OpenAI Build Week 2026 beta

Built with GPT-5.6 and Codex.

GPT-5.6 provides adaptive reasoning and language. VocoFlo supplies the coaching method: bounded missions, evidence rules, response limits, and a clear direction, so the model guides one real speaking moment instead of drifting into open-ended chat.

VocoFlo Moment Coach

Real-world speaking coaching for people who stutter.

Plan one upcoming speaking moment, try one bounded evidence mission, then report back to compare what fear predicted with what actually happened and take one next step.

Each mission is deliberately bounded: one mission, up to three report-backs, then a clear close. You can start a new mission at any time.

Educational speaking coaching only. This beta does not diagnose, treat, cure, record speech, or assess medical conditions.

Works in this beta

  • Text-based upcoming-moment coaching
  • One AI-generated mission
  • Report-back evidence
  • One next step
  • Local browser history

Planned for VocoFlo

  • Speech recording and speech-event analysis
  • Longer guided learning paths
  • Deeper progress patterns and personalization
  1. Situation
  2. Mission
  3. Report
  4. Next step

The direction

This coach is not trying to make every sentence perfect, and it will not score your speech or hand you a pile of techniques. It helps you notice the loop that can make speaking feel high-stakes: checking, predicting, controlling, correcting, avoiding, and reacting to your speech.

The coach moves you forward through your own evidence. One real speaking moment at a time, it helps you compare what fear predicted with what actually happened, notice whether you continued and how the listener responded, and gradually become less governed by the urge to monitor every word.

In the larger VocoFlo journey, this observing-and-checking movement is later called 'the watcher' — after you have begun to recognise it for yourself.