Keep pulling the thread on Tanay Kothari.
Wispr Flow pivoted from being a hardware company to its current software-as-a-service model.
Wispr Flow's predecessor company spent three years with a team including 40 PhDs to successfully build a wearable device that could translate thoughts to text.
In August 2024, Wispr Flow pivoted to focus entirely on its software product because management concluded that building a "habit of voice" was a prerequisite for hardware adoption.
Following its pivot from hardware to software, Wispr Flow reduced its team size from 40 people to approximately five.
Tanay Kothari asserts that Wispr Flow's voice model is the best in the world across 80 languages when benchmarked on latency and accuracy.
Wispr Flow's primary internal metric is building the "habit of voice" for its users, rather than tracking ARR or total user count.
Tanay Kothari claims Wispr Flow is the only company that has built a "dictation model," which translates speech to intended written text, as opposed to a "transcription model" that performs literal speech-to-text.
The most difficult part of Wispr Flow's technical stack is its contextual engine, which learns a user's short- and long-term context to build a personalized model.
Tanay Kothari believes Wispr Flow has the highest retention numbers of any product for both paying users and overall usage.
The long-term vision for Wispr Flow is to evolve into a "Jarvis"-like proactive assistant that can write, perform actions, and offer help across all applications.
By implementing a single-decision-maker model, Wispr Flow is running 20 parallel projects with a 25-person team and has increased its development velocity by 2 to 2.5 times in the last couple of months.
Tanay Kothari believes high churn from "AI tourists" is a major problem in the AI software industry, making retention a more critical metric than ARR growth.