Affinity Atlas

Affinity Atlas — Evidence-Grounded Cultural Programming

What if cultural programming could be as imaginative as brainstorming, but as transparent as research?

Affinity Atlas is a Qloo-powered cultural research and programming agent that helps creative professionals turn unexpected cultural connections into explainable event concepts.

Instead of simply recommending what people might like, it helps answer a more useful question:

What cultural experience should we develop, why does it fit our creative brief, and what evidence supports that decision?

Inspiration: The problem we wanted to solve

Cultural programmers, festival curators, museums, and creative teams regularly face a difficult challenge: designing original experiences that connect artists, audiences, disciplines, and places.

Creativity is essential, but inspiration alone does not establish whether a connection is meaningful. Generic AI brainstorming can generate compelling concepts without distinguishing real cultural signals from invented associations.

We wanted to explore a different approach: creative decision-making grounded in cultural intelligence, with evidence and uncertainty made visible.

That became Affinity Atlas.

What we built

Affinity Atlas transforms a simple cultural brief into three research-grounded programming proposals.

Users provide:

• An artist: the cultural starting point.
• A city: the geographical context.
• A programming objective: the creative or curatorial goal.

The system researches cross-domain cultural affinities through Qloo, evaluates possible programme formats, and produces three concepts with supporting evidence, editorial reasoning, and limitations.

The user can inspect not only the proposals, but also the reasoning process that led to them.

For example, a programmer can explore Nils Frahm in Madrid with objectives such as a film programme, a sound installation, or an educational workshop.

The objective influences the editorial decision process, while the underlying cultural research supplies evidence rather than imagined relationships.

Why Qloo is essential

Qloo is the cultural intelligence foundation of Affinity Atlas, not a decorative API integration.

Our research process uses Qloo's Taste Graph to discover cross-domain cultural relationships.

The application:

1. Resolves the input artist through Qloo's Search API.
2. Queries Qloo Insights across three domains: movies, brands, and places.
3. Applies geographical context to the place research.
4. Requests explainability information and preserves retrieved cultural entities as evidence.
5. Uses those results to inform subsequent editorial decisions.

This approach creates an important distinction between what Qloo provides and what our application proposes.

Qloo supplies cultural entities and affinity signals. Affinity Atlas interprets those signals into potential cultural experiences.

Without Qloo, the system would lose the cross-domain, source-grounded research that makes its proposals explainable. A generic language model might still invent event ideas, but those ideas would not have the same Qloo-derived evidential foundation.

How the agent works

Affinity Atlas follows a bounded, inspectable workflow:

1. Research

The system resolves the artist and retrieves related cultural entities from Qloo across films, brands, and places.

2. Decide

A deterministic decision engine evaluates predefined cultural-programming formats against the user's objective and the available research.

3. Critique

The application checks format constraints, evidence requirements, diversity, and limitations. It avoids presenting unverified venues as confirmed event locations or cultural brand affinities as commercial partnerships.

4. Revise

An optional external LLM reviewer can select among predefined programme formats. Its output is validated against strict rules.

If the reviewer is unavailable or returns invalid output, the system uses a deterministic fallback rather than trusting unsupported content.

5. Explain

The interface presents programme proposals alongside supporting evidence, reviewer status, editorial decisions, and a research trace.

The result is bounded agentic decision support, not an unrestricted autonomous event planner.

What makes Affinity Atlas different

Most cultural recommendation systems concentrate on identifying things people may enjoy.

Affinity Atlas shifts the focus from recommendation to creative programming decisions.

Its distinctive characteristics include:

• Cross-domain discovery: connecting music with film, brands, and places.
• Goal-aware programming: evaluating creative formats in the context of a specific brief.
• Evidence-grounded concepts: connecting proposals to retrieved Qloo entities.
• Transparent reasoning: exposing decision stages, critique, and supporting signals.
• Responsible uncertainty: clearly separating potential ideas from verified practical facts.
• Resilient AI assistance: retaining deterministic decisions if external model review fails.

