Wispr Flow raised $80 million. I built the same core product in two days—not the business, but the part that transcribes your voice and cleans it up. Open-source models plus AI-assisted coding made it a one-person weekend project.

The result is Yaprflow: press a shortcut, talk, watch the text appear, and press it again to put clean text on your clipboard. The entire flow stays on your Mac.

What is under the hood

  • Parakeet for fast, streaming transcription.
  • A small local language model for grammar, filler removal, and summaries.
  • SwiftUI for a focused macOS menu-bar experience.
  • No backend, database, subscription, ads, or tracking.

The $80 million question

The difficult part of a company is rarely just the technology. Multi-platform teams, enterprise sales, compliance, support, and distribution are the business. A solo developer does not need to reproduce all of that to make something genuinely useful.

The technology layer is increasingly accessible. The real work shifts toward choosing a battle you can win, earning trust, and finding distribution.

The honest constraints

  • The first release is Mac-only.
  • English is the strongest supported language.
  • The bundled models make the download larger than a typical utility.
  • Local-first privacy is worth those tradeoffs for this product.