A compact, push-to-talk hardware companion that converts desktop AI interactions into a voice-first workflow: press to speak, glance to read. It routes dictation and spoken prompts system-wide into whatever app has the cursor, surfaces short, always-on glances (rate limits, CI status, approvals) on a small AMOLED face, and lets agents push actionable items back to the device so you can respond without breaking flow. The device supports bring-your-own-model or cloud-based backends, keeps audio and state encrypted and on-device by default, and emphasizes privacy and low-latency local-first operation. The pitch is a $25 DIY alternative to $159 AI voice recorders, aimed at restoring quick, distraction-free voice control for modern productivity.
Targets include everyday users who want instant dictation, developers who want voice-driven agent control and deploy approvals, teams that need an AI-era pager with provisioning and SSO, and makers who can build sandboxed WASM apps via a single SDK that runs across every Zeph board. Integrations listed include Claude, Google Gemini, GitHub Copilot, Notion, Obsidian, Perplexity, Ollama, Deepgram and ElevenLabs. Hardware runs on Waveshare ESP32-S3 boards with AMOLED options (2.16" 480×480 and 1.8" 368×448), ships through third-party stores, and offers docs, downloads, and a developer-focused SDK for building portable, sandboxed apps.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.