The application is designed to support human curatorial judgment, not replace it.

How we built it

Affinity Atlas is a lightweight web application with a Python backend and a responsive web interface.

The backend uses Python's standard library to serve the application and coordinate research requests.

Its main components include:

• Qloo integration: artist resolution and cross-domain insights.
• Concurrent research: retrieval across movie, brand, and place domains.
• Decision engine: predefined formats, editorial selection, and evidence constraints.
• Bounded AI reviewer: structured output validation with deterministic fallback.
• Research trace: visibility into the stages of the workflow.
• Public deployment: a browser-accessible demonstration hosted on Render.

API credentials are managed on the server in public-demo mode rather than embedded in the browser.

At the documented October 14 release checkpoint, the project had 34 passing automated tests, covering its decision logic, reviewer safeguards, and HTTP behavior. This is a historical test result, not a claim about every future deployment.

The source code is published under the MIT license.

Challenges we faced

Making AI creative without inventing evidence

A major challenge was preventing generated ideas from being mistaken for verified facts.

Our solution was to separate cultural research from editorial interpretation. Qloo supplies discoverable entities and signals, while deterministic code controls how those signals appear in proposals.

Balancing flexibility with reliability

A completely open-ended reviewer could produce unsupported formats or inconsistent results.

We therefore constrained the AI reviewer to predefined choices, validated its structured responses, and implemented a deterministic fallback.

Explaining agent decisions

A useful programming concept should be understandable and defensible.

We designed the workflow to expose research evidence, selection reasoning, and limitations instead of presenting recommendations as unexplained answers.

Building a practical public prototype

The application also needed to keep API credentials server-side, handle failures, and provide a usable browser interface without depending on a complex runtime stack.

What we learned

Building Affinity Atlas reinforced an important insight:

Agentic AI becomes more useful when it can explain where its evidence ends and its judgment begins.

We learned that grounding creative decisions in external cultural intelligence requires more than retrieving interesting entities. Those entities need to be interpreted against a human objective, checked against constraints, and presented with clear limitations.

We also found that narrowly scoped AI assistance can complement deterministic safeguards without requiring an unrestricted autonomous agent.

Potential impact

Affinity Atlas is designed for cultural programmers, festival organizers, museums, creative studios, and interdisciplinary arts teams.

It could support early-stage programme research by helping professionals explore connections beyond their own immediate knowledge, compare creative directions, and communicate the reasoning behind a proposed concept.

Its potential value lies in making cultural ideation more exploratory, explainable, and evidence-aware.

We have not conducted a user study or demonstrated quantitative time savings or commercial outcomes. These remain opportunities for future validation.

Responsible use and limitations

Affinity Atlas produces research-grounded creative hypotheses, not verified event plans.

• Qloo affinity signals are not probabilities of event success or percentages of compatibility.
• Suggested places are not confirmed or bookable venues.
• Brand associations do not imply sponsorships or partnerships.
• Accessibility, budgets, permissions, licensing, and availability require independent verification.
• The external AI reviewer is optional and operates within bounded constraints.

Human curators remain responsible for final decisions and real-world verification.

What's next

Our priorities are to further validate the public demonstration, improve the clarity and accessibility of the interface, and assemble reproducible examples showing how different programming objectives influence the resulting concepts.

Longer term, we would like to explore structured feedback from cultural professionals and additional ways to communicate evidence and uncertainty.

Try Affinity Atlas

Live demo: https://affinity-atlas-qloo.onrender.com

Source code: https://github.com/MMLABSHQ/affinity-atlas-qloo

License: MIT

Affinity Atlas is built by M&M Labs for the Qloo Agentic Hackathon 2026.

Evidence first. Imagination second.

Built With

• css3
• html5
• javascript
• openrouter
• pyth
• qloo
• qloo-api