Blind Spot
A conversational AI that helps you find the gaps in your own thinking — through questions, not answers.
The Challenge
Most AI tools are built to answer. Blind Spot is built to ask.
The problem it addresses is not a lack of information — it is a lack of self-awareness. People hold beliefs shaped by tribal thinking, identity-protective reasoning, and unchallenged assumptions. These are not failures of intelligence. They are blind spots: the things you do not know you do not know.
Existing AI tools make this worse. They confirm, summarize, and generate. They do not push back. They do not notice when you generalize, when your reasoning is circular, or when your framing has narrowed before you have even asked the question.
Blind Spot is designed to do the opposite. It uses the Socratic method to surface the assumptions underneath what you say — not to win an argument, but to help you think more clearly about your own position.
Approach
Blind Spot is built on a clean architectural boundary between two layers.
The Conversation Layer handles the interaction in the moment — real-time voice and text, turn-taking, tone, cadence, and presence. It is designed to feel like a conversation partner, not a tool. This layer is intentionally restartable and replaceable. If the underlying model changes tomorrow, the experience should remain coherent.
The Cognition Engine handles what is happening across time. It tracks frames, beliefs, and tension. It decides when to apply pressure and when to hold space. It selects the type of question — clarifying, counterfactual, scope-limiting, values-reflecting — and passes a structured intent to the Conversation Layer, which then phrases it appropriately for voice and context.
The two layers never collapse into one. Cognition never speaks directly to the user. Conversation never decides what to challenge. That boundary is the core design decision that makes Blind Spot coherent as a product rather than just a prompted chatbot.
The tool tracks a belief ledger across a session — claims the user makes, not a transcript of the chat. It monitors quantifier-level language, detects tribal and identity-protective frames, and modulates pressure based on emotional intensity and openness. The goal is not to debate. It is to invite reflection at the right moment, in the right way.
Outcomes
Blind Spot is in active development and running as a live prototype. Current capabilities include real-time voice conversation, session-based cognition persistence, and a timeline that captures how a user's thinking evolved across a session.
The longer-term goal is more ambitious: a daily thinking companion for people who want to reason better — not a productivity assistant, not a chatbot, but a cognitive mirror. Something that contrasts with every other AI tool by refusing to give you the answer, and instead helping you find it yourself.