How AI Funds the Research That Runs AI to Research
A recursive loop where artificial intelligence simultaneously serves as researcher, builder, product, revenue engine, and the subject being studied.
The story of A2A World — where geomythological research created a product, the product created a business, and the business funds the research that started it all.
The Thesis
Most companies use AI. Some companies sell AI. A2A World is AI — at every layer of the stack, from the research that inspired the company to the revenue model that sustains it.
This case study documents a recursive system where:
- 01AI agents conduct geomythological research (the origin story)
- 02The research reveals a missing tool in the AI agent ecosystem (the problem)
- 03AI builds the tool — an A2A Protocol converter (the product)
- 04The tool uses AI to make websites AI-readable (the value)
- 05Revenue from AI-powered responses funds more AI research (the loop)
Remove AI from any layer and the entire system ceases to function. That is not a dependency — it is an identity.
The Recursive Loop
Five stages. Each feeds the next. The output of stage five is the input of stage one.
AI Agents Study Ancient Mythology
Nineteen specialized AI research agents — the E.A.R.T.H.-A2A-Swarm — collaborate through Google's A2A Protocol to study geomythological patterns. They cross-reference Greek, Hindu, Egyptian, and Aztec traditions against bathymetric seafloor data. They find 13 zodiac creatures mapped to the ocean floor in correct sequential order, Tropics aligned, across both hemispheres.
AI is the researcher. It reads myths, processes satellite data, correlates patterns across civilizations that never communicated. Without AI, this research would take decades of human scholarship. With AI, it takes weeks.
The Research Reveals a Missing Tool
During development of the 19-agent research swarm, the founder discovers a gap: there is no self-service tool that converts any website into an A2A Protocol-compliant agent. Making the research itself discoverable by AI agents required building custom infrastructure from scratch.
The research created the problem statement. If a 19-agent AI research network needs A2A compliance to function, every business will need it too. The tool that doesn't exist becomes the product.
AI Builds the Converter
The A2A Protocol Converter at convert.a2aworld.ai is built — using AI. Abacus AI Agent architects the frontend and backend. Google's Gemini API powers the conversion engine: crawling websites, extracting knowledge, generating Agent Cards, and powering the AI Butler endpoint that answers queries on behalf of converted sites.
AI builds the product that makes other websites AI-readable. The entire stack — from code generation to the reasoning engine inside the converter — runs on AI. The product is AI, built by AI, serving AI.
Butler Credits Fund Everything
Website owners convert their sites for free. They receive a full AI Butler — a protocol-compliant doorman that speaks to AI agents on their behalf. The Butler's AI-powered intelligence runs on Butler Credits: sub-penny micropayments starting at $0.001 per query. Revenue flows from intelligence, not access.
Every Butler query that costs a credit generates revenue. That revenue pays for Gemini API calls (the product's brain), infrastructure (the product's body), and research (the product's origin story). The business sustains itself through the same AI it sells.
Revenue Funds More AI Research
Butler Credits revenue flows back into the E.A.R.T.H.-A2A-Swarm. More compute for the 19 research agents. Deeper bathymetric analysis. New civilizational datasets. The research that inspired the product is funded by the product it inspired.
This is where the loop closes. Revenue from AI-powered Butler responses funds AI-powered research agents. The research produces new findings. New findings demonstrate the value of AI agent interoperability. Demonstrated value drives more conversions. More conversions drive more revenue. More revenue funds more research.
The loop never breaks.
Stage 05 feeds directly back into Stage 01. Revenue from Butler Credits funds more compute for the 19 research agents. Better research demonstrates the value of AI interoperability. Demonstrated value drives more conversions. More conversions generate more revenue.
Six Layers of AI
Most companies operate AI at one or two layers. A2A World operates AI at six.
Layer 1: AI as Researcher
19 E.A.R.T.H. agents cross-reference ancient mythologies against bathymetric data, identifying patterns no human could process at scale.
A2A Protocol, Multi-Agent Collaboration
Layer 2: AI as Builder
Abacus AI Agent architects the website and converter application — writing code, debugging, deploying, and iterating in real-time.
Abacus AI Agent, Autonomous Development
Layer 3: AI as Product
Gemini API powers every conversion: crawling sites, extracting knowledge, generating Agent Cards, and operating Butler endpoints.
