Modern Technology Is Reaching Beyond Software — Into the Heart of Japanese Parliament
Modern technological innovation is making waves not only in the world of software engineering, but in the halls of the Japanese parliament. The Mirai Gikai project, which we introduce here, is a groundbreaking initiative that leverages cutting-edge web technologies like Next.js and Supabase, alongside LLM (large language model) agents, to make parliamentary information open and transparent. Developer Kenta Murai and his team — working alongside the parliamentary affairs team, secretaries, and other political experts — built the system while developing a genuine understanding of the nature of the Japanese parliament, the legislative process, and its complex specialized terminology.
In the early stages of development, the engineering team faced the reality of their own limited understanding of parliamentary procedures and the complexity of the bill-drafting process — at times feeling completely overwhelmed. By harnessing AI tools like ChatGPT to convert dense legalese into accessible language, they turned challenge into learning. This accumulation of challenges and growth exemplifies what it means to develop as an engineer in the AI era, and demonstrates new possibilities that emerge when cutting-edge technology and civic participation come together.
This article provides a comprehensive look at the challenges Mirai Gikai encountered and how they were overcome, the performance optimization techniques enabled by Next.js and Supabase, the practical application of AI agents, and the approach to cost management — all in detail. You will come away with a glimpse of a future where politics and technology converge, and a deep understanding of the technological innovations defining the AI agent era, told through the lived experiences of the development team.
The Initial Barrier: Engineers Who Didn't Know Parliament — How a "Sea of Jargon" Was Overcome with ChatGPT Next.js + React Server Components: The Ultra-Fast UI That Works Even on Slow Connections The Key to LLM Deployment Is Langfuse — Practical Know-How on Cost Management, Prompt Optimization, and AI Evaluation Technology Meets Politics — The Road to the AI Agent Era and an Open Information Society
Looking to optimize community management?
We have prepared materials on BASE best practices and success stories.
The Initial Barrier: Engineers Who Didn't Know Parliament — How a "Sea of Jargon" Was Overcome with ChatGPT
When development of the Mirai Gikai project began, the biggest obstacle was the sheer complexity of the parliamentary system itself. Developer Kenta Murai and the rest of the engineering team confronted the reality that they lacked not only deep expertise in software development in the civic domain, but even a basic understanding of political institutions. Their knowledge of parliament came largely from television news and general information sources. Fundamental details — the difference between the House of Representatives and the House of Councillors, how committees work, how bills become law — were far from clear. The team received lectures from parliamentary affairs staff including Mr. Yasu and Mr. Sugata, and secretary Ms. Furukawa, on the legislative process, types of bills, and the role of committees. Only through this collaboration did it become possible to approach the project from both a technical and political perspective.
Concretely, the team worked to translate the technical documentation produced by bureaucrats — filled with specialized terminology — into language that middle or even elementary school students could understand. For example, in the lead-up to an extraordinary session of parliament, team members read the original text of upcoming bills together, wrestling with questions like "What actually is a bill? Why does it require such complex procedures?" One member described the experience of being confronted with a document packed with specialized terms as feeling like "a wall suddenly appearing in front of me" — "my mind went completely blank," as one team member recalled, an experience that was no trivial matter.
Because a lack of knowledge about parliament could affect the design of the system and how data was handled, engineers focused first on re-learning the actual flow of parliament and legislation. As each member worked to read actual bill documents and understand the specialized terminology and legal phrasing, an AI tool known as ChatGPT proved enormously useful. For example, when the team asked it to convert the specialized terminology of a bill into "language a middle schooler could understand," the AI instantly rewrote the text in simple terms, dramatically improving overall comprehension. This experience became a precious opportunity for the development team to see firsthand the potential that comes from fusing technology and politics — watching difficult, inaccessible bill text transform into information accessible to everyone, with AI's help.
Through the efforts and pioneering spirit of the engineering team, and with the support of AI technology, this project — which converted opaque, closed parliamentary information into an open, accessible platform — has attracted significant attention as a new model for civic participation and public information disclosure.
Next.js + React Server Components: The Ultra-Fast UI That Works Even on Slow Connections
The Mirai Gikai development team adopted the latest web technologies to maximize user convenience and system performance. Adopting Next.js wasn't just about speeding up the front-end display. By using React Server Components, the team achieved server-side component rendering, delivering minimal HTML to users and building a system where information appears quickly even on slow connections or in areas with poor signal. This approach contributes significantly to a higher quality user experience.
Using Supabase streamlined database and authentication functions, making team collaboration easy in a way similar to services like Firebase. Unifying the technology stack reduced the cost of coordination across projects within the team, allowing all projects to move forward rapidly. For example, by using the same technology stack across other products like the Political Funds Transparency and Action Board projects, technical collaboration could happen horizontally across the team, and pull requests could be merged immediately. This setup — where engineers can freely collaborate without being constrained by environment differences — dramatically increased the development pace of the Mirai Gikai project.
By utilizing React Server Components, the system efficiently generates HTML on the server side, enabling rapid display of only the data needed.
Another strength of these modern technologies is cache management and database access optimization. Even on pages where content changes dynamically, caching database query results for a set period prevents unnecessary reloading and reduces the overall system load. The combination of Next.js's caching capabilities and Supabase's system-wide integration delivered significantly better performance than previous products — a point that has been widely praised.
The benefits of these technologies are immediately apparent to users during actual system operation. When searching for bill content or parliamentary information, for example, the required data appears instantly, allowing users to find what they need without frustration. Attention has been paid to display speed even on mobile devices and under slow connection conditions, earning positive reviews for smooth operation even during peak hours.
