Since launching the beta in December 2025, we have provided JijZept AI to several companies, gathering extensive feedback from real-world optimization workflows. The quality of our mathematical modeling support has been particularly well-received, with confirmed cases where AI-generated models produced results comparable to or better than those created manually.

Based on this feedback, our February development focuses on increasing opportunities for interaction between users and the AI, fostering context formation and deeper understanding of problems and processes.

More Opportunities for User-AI Interaction

JijZept AI retains its ability to automatically execute workflows from modeling to visualization. As a development direction going forward, we are strengthening the design so that users can provide feedback and approval at each step.

• A partner for formulation and modeling: Work through optimization problem formulations step by step through dialogue with the AI.
• User participation in decision-making: Review and approval points are introduced before the AI executes its next action.
• Context formation: Deepen your understanding of problem backgrounds and constraints through conversation, leading to more appropriate models.

We plan to realize this direction through the following improvements.

• Tool Selection & Approval: A new UI that lets users review and approve the tools the AI uses
• Expanded File Attachments: Support for Word and Markdown files, drag & drop, and clipboard image paste
• Faster Responses: Improved response experience through agent architecture revisions and other measures
• UI Improvements: Plain text copy for AI output, Python code display fixes in dark mode, and more
• Visualization Feedback: Annotate and provide feedback directly on visualization results

Through these improvements, we aim to make JijZept AI an even more effective partner for mathematical optimization. We welcome your continued feedback.