Versus a generic podcast
A topic can be requested on demand, its treatment can use personal context, and when it ends you can keep exploring without starting over.
How it works
About
Five Cents is an app for personalized podcasts generated with AI. It turns a question, a news story, or a destination into a short podcast that is researched and written for listening. It also learns from user interactions to improve personalization and editorial quality day by day.
From intent to a podcast ready to play
- 01Question and profile
- 02Intent and research
- 03Script and evaluation
- 04Voice and publishing
- 05Feedback and continuity
Personalized podcasts
A team of agents for every request
One agent interprets intent and format; another gathers context and research; the writing agent turns that material into a spoken narrative; and evaluation agents review evidence, structure, clarity, and alignment with the request before the voice layer produces and publishes the podcast.
Personalization combines the listener's question with selected and free-form interests, language and edition, chosen voice, listening history, and context from related podcasts. As the listener uses Five Cents, feedback, progress, completions, saves, and shares enrich a profile that helps decide what to research, which angle to take, and how to explain it better next time.
Learning and interactive continuity
Context connects one podcast to the next
After listening, Five Cents suggests questions that expand the topic through a consequence, practical case, comparison, or deeper angle. The listener can choose a suggestion or write a new question, while the system uses the previous podcast as context so the conversation can continue without starting over.
Written or dictated feedback and signals such as progress, completions, saves, and shares help reveal what matters and how to explain it better. Individually, they enrich personalization; in aggregate, they guide the Popular catalogue and future versions of prompts, structures, duration, and tone. A continuity check compares related scripts and requires a new angle when it detects too much repetition.
Weekly content
Your news, researched for you
Every Monday, a group of agents researches current events and produces two formats for every listener: Topics of the Week, which explain one story with context, and six news channels with their own editorial structures.
Selection is compared with previous weeks to avoid repetition; when a story continues, it must add a real development or a new angle. Specialized evaluation agents review source coverage, relevance, spoken structure, and compliance with each channel's format. Automatic retries, traceability, and aggregate feedback help preserve a recognizable voice and improve future editions.
Travel guides
Expert knowledge turned into reproducible routes
Travel guides combine a shared editorial structure with destination-specific knowledge built from recommendations by expert travellers and long-term residents. The result follows a familiar format without forcing different destinations into the same route.
A route agent identifies bases, stops, days, and connections from the script; grounds every point in textual evidence, validates coordinates, and generates both the map and the day-by-day itinerary. When destination-specific knowledge is unavailable, the system reuses the editorial structure—never facts from another place.
Continuous optimization
Every script helps improve the next version
The system analyzes generated scripts alongside explicit feedback and listening signals from users. Evaluation agents identify patterns in quality, clarity, usefulness, coverage, and repetition; those findings are used to update versioned prompts for research, selection, writing, and review. The system then regenerates representative scripts, evaluates them again, and compares the results. This loop improves agents and formats with evidence, preserves the history of every change, and makes it possible to restore an earlier version when an update does not perform better.
- 01Scripts and feedback
- 02Evaluation
- 03Versioned prompts
- 04Regeneration
- 05New evaluation
Agent infrastructure
An observable architecture from end to end
Each stage records its status, agent version, timing, response identifiers, and cost. This makes it possible to locate failures, compare editorial changes, and measure generation quality. Weekly content is produced once for the full audience, while personalized content is generated on demand: two distinct flows running on the same measurable infrastructure.
Distribution
Five Cents on other platforms
A selected part of the Five Cents editorial catalogue is published beyond the app to reach new audiences. Current-affairs stories, travel guides, and popular episodes are adapted to each channel with their audio, artwork, and description; the app remains the complete, personalized, and interactive experience.
Why it matters
More personal than a catalogue. More editorial than a chatbot.
Versus a generic chatbot
There is research, spoken structure, a versioned editorial line, continuity controls, and a product designed to work without looking at a screen.
Without always choosing what comes next
Topics of the Week, news, Popular, and travel guides create a curated feed the listener can simply open and play.
