AI in the Information Desert of Rare Diseases

There are pathologies with fewer than 200,000 diagnosed patients worldwide. For them, clinical trials are scarce, specialists can be counted on one hand per continent, and press coverage is virtually non-existent. In that context, an article published this week in The Kabuki Papers – an independent media outlet dedicated to Kabuki syndrome – poses an uncomfortable and necessary question: who covers the news when no one is willing to cover it?

Kabuki syndrome is a rare genetic disease affecting development, with manifestations ranging from intellectual disability to cardiac and skeletal anomalies. Its estimated prevalence is around 1 in every 32,000 births. It does not generate enough audience for major health media outlets, general medical journalists barely mention it, and affected families have spent years relying on forums, WhatsApp groups and conference PDFs to stay informed.

The gap no one wants to fill

The Kabuki Papers is one of those initiatives born from necessity: an editorial project maintained by and for the community of patients and families with Kabuki syndrome. The linked article – which was picked up this week on Hacker News – does not simply describe the problem. It documents how large language models are functioning as a kind of on-demand newsroom for groups that would otherwise have no access to updated syntheses of scientific literature, conference summaries, or translations of technical studies.

This is not trivial. For a family with a newly diagnosed child, the difference between understanding and not understanding a paper on KMT2D gene variants can have real clinical implications. Current models are capable of processing that document, explaining its conclusions in comprehensible language, and flagging its methodological limitations. They do not replace the specialist, but they do narrow the gap between published research and family understanding.

Who this matters to, and why

The case of Kabuki syndrome is representative of a pattern that repeats across hundreds of rare diseases. Language models with wide context windows like Claude Opus 4.7's million tokens allow the ingestion of extensive documentation – clinical guidelines, longitudinal studies, patient registries – and generate responses grounded in that corpus without needing a permanent editorial team.

This has concrete applications for several profiles:

• Families and patients: access to comprehensible syntheses of recent research without depending on someone to translate or summarise it.
• Patient associations: ability to publish informative content with minimal resources, using AI as a supporting editorial tool.
• Clinical researchers: more efficient dissemination of their results to communities who rarely read indexed journals.
• Independent health journalists: reduced documentation time in hyper-specialised areas.

The limit, of course, is hallucination. Models can confuse genetic variants, misinterpret prevalence figures, or present disputed information with unwarranted confidence. In rare diseases, where the evidence base is small and consensus is fragile, that risk is greater than in well-documented pathologies. The Kabuki Papers seems aware of this: AI use is framed as editorial assistance, not substitution for clinical or journalistic judgment.

A model others should observe

What is interesting about this case is not the technology itself, but its application. There are hundreds of rare diseases – many without any dedicated media, without an active association, without a moderated forum – where this approach could be replicated. The combination of a model with strong reasoning capacity over technical documents, an MCP server connected to databases like PubMed or Orphanet, and a small team with clinical expertise could constitute a functional newsroom for communities that currently have none.

From ClaudeWave we have explored similar integrations in other niche contexts, and the main bottleneck is not technical: it is the absence of an organised community with the will to maintain the project. When that community exists – as seems to be the case around Kabuki syndrome – AI ceases to be an experiment and becomes infrastructure.

The Kabuki Papers article does not claim to be a manifesto or implementation guide. But it honestly describes a real problem and a practical response. It is worth reading, especially if you work in digital health or tools for specialised communities.