The Vibe Coder’s Manifesto: Transforming Complex National Health Data into Accessible Public Insights

In the digital age, government data is often described as a "treasure trove," yet in practice, it frequently resembles a labyrinth. The Philippine Statistics Authority (PSA) recently released the 2025 National Demographic and Health Survey (NDHS), a monumental report tracking critical indicators ranging from maternal care and childhood mortality to intimate partner violence and fertility rates. While this report is arguably the most vital health document in the country, the public discourse surrounding it has been remarkably narrow. For over a month, headlines have focused almost exclusively on a single statistic: the national total fertility rate dropping to 1.7, below replacement level.

While this demographic shift is a headline-worthy national story, the true depth of the NDHS remains largely locked away in dense, slow-loading Excel spreadsheets. Bridging this gap between raw data and public understanding usually requires a team of data scientists, web developers, and graphic designers. However, in a groundbreaking experiment, one journalist managed to build a comprehensive data visualization and analysis dashboard in just one week—all through the nascent practice of "vibe coding."

The Genesis of a Digital Dashboard

"Vibe coding"—the process of describing desired outcomes to an Artificial Intelligence coding agent and allowing the AI to generate the underlying programming—has fundamentally altered the barrier to entry for data journalism. The project began not with the intent of creating a full-scale dashboard, but as a technical experiment to create interactive, drill-down maps of the Philippines.

Utilizing a GitHub repository maintained by Filipino developer James Faeldon, which maps administrative boundaries, the project sought to allow users to navigate from the national level down to regions, provinces, and finally, municipalities. By utilizing Claude Code, the journalist translated a vision of hierarchical maps into reality. The process, however, was not without its "growing pains."

The initial hours of development were characterized by the "ninety-ninety rule" of programming: the first 90% of the code consumed the first 90% of the development time, while the remaining 10% consumed the other 90%. Simple geographic complexities—such as Highly Urbanized Cities (HUCs) that are geographically embedded within provinces but politically independent—caused the AI to render maps that looked like jigsaw puzzles with pieces floating in the sea. It required manual intervention, a deep understanding of local administrative structures, and constant iteration with the coding agent to refine the visual output.

A Chronology of "Vibe Coding" Development

The development cycle was a rapid succession of pivots and expansions. Once the base map template was stabilized, the project moved quickly from proof-of-concept to a data-driven tool:

• Phase 1: Data Extraction. Using the coding agent, the project moved away from manual cleaning. By pointing the AI toward PSA spreadsheets—ranging from poverty incidence to GINI coefficients—the agent successfully extracted and structured data, automatically identifying hierarchical patterns without the need for traditional, cumbersome data manipulation.
• Phase 2: The NDHS Pivot. Recognizing the potential of the NDHS dataset, the project shifted focus. What began as a map project grew to include tab systems for teenage pregnancy, vaccination status, and contraceptive preferences.
• Phase 3: Editorial Refinement. As the data grew, so did the design challenges. To avoid overwhelming the user, tables were split into logical segments. The "vibe coding" allowed for near-instant design pivots, such as switching from standard bar charts to cleaner, more effective lollipop charts with a single instruction.
• Phase 4: Architectural Standardization. As the project expanded, the initial "messy" code generated by AI became a bottleneck. The developer introduced a structured architecture—separating the data layer from the rendering layer—which increased efficiency and reduced token usage. By the end, the project transitioned to utilizing more cost-effective models like GPT-5.3-codex, proving that high-level AI orchestration can be both powerful and economical.

Supporting Data and the Power of Visualization

The dashboard currently serves as a bridge for complex information that would otherwise remain dormant in government portals. By transforming raw spreadsheets into interactive choropleth maps, the tool allows users to visualize variation across regions at a glance.

The inclusion of "waffle charts" to represent contraception preferences among Filipino women demonstrates how sophisticated data visualization can distill sixteen different dimensions into a user-friendly interface. Furthermore, the dashboard doesn’t just present numbers; it incorporates AI-generated insights. These insights are designed to highlight patterns that might be missed by a casual observer, such as specific correlations in vaccination status or teenage pregnancy rates across different administrative divisions.

Ensuring Accuracy: The Fact-Checking Pipeline

The greatest criticism of AI-generated content in journalism is the risk of "hallucination." To mitigate this, the project implemented a rigorous, multi-layered auditing pipeline. Initially, the system attempted to check individual claims, but the process proved repetitive.

The developer then pivoted to "block-level" fact-checking. Each insight generated by the AI was scored against the original dataset as "supported," "mostly supported," "partly supported," or "unsupported." Strong guardrails were established: every evidence line was required to point to an exact data pointer and value in the source document. If a piece of data could not be verified, the output was rejected and re-run. In the final audit, 273 blocks were verified, with the vast majority showing high levels of support, demonstrating that when paired with a strict audit pipeline, AI can be a reliable tool for data-driven storytelling.

Expert Perspectives and Official Implications

The potential for this tool has already garnered attention from key stakeholders in the health and sociology sectors. Dr. Romelei Camiling-Alfonso, a health innovation expert, noted that the dashboard holds significant value for "doctors to the barrios" and municipal health officers who are on the front lines of primary healthcare delivery.

Sociologist Athena Presto praised the "low-sensory" design of the interface, noting that it allows users—whether they are researchers or the general public—to access critical health information without needing to navigate complex, high-friction government websites. Tricia Aquino of PumaPodcast emphasized the collaborative potential of the project, noting that by making the NDHS interactive, newsrooms can now find connections and patterns that were previously invisible.

The Future: Toward a More Accessible Democracy

The implications of this experiment are profound. If a single journalist can build a high-functioning, fact-checked health dashboard in one week using "vibe coding," the potential for civil society to hold governments accountable and democratize data access is unprecedented.

The next phase of the project envisions a conversational layer. By layering a chatbot on top of the existing architecture, the dashboard could soon allow users to ask natural language questions—such as "Which region has the highest rate of unmet family planning needs?"—and receive answers grounded in verified data.

However, the creator remains clear-eyed about the limitations. While "vibes" and AI agents can take a project from zero to one hundred with remarkable speed, they do not replace the need for an editor’s eagle-eye. Human judgment remains the final arbiter of design, editorial tone, and, most importantly, the integrity of the data.

This project stands as a testament to the fact that the tools of the future are here, and they are accessible. By combining the speed of AI with the traditional rigors of journalism, the "vibe coder" has provided a blueprint for a new era of data transparency—one where the most important reports in the country are not just published, but truly understood.