An autonomous multi-agent ecosystem that transforms daily news into structured, debate-driven podcast episodes.
The Autonomous Debate Ecosystem (ADE) is an AI-native editorial and podcast production framework designed to simulate the workflow of a real debate podcast.
demo link: https://infinite-debate.vercel.app
Rather than generating conversations directly, ADE models the complete production pipeline:
- Discover daily news.
- Extract factual knowledge.
- Generate multiple debate perspectives.
- Evaluate editorial quality.
- Select unique discussion topics.
- Direct multi-agent debates.
- Build structured argument flows.
- Detect viral moments.
- Produce fully voiced podcast episodes.
The project explores how autonomous AI agents can collaborate to perform the roles of researchers, editors, hosts, debaters, producers, and audio engineers.
Daily News
↓
Editorial Intelligence
↓
AI Podcast Director
↓
Multi-Agent Debate
↓
Argument Intelligence
↓
Viral Intelligence
↓
Audio Production
↓
Podcast Episode
The first stage discovers and prepares debate topics.
- Fetch >10 daily articles.
- Store article metadata.
News API
↓
Articles
Each article is processed by an LLM to extract:
- Facts
- Entities
- Themes
Article
↓
LLM
↓
Facts
Entities
Themes
Each article generates multiple possible debate angles.
- Philosophical
- Controversial
- Viral
- Emotional
Each mode can be adjusted using:
- Heat
- Chaos
- Intensity
Input:
Facts
Entities
Themes
Mode
Intensity
Output:
~10 Articles
×
4 Modes
=
~40 Candidate Debates
Every candidate debate is evaluated using:
- Original article
- Extracted facts
- Debate topic
Scoring dimensions:
| Metric | Description |
|---|---|
| Faithfulness | Consistency with source |
| Relevance | Importance of discussion |
| Controversy | Potential disagreement |
| Framing | Quality of debate setup |
| Coherence | Logical structure |
| Overall | Aggregate quality |
Average scores are calculated for each candidate.
Selection rules:
average > 9
selected == 0
Group by original article title.
Each article can only contribute one debate topic.
One candidate is randomly selected and marked as used.
This allows multiple unique debate episodes to be generated from a single daily news cycle while preventing repetition.
For demonstration purposes:
1 Scraper
↓
1 Debate
↓
1 Podcast Episode
This constraint exists primarily because of voice synthesis costs.
The AI Podcast Director is responsible for transforming a news topic into a structured debate experience.
Rather than simply coordinating speaker turns, the director performs editorial planning before the discussion begins.
For every episode, an LLM creates a custom debate package based on the selected topic and its factual background.
The package defines:
Topic + Background
↓
Debate Package
The generated package contains:
- Central question
- Core conflict
- Real-world stakes
Example:
Should AI-generated content replace human creativity?
Central Question:
Can AI democratize creativity without harming artists?
Key Conflict:
Innovation vs creative ownership.
Stakes:
Economic, cultural, and ethical consequences.
Instead of fixed personalities, each participant receives a topic-specific briefing.
Each debater is assigned:
- A stance
- A debate role
- Strategic objectives
- Supporting evidence
- Preferred rhetorical techniques
- Arguments to avoid
- Likely opponent strategies
For example:
Alex
↓
Defend technological progress
Sarah
↓
Defend human and cultural values
The exact positions change depending on the debate topic.
The moderator also receives a private briefing.
This includes:
- Opening questions
- Follow-up prompts
- Areas of tension
- Pressure points for both sides
The moderator uses these signals to maintain pacing and encourage deeper discussion.
During the episode, the Podcast Director continuously evaluates debate state and decides:
- Who should speak next.
- Whether clarification is needed.
- When to challenge weak arguments.
- When to transition topics.
- How to maintain balance and engagement.
The director acts as the producer of the debate rather than as a participant.
The debate itself is performed by multiple autonomous agents.
Unlike traditional role-playing systems, speaker identities are not permanently fixed.
Each episode generates new strategic briefings tailored to the selected topic.
Debate Package
↓
Role Briefings
↓
Multi-Agent Discussion
Alex typically approaches problems from an analytical perspective.
Depending on the debate package, Alex may become:
- A policy advocate
- A technology optimist
- A legal analyst
- A scientific skeptic
- An economic pragmatist
Alex follows the assigned strategy while adapting to new arguments during the discussion.
Sarah generally represents human-centered perspectives.
Depending on the topic, Sarah may become:
- An ethicist
- A social advocate
- A cultural critic
- A community representative
- A defender of individual rights
Her objectives are dynamically generated for every episode.
Marcus serves as moderator.
Rather than following a scripted sequence, Marcus receives editorial guidance and reacts to the evolving conversation.
Responsibilities include:
- Introducing the debate.
- Asking prepared and spontaneous questions.
- Challenging both sides.
- Managing pacing.
- Summarizing key disagreements.
- Maintaining neutrality.
Every speaking turn is generated using multiple context layers:
Topic + Background + Debate Framing + Role Briefing +
Conversation Memory + Previous Speaker + Current Task
↓
Agent Response
This allows participants to:
- Stay consistent with their assigned roles.
- Reference previous arguments.
- Adapt to strong counterpoints.
- Maintain long-form coherence.
- Avoid repetitive exchanges.
The result is a debate where strategies are generated for each episode rather than hardcoded into the system.
The debate is represented as an evolving argument graph.
Claim
├── Support
├── Attack
└── Summary
This structure enables:
- Better conversational memory.
- Reduced repetition.
- Stronger debate coherence.
- Easier post-processing.
The system identifies high-impact moments.
Potential signals include:
- Strong disagreements.
- Emotional peaks.
- Memorable quotes.
- Unexpected perspectives.
- High-engagement exchanges.
These moments can later support:
- Episode highlights.
- Shorts.
- Social media clips.
- Viral scoring.
The completed debate enters the production pipeline.
Transcript
↓
Voice Synthesis
↓
Audio Chunk Generation
↓
Audio Assembly
↓
Podcast Episode
Outputs include:
- MP3 audio
- Transcript
- JSON metadata
- CSS styling
- Episode assets
The Autonomous Debate Ecosystem follows several principles.
Topics should be curated, not randomly invented.
Conversations are more engaging than summaries.
Argument graphs maintain logical consistency.
Different agents provide competing viewpoints.
Independent AI components collaborate to produce a complete debate podcast.
cd backend
cp .env.example .env
# Edit .env and set relavant contentcd backend
npm install
npm run dev # provide api for frontend to display audio
npm run scraper # to scrap and generated multiple news debate topic
npm run programme # to generate debate conversation and audio
npm run generate # run scrap and generate 1 topic for debate conversation and audio
In new terminal do backend set up:
cd frontend
npm install
npm run devThe Autonomous Debate Ecosystem explores the idea that podcast production can become an autonomous editorial process.
Instead of simply generating text or speech, the system coordinates specialized AI agents that perform the roles of journalists, editors, moderators, debaters, producers, and audio engineers.
The result is a fully automated pipeline capable of transforming daily news into structured, engaging, and debate-driven podcast episodes.