Founder-engineer case study · 2026
VocalCode.
Local push-to-talk dictation for AI coding. Hold a configured key or mouse button, speak, and release — clean text lands in the active field in most desktop apps.
The product
AI coding agents work best with context-rich prompts, but typing several paragraphs interrupts the work. VocalCode reduces the interaction to three deliberate steps:
It is a focused dictation tool, not a meeting recorder or a general voice assistant.
What I built
I designed and shipped the desktop product, local speech pipeline, cross-platform input layer, settings experience, licensing service and release automation.
Privacy boundary
Recognition audio and transcripts stay on the user's machine. Internet access is still used where the product genuinely needs it: initial trial provisioning, model and update downloads, checkout, activation and periodic paid-licence validation.
The public privacy notice separates those network operations from the local recognition path instead of reducing the claim to a vague “offline” label.
Release discipline
Current releases are distributed through direct, versioned downloads plus maintained package manifests. The release pipeline verifies platform builds, signatures, notarization, update metadata and the public website before promoting a release.