14-Year-Old Dev Ships Single-Model Dictation App Running on 4GB VRAM

A 14-year-old developer replaced the two-stage Whisper+LLM dictation pipeline with a single model, achieving up to 36x speedup and 0.5-second output on a 4GB RTX 3050 Ti, and open-sourced the app Typd.

2026-11-14 ~ 2026-11-14 · 2 related posts

• Teen dev replaces Whisper + LLM dictation pipeline with one 1.7B model, 0.5s vs 36s — Mehmet_dedeler · 2026-11-14
• A 14-year-old trained a single model to replace the Whisper+LLM dictation pipeline: finished text in ~0.5s on 4GB VRAM — Mehmet_dedeler · 2026-11-14