Turning Climate Science Into Instant Intelligence: The ClimateBlocks Project

Organisations involved

Main Participant: The Newsroom is a Portuguese media innovation startup developing AI tools that support journalists to navigate complex information and data with speed and clarity. The company leads the design and deployment of ClimateBlocks, bringing together expertise in AI, knowledge engineering and applied climate communication.

The challenge

Climate science is expanding at a pace that outstrips the ability of journalists and analysts to keep up. Research remains fragmented across hundreds of thousands of peer‑reviewed papers, making it difficult to extract clear, reliable insights for publicly available information and content. Existing tools cannot translate this high‑volume, highly complex scientific landscape into transparent, traceable knowledge suitable for responsible reporting.

The Newsroom identified an opportunity to move beyond traditional media-monitoring AI models towards deterministic, evidence-linked knowledge graphs—structured networks that explicitly connect facts and their sources—which could organize and interpret vast repositories of scientific information. This approach promises to enhance the accuracy and transparency of information extraction, enabling researchers and journalists to trace scientific claims directly to their original sources.

Achieving this required large‑scale computational resources to process, analyse and train models on 500,000 papers—far beyond what commercial hardware could handle.  Utilising HPC resourcesenabled the company to build the foundations of a continuously evolving, scientifically grounded climate intelligence engine.

The Solution

The Newsroom developed ClimateBlocks, an AI‑powered application that connects and interprets more than 500,000 climate research papers through a transparent knowledge graph. Using a total of 3,000 node‑hours on GPU accelerated HPC clusters, the team fine‑tuned a specialised large language model on thousands of annotated papers. The system extracts key insights, tracks concept evolution pathways, and identifies domain experts within seconds.

By combining model accuracy with a structured, auditable data backbone, ClimateBlocks delivers rapid, replicable intelligence tailored to the needs of climate journalism.

Impact 

More than 150 journalists already use ClimateBlocks to verify findings and cut research time from hours to seconds, improving the quality and rigour of evidence‑based reporting. A partnership with the European Journalism Centre is accelerating adoption and strengthening resilience against misinformation—an urgent need highlighted by the Declaration on Information Integrity at COP30. By enabling transparent, auditable AI rooted in European HPC, the project supports societal understanding of climate issues and advances Europe’s digital sovereignty.

The Newsroom built their initial AI‑powered knowledge graph of 500,000 climate papers, providing state‑of‑the‑art insights at lower cost and environmental impact than proprietary alternatives. This new capability positions the company to expand its turnover and enter the growing climate intelligence market, serving both media organisations and adjacent sectors.

Benefits

• Trained a specialised AI model 96% smaller than general‑purpose systems, reducing inference costs by 12–17×.
• More than 150 journalists now use ClimateBlocks for instant, evidence‑linked climate insights.
• Secured a strategic partnership with the European Journalism Centre, accelerating adoption.
• Built the first AI‑powered database of 500,000 climate papers with millions of structured links.
• Improved inclusion of Global South research, offering a more diverse and complete scientific perspective.