How the EQ Agent is built
Chat, Podcast, and Voice are three different ways into the same coaching panel. Here's what's actually running behind each one.
The experiences you use
Every mode draws on the same curated base of leadership interviews and frameworks — Dr. Alistair and Sarah, the two personas you talk to, stay consistent across all three.
Chat with Experts
Think through a hard conversation, a team dynamic, or a technical decision with a panel modeled on hundreds of experienced leaders.
- Language models routed per request, chosen for cost and quality
- Live web search for anything outside the curated base
- Conversation history so a thread survives across sessions
Podcast Episodes
A two-host podcast generates a fresh scripted episode from the knowledge base and narrates it to you, segment by segment.
- Dialogue scripted live by a language model per segment
- Distinct neural voices per host
- Opening segments pre-generated for instant playback; the rest streams in just ahead of where you're listening
Live Voice
Talk with Alistair or Sarah in real time — an unscripted, spoken back-and-forth, not a script read aloud.
- Realtime voice model, streamed directly to your device
- Short-lived, single-use session credentials — nothing long-lived ever reaches the browser
- Same knowledge base grounds what they say
Growth & profile
A short personality and EQ profile shapes how the panel responds to you, and an opt-in growth log resurfaces patterns over time.
- Big Five (TIPI) and EQ self-assessments
- Weekly growth snapshots, generated only with consent
- Nothing shared beyond your own account
The platform underneath
The parts that stay the same no matter which experience you're using.
Authentication
Every session is signed in before it can reach a page or generate a response.
Managed database
Profile, history, and consent state live in a managed Postgres database, not on our servers.
Object storage
Generated audio is versioned and stored in cloud object storage.
Edge hosting
The app is deployed on a global edge network for fast response times wherever you are.
Roughly: your request reaches an authenticated page, the page calls a small set of purpose-built endpoints, and those endpoints call out to the language, voice, and storage systems above — each scoped to exactly what that request needs, nothing more.