• Completed
  • Machine Learning / Deep Learning
  • Service Planning and Development
  • News Analysis
  • Hackathon

Maeil Business Newspaper × Anthropic AI Hackathon

We designed and built an AI service connecting the timeline of news events with stakeholder relationships, reaching the finals of the Maeil Business Newspaper × Anthropic AI Hackathon.

2026 · Completed

Turning the experience of reading news into knowledge

At the Maeil Business Newspaper × Anthropic AI Hackathon, we designed and built a service to help readers understand the context and relationships behind the news.

A single article can rarely capture every stakeholder’s perspective.

Some facts emerge over time, and articles also reflect the views of their authors.

By finding related articles and drawing on different frameworks for thinking,

we wanted to help readers explore how news events unfold alongside the perspectives of the people, companies, and other stakeholders involved.

Defining the problem and the service direction

We did not start and stop with the observation that news is consumed quickly, while its connections to past events and different stakeholder perspectives can be difficult to grasp.

When solving a problem, it is also important to consider whose problem we are solving. We therefore approached it from the perspective of Maeil Business Newspaper, which would put the solution into practice.

We began with Maeil’s ‘Make Knowledge’ vision and considered how to turn fragmented information into knowledge readers could make their own.

At first, we explored simply interpreting news through the thinking styles of well-known figures. We then developed the idea toward applying event histories and a range of frameworks, so that more readers could use it easily.

  • Extract people, companies, and events from articles and connect their relationships (the fundamental problem we wanted to solve).
  • Design a user flow that adds buttons and overlays to the existing news service (integrating it into Maeil’s service).
  • Make key information available without signing in and reuse each analysis for multiple readers (encouraging people to use the service).
  • Explore extensions such as chat and public discussion on top of the core experience (encouraging incoming users to stay).
Prototype

Validating the design with data

We analyzed average article length to calculate the Claude API processing cost per article, and examined responses by category to prioritize development.

We focused on extraction criteria and consistent output so that the results would make sense across different kinds of news.

An analysis of the distribution and characteristics of article data

Implementation and verification

We built the service using Claude Code.

While managing context and recording our work in pull requests and commits, we used tests and a separate test interface to check functionality and whether the data was reflected correctly.

As development accelerated, review became more important.

We encountered cases where existing features disappeared unintentionally or data structures became harder to maintain. Those experiences taught us that AI-generated work needs a verification process.

We also prepared a product plan and promoted the service through Instagram Reels.

A test interface created to verify the implementation

Results and lessons

We reached the finals and presented our work.

Although we did not win a final award, we experienced the entire product development process, from selecting an idea and defining requirements to data analysis, implementation, verification, and presentation.

We focused on whether the service could be integrated into a real product, whether operating costs and maintenance would be manageable, and whether AI outputs would be consistent.

We also spent a great deal of time considering how to communicate the core value to a non-specialist audience within a short presentation.

Participation journal

The Journal describes in detail how we narrowed down our ideas and completed the service.