The technical ingenuity demonstrated in this development environment is likely to influence many engineers as a model for future web application development. The adoption of new technologies and their appropriate application have proven to be a decisive key to enabling flexible, efficient information delivery even on a complex topic like political information. The Mirai Gikai project, through the use of these technologies, represents a groundbreaking initiative — opening the door to a bright, open model of civic engagement where closed parliamentary information is transformed into something anyone can access and understand.
The Key to LLM Deployment Is Langfuse — Practical Know-How on Cost Management, Prompt Optimization, and AI Evaluation
In the Mirai Gikai development process, LLM (large language model) agents play a major role in maximizing AI potential. In particular, the automation and efficiency gains enabled by LLM agents are directly linked to faster development and higher overall quality. However, this brings two major challenges: rising usage costs and the management of prompt operations. This is where Langfuse — an LLM management tool — enters the picture.
Langfuse was adopted in the Mirai Gikai project as a tool for centrally managing AI service usage costs, prompt management, and LLM performance evaluation. On the cost side, the service operator always faces the problem of expenses growing in proportion to conversation volume. Without proper management of overall costs and individual user usage, the entire system's operation could be put at risk. Langfuse's management interface enables real-time monitoring of overall usage and per-user usage, allowing the team to set daily usage caps and prevent malicious users from driving up costs across the system.
Prompt management was another major challenge. Previously, engineers embedded prompts in source code and had to make changes themselves whenever adjustments were needed. This made it difficult for business-side users and non-engineer team members to participate in prompt refinement, slowing down rapid improvement. By using Langfuse, the parliamentary affairs team and other stakeholders — those with deep political knowledge — can directly version-control and modify prompts themselves. Creating an environment where both developers and business stakeholders can be involved directly improves the user experience and enables quantitative evaluation of user feedback, similar to A/B testing. The system actually collects user responses (positive and negative ratings) by prompt version, enabling objective judgment of which prompts are most effective.
Key features of Langfuse that have been particularly well received include:
- Real-time measurement of overall and per-user usage costs
- Prompt version management and A/B testing that enables user-feedback-driven improvement
- An LLM self-evaluation function (LLM-as-a-judge) that reduces the burden on developers while improving the accuracy of performance evaluation
With these features, the Mirai Gikai project is able to efficiently operate LLMs while consistently extracting optimal performance. In the development environment, a culture has taken root of comparing multiple LLMs — such as Claude Code and models provided by OpenAI — and selecting the best tool for the situation. Developer Murai himself has said there is no absolute standard for which LLM agent to use, and that he switches depending on daily tasks and context — a flexibility that contributes to the overall success of the project.
Using Langfuse, the team also built a user interface where non-engineers can easily query and adjust AI. For example, pressing and holding a piece of text on a smartphone immediately brings up an AI explanation — a feature that many users find extremely attractive. This means that even when users encounter specialized terminology or confusing expressions, they can immediately get a clear answer, allowing people unfamiliar with politics to navigate the information with confidence. Careful UI customization also added a feature that displays furigana (phonetic readings) for kanji characters, extending accessibility to users who struggle with complex Japanese.
The use of AI agents has contributed greatly to overall system efficiency, improved user experience, and higher developer productivity. Through the repeated process of fine-tuning prompts and evaluating LLM performance, the AI itself continues to improve, enabling ever-higher-quality service delivery. In the domain of political information, the introduction of these cutting-edge technologies has proven to be a decisive factor in breaking down traditional barriers and realizing an information platform that all users can access and understand with ease. The result of everyone involved in the project recognizing their own role and working together to effectively leverage LLMs and Langfuse is that the Mirai Gikai project has attracted broad attention as a new model for political information delivery.
Technology Meets Politics — The Road to the AI Agent Era and an Open Information Society
The development of the Mirai Gikai project was an attempt to bring technology to bear on a field — politics — that has traditionally been complex and difficult to access, creating an open and accessible information platform. In the early stages, the team struggled to understand the nature of parliamentary institutions and the bill-drafting process, but through collaboration with the parliamentary affairs team and other specialists, and with the support of AI tools like ChatGPT, they succeeded in converting all information into a form understandable even to middle school students. The engineers, confronted with specialized terminology and complex legislative processes, persevered and continued learning and practicing until they succeeded in building a system that delivers political information in a form accessible to everyone.
The adoption of Next.js, Supabase, and other modern technologies delivered a fast, easy-to-use system. The use of LLM agents and the introduction of management tools like Langfuse realized cost management and prompt version management and optimization. This prevented the risk of malicious use driving up system-wide costs, while creating an environment where business-side users can easily adjust prompts. The AI's self-evaluation function (LLM-as-a-judge) also ensures consistently high-quality responses, contributing to the accuracy and reliability of information — a point that should not be overlooked.
Overall, the Mirai Gikai project is a pioneering effort at fusing politics and technology, and a strong example of engineers and domain experts sharing knowledge and building a robust foundation for delivering complex parliamentary information to users. The modern technologies and effective LLM operations demonstrated in this project will undoubtedly serve as a reference for system development in other fields as well. From this initiative, we can glimpse a future where political information — rather than hardening into impenetrable specialized knowledge — becomes something open and accessible to everyone.
Reference: https://www.youtube.com/watch?v=vOTVDplC3Z0
Streamline Event Management with AI | TIMEWELL Base
Struggling with the complexity of managing large-scale events?
TIMEWELL Base is an AI-powered event management platform.
Track Record
- Adventure World: Managed Dream Day with 4,272 attendees
- TechGALA 2026: Centralized management of 110 side events
Key Features
| Feature | Impact |
|---|---|
| AI Page Generation | Event page ready in 30 seconds |
| Low-Cost Payments | 4.8% fee (among the lowest in the industry) |
| Community Features | 65% of attendees stay engaged after the event |
Feel free to reach out for a consultation on streamlining your event operations.