Pocket FM (Sherpa)

Vibe write your story with Sherpa

Pocket FM (Sherpa)Sep 18, 2026

Vibe write your story with Sherpa

Pocket FM (Sherpa) launch preview
4323603.1K2.5M4.1K

Tags

1M-3MB2CCinematicIndiaGlobalSeries C+Feature launchUSVertical AI

TLVC Rating

4.00
Hook
Editing / Creativity
Copy
Sentiment of launch
Distribution strategy

Community Rating

No ratings yet

to rate this launch.

Sherpa is a fiction-writing platform from Pocket FM designed to help creators build long-form stories spanning anywhere from 10 to 1,000 episodes , and it launches alongside a broader push to make AI the default production layer for serialized audio drama. The tool is aimed at the writers already inside Pocket FM's ecosystem, which reaches more than 250 million listeners across 20+ countries with over 550,000 creators and a library of more than 770,000 audio series . According to founder Rohan Nayak's launch post, Sherpa helped increase content production by 1,200% in a year while ARR moved from $250M to $500M, which is the commercial backdrop this launch is meant to justify. Under the hood, Sherpa uses multiple specialised agents across ideation, research, story planning, prose generation, feedback, and narrative memory, intended to maintain consistency across characters, relationships, conflicts, and storylines extending over hundreds of episodes . Its planning and prose models use Pocket FM's proprietary data, while its assistant layer can draw on frontier models such as OpenAI's GPT and Anthropic's Claude . The pitch to writers, per cofounder Prateek Dixit, is that general-purpose chatbots can draft an episode or two but break down when a story has to hold together across a hundred or more. The platform is currently free for all writers, and Pocket FM does not expect it to become a revenue generator in the near term . Pocket FM was founded in India in 2018 by Rohan Nayak, Nishanth KS and Prateek Dixit, and is headquartered in Bengaluru and Culver City, California, with offices in San Francisco, Berlin and Mumbai . Alongside Sherpa, the parent company Pocket Entertainment is planning to enter Japan and South Korea over the next six to eight months and exploring markets such as the Netherlands, Belgium, Turkey, and the Scandinavian countries , and continues to build Pocket Saga, its AI-generated anime adaptation of existing IP. For operators watching AI in consumer entertainment, Sherpa is a concrete case study in training on proprietary retention data rather than competing with frontier labs on general capability.

Comments (11)

Sign in to join the discussion.
N
Nnamdi Okafor25d ago

1200% content production in a year is either a miracle or a lot of very tired writers pretending an AI helped them. Which one is it?

P
Priya Vaidya25d ago

Minute by minute retention data as training signal is actually the interesting part of this tweet. Everyone else is training on vibes and hoping.

Z
Zeynep Aksoy25d ago

One more thing for the roadmap: let Sherpa suggest the exact minute to drop a cliffhanger based on the retention curve. You're already sitting on the data.

Ł
Łukasz M.25d ago

The thumbnail says 'Sherpa' but the vibes say 'we needed a mascot and the mountain guide was available'.

G
Gia Bertolucci25d ago

Tagline rewrite, on the house: 'Sherpa writes the chapter your listeners can't stop.' You're welcome.

D
Devansh R.25d ago

Building an adjacent tool for indie audio drama writers and honestly this is the wake up call I needed to ship faster. Respect for the retention data moat.

R
Ravi Shankaran25d ago

Reminds me of what one of my portfolio bets is doing for short video scripts, except you actually have the distribution to close the loop. That's the unfair part.

M
Mireille Kadi25d ago

Reporter here, would love to chat about how creators are actually using Sherpa day to day. DMs open whenever.

砂
砂川 tomo25d ago

Fiction AI trained on what keeps people listening is going to produce the most addictively mid stories ever written and I mean that as a compliment.

B
Benoît Aumont25d ago

Curious how creators own the output when the model is trained on aggregated retention behavior across the whole catalog. Asking for every lawyer watching this launch.

M
Mika Solberg25d ago

Serialized fiction listening peaked already and tools like this just squeeze more out of a plateauing audience. Prove me wrong in 18 months.