Fully Informed — Several AI models answer, argue, and a judge gives one verdict with the disagreements shown.
Multi-model decision intelligence. Frontier models each answer a question, debate one another, and a judge model delivers one verdict with the disagreements shown, not hidden.
The problem
A single model can be confidently wrong without saying so, and comparing models means copying the same prompt across several tabs and judging the answers yourself.
What we built
A fan-out, debate and verdict pipeline streamed live to the user, a judge that reports what every model missed, image models judged by an art-director model, and pay-as-you-go billing at model cost plus a margin.
What it does today
One answer you can weigh: what each model said, where they disagreed, and what all of them missed — paid for by use, with no subscription.
— Under the bonnet
What makes a product like this hard to build
- 01
Many models at once
Every question goes to several frontier models in parallel, each with different APIs, limits, speeds and failure modes. One slow or failing model can't hold up the answer.
- 02
A fair debate
Models answer, read each other's answers and respond. Keeping that structured, bounded and readable, rather than an endless loop, is a design problem as much as a technical one.
- 03
Judging, not averaging
A judge model has to weigh the arguments and report the disagreements and what everyone missed, without simply picking the most confident answer.
- 04
Streaming and cost control
The debate streams live, and every token costs money. Usage is metered per request and billed at model cost plus a margin, with no surprises for the user.
— Estimated build cost
What a build like this costs
£15,000–£20,000
Multi-model AI reasoning product · indicative range for a comparable build
Sized from the codebase we built: 17k lines of production code, 22 database tables, 37 API routes.
A product like Fully Informed is a bespoke build, not a template: most of the cost is in the hard parts above, plus testing against real data before anyone relies on it. Where your version lands in the range depends on the factors listed here. You get a fixed price once scope is agreed — usually after a short discovery sprint. Hosting and AI model usage are billed at cost on your own accounts.
Get an estimate for yoursWhat moves the price
- Number of model providers supported
- Depth of the debate (rounds, judges, image models)
- Live streaming interface
- Usage metering and billing
- Rate-limit and failure handling across providers
Built with
- TypeScript
- Fastify
- PostgreSQL
- OpenAI-compatible model APIs
- Stripe
Designed, built and run by our Manchester team. Want something similar? See our AI development services, AI automation and pricing.
