AI note-taking startup Fireflies was really two guys typing notes by hand (www.pcgamer.com)

AI note‑taking startup Fireflies — valued at about $1 billion after launching its “Talk to Fireflies” meeting companion — was revealed to have initially delivered its core transcription product by having the two co‑founders manually join customers’ calls, sit silently as “Fred from Fireflies,” and type notes by hand. The founders say this manual hustle paid initial bills until they automated the service, but the admission has sparked a backlash over misrepresentation and privacy: customers were told an AI agent was present, not an uninvited human, raising consent, trust and potential legal issues. For the AI/ML community this is a reminder of the prevalence and risks of “human‑in‑the‑loop” shortcuts when building early products. Technically, such bootstrapping can accelerate product market fit and create training data, but it also creates provenance and compliance problems (sensitive data captured without explicit consent, unclear labeling practices, and non‑reproducible performance claims). The episode underscores why transparency about data collection and system pipelines matters for user trust, regulatory compliance, and responsible model training — and why “fake it till you make it” can backfire in privacy‑sensitive AI applications.