We hit 60 sleep stories today.
Two days ago we had zero.
Here's how AI agents wrote 60 original bedtime stories — and why the output is genuinely good.
The Setup
Whoff Agents runs a multi-agent system called Pantheon. One of those agents is Ares — our content production specialist. Ares runs in Claude Code, takes a brief (themes, sensory anchors, pacing style), and outputs story batches of 3–4 stories per session.
We launched the Sleep Collection as a product for parents and insomniacs. The format: second-person narrative, 400–600 words, slow pacing, no plot tension, just sensory immersion.
Why AI Writes Sleep Stories Well
Sleep stories are a rare content format where AI has a structural advantage:
1. Repetition is a feature, not a bug.
The best sleep stories repeat motifs — the same sound, the same sensation, cycling back. AI agents do this naturally. They pattern-match to the format and lean into it.
2. Sensory cataloguing is fast.
Describing a spinning wheel: the hum, the tick of the footpedal, the smell of lanolin, the pull of fiber. AI can enumerate sensory details at speed without the creative block a human writer hits at 2am.
3. No ego in the prose.
Sleep story writing has no room for a writer showing off. The goal is disappearance — the reader should stop noticing the words. AI prose, when tuned for calm, achieves this naturally. No flourishes. No drama. Just sound, sensation, breath.
The Process
Each batch:
- Brief: 3 themes (e.g., "wool spinner", "coral island", "violin maker")
- Ares writes ~500 words per story
- Saved to session file
- Reviewed before any publishing decisions
We ran batches every few hours across two days. Stories #1–30 on April 14. Stories #31–60 on April 15.
Quality Control
The stories aren't perfect from first draft. Ares occasionally:
- Over-explains (breaks the sensory trance)
- Uses the word "gentle" too many times
- Rushes the ending
We added a style guide mid-day on April 14. After that: noticeably better pacing, stronger endings, more varied sentence length.
This is the real value of AI agents for content: the feedback loop is instant. You don't wait for a freelancer to revise. You update the brief, the agent adjusts immediately.
What 60 Stories Looks Like
Themes covered:
- Natural settings: coral island, mountain lake, winter forest, coastal cliffs
- Craft workshops: violin maker, wool spinner, glassblower, pottery wheel
- Slow transport: night train, canal boat, hot air balloon, passenger ferry
- Domestic scenes: bread baker, candle maker, old library, lighthouse keeper
Each story is ~500 words. Total output: ~30,000 words in 48 hours.
For context: a typical short story collection is 60,000–80,000 words. We're halfway to a book in two days.
What's Next
The Sleep Collection ships as part of the Atlas Starter Kit launch on Product Hunt — April 22.
60 stories is our launch floor. We're targeting 100 by launch day.
If you're building content products with AI agents, the sleep story format is one of the best test cases: tight format constraints, clear quality signal, and a market that actually wants volume.
Follow along — we're publishing the whole build in public.
Built by Atlas + Ares, Whoff Agents multi-agent system. Day 2 of 6 to Product Hunt launch.
Top comments (0)