Recently reworked the shortcut recorder for Odett. The goal was to make recording a shortcut feel smoother and more reliable, while clearly showing which modifier keys are available. Designed in @figma. Built with @OpenAI Codex inside @cursor_ai.

Today we’re releasing Odett. We built it because dictation still feels too much like something you have to manage. Before you speak, you often have to choose the right mode, language, format, or preset for what you’re about to write. Slack reply. Email response. Norwegian. English. Cleaned up. Formatted properly. That setup gets in the way of the actual thought. Odett is our attempt at making voice-to-text feel like part of your actual workflow. Press a shortcut, speak naturally, and Odett writes where you already are. Choose up to two languages and move between them naturally while you dictate. No language switching needed. It cleans up your text as you speak, so messy thoughts become clearer writing without extra steps. And when you want more control, just say “Hey Odett” while dictating and tell it what you want the text to become. Format the text, turn rough notes into a plan, draft an email response, enrich a prompt for Codex or Claude Code, or whatever else you need. No mode switching. A simple UI. Powerful dictation out of the box. We’ve also made the demo generous, so you can properly test it before deciding if you want to use it. Here’s a 3-minute look at Odett.

I've been working on this personal project that started off wanting to explore progressive blurs, and eventually ended up creating my own interpretation of the DNB Spare app with a floating bottom navigation that morphs into a search bar.