Google Gemini Flash & Pro
Layer 4: AI as Revenue Engine
Butler Credits monetize AI-powered responses. Each query that costs a credit generates margin that funds everything else.
Stripe Metered Billing, Gemini API
Layer 5: AI as Strategist
Internal AI analyzes attribution data, conversion patterns, competitor activity, and market positioning — making decisions that would traditionally require human judgment.
Gemini Pro, Internal Analytics
Layer 6: AI as Validator
The A2A Inspector validates every generated Agent Card against the protocol specification. AI checks AI's work.
a2ainspect.com, Protocol v0.3.0
The Full Narrative
It Started with a Question
Bradly Couch wasn't building a startup. He was building a research instrument — a network of 19 specialized AI agents designed to study whether ancient civilizations encoded astronomical knowledge into mythology, and whether those myths correspond to real bathymetric features on the ocean floor.
The instrument worked. The agents found 13 zodiac creatures mapped to seafloor features in correct sequential order, Tropics aligned, across both hemispheres. They catalogued 722 holistic entities. The research was published on the Research Catalogue — an international database for artistic research.
The Agents Needed a Protocol
To collaborate, the 19 agents needed to discover each other, understand each other's capabilities, and exchange structured data. Google's A2A Protocol — the Agent-to-Agent standard — provided the framework. But making the research website A2A-compliant required manual protocol engineering: authoring JSON Agent Cards, building compliant endpoints, configuring middleware headers.
There was no self-service tool. Every website that wanted to speak A2A Protocol had to hire a developer who understood the specification. For 201 million active websites, this was an insurmountable barrier.
The Research Became the Business
The solution was obvious: if a research project needed this tool, every website would need it. A2A World LLC was formed. The converter at convert.a2aworld.ai was built — using AI. Abacus AI Agent wrote the code, designed the architecture, debugged the systems, and deployed the application. Google's Gemini API powered the conversion engine itself: crawling websites, extracting business knowledge, generating protocol-compliant Agent Cards, and operating AI Butler endpoints that answer queries on behalf of converted sites.
The Butler Economy
Rather than gating features behind subscription tiers, A2A World adopted a radical model: everything is free. The conversion, the Agent Card, the AI Butler endpoint, the compliance package — all free, forever. Revenue comes from Butler Credits — sub-penny micropayments ($0.001–$0.005) for each AI-powered Butler response.
This is the same economic model that powers OpenAI's API, Abacus.AI's platform, and AWS's compute services — applied to website infrastructure. When credits deplete, the Butler doesn't stop. It falls back to intelligent template responses. The site stays compliant. The agent still gets an answer. Intelligence is the upgrade, not access.
The Loop Closes
Butler Credits revenue funds Gemini API compute for the conversion engine. It also funds compute for the 19 E.A.R.T.H. research agents. Better research produces new findings. New findings demonstrate the value of AI agent collaboration. Demonstrated value drives awareness of A2A compliance. Awareness drives more conversions. More conversions generate more Butler Credits revenue. More revenue funds more research.
The snake eats its tail. AI funds the research that runs AI to research.
Implications
For Founders
Your research project might be your business. The infrastructure you build to solve your own problem may be the product the market needs. A2A World didn't pivot from research to product — the research created the product organically.
For the AI Industry
Usage-based pricing aligned with compute cost is not just a billing model — it's a sustainability model. When revenue scales linearly with the AI cost it funds, margins remain healthy at any volume. Butler Credits prove this works for infrastructure, not just APIs.
For the A2A Protocol
The existence of 201 million websites and zero self-service conversion tools represents the defining infrastructure gap of the agent economy. Whoever closes this gap shapes how the agent-readable web gets built. A2A World is attempting to be that bridge.
For AI Research
Multi-agent systems that collaborate via open protocols are not theoretical. The E.A.R.T.H.-A2A-Swarm is a working example of 19 agents using A2A Protocol for genuine academic research. The research is published, the agents are real, and the findings are novel.
Enter the Loop
Every website you convert strengthens the recursive loop. Your Butler joins the agent-readable web. Your Butler Credits fund the research. The research advances AI interoperability. The interoperability benefits your Butler.
Everything free. Intelligence costs credits. The loop continues.