Working with AI-Collaborative Format

Account

This page explains how to proceed with the AI-collaborative test format.

Overall Test Flow

Step 1: Launch Development Environment
  ↓
Step 2: Implementation of Tasks
  ↓
Step 3: End of Test (Automatic Submission)

Step 1: Launch Development Environment

When you start the test, a browser-based development environment (VS Code) will launch automatically. It may take a few minutes to start up, but during this time, you can still review the problem statement or use the "Messages" function to communicate with AI for interactive-type problems.

Exam Rules and Guides

  • Exam Rules: At the start of the exam, a modal will display the exam rules, including important precautions. You can re-check these at any time via the "Exam Rules" tab during the test.
  • Exam Guide: Supplementary information previously located in the "Check Issue" step has been moved to the "Exam Guide" tab.

Available Tools

  • Terminal: You can execute commands from the Terminal located at the bottom center of the screen.
  • GitHub Copilot Chat: You can ask questions or consult with the AI via the chat on the right side of the screen.
  • Claude Code / Codex: These are pre-installed in the development environment. You can use them within the "GitHub Copilot Chat" or by launching them in the "Terminal."
  • Messages (Interactive problems only): You can communicate with AI stakeholders through the "Messages" section at the bottom left of the screen to gather requirements and other information.

Available AI Models and AI Usage Limits

The scope of available AI models and the AI usage limit are configured by the company for each problem. You