of where they grew up, what they believe, who raised them, what decade shaped them. I wanted a system that treated that as load-bearing, and that relied on real contributors, not a language model's guess, to fill in the substance.