Pocket FM launches Sherpa, an AI fiction tool with 89.8% story memory vs 57.4% baseline
testingcatalog · x · 2026-09-19
Audio-fiction platform Pocket FM launched Sherpa (beta), an AI tool for writing long serialized stories. The underlying model was reportedly trained on 100M+ hours of retention, coin-spend, and drop-off data; on the company's own 176 cross-chapter questions about characters, event order, and plot, its story memory scored 89.8% vs 57.4% for a graph-memory baseline.
Sherpa ships with purpose-built agents: a Planner keeps characters, arcs, and episode plans in context, plus storyboard, feedback, and localization steps. Drafts can be localized into 7 languages, narrated with one voice per season, and published directly as Pocket FM episodes.
More from Multimodal
- fal launches H3 Max Lip Sync: photo + audio to lip-synced video in 11 seconds — jfischoff · 2026-09-19
- Redditor shares AI hunting short film made with Seedance 2 — Ok-Nerve941 · 2026-09-19
- Microsoft's MoGe-3 3D geometry model lands in ComfyUI with point, depth and normal outputs — petewoodbridge · 2026-09-19
- Training an LLM to paint by writing code: RL meets creative tasks with a judge model — measure_plan · 2026-09-19
- Astra Shows Off Impressive Comic Creation Skills — Algoartist · 2026-09-19
- Poet's Workflow: Feeding Astra Scene Bookends Plus Anchor Lines Yields Night-and-Day Lyrics — HeIsYour · 2026-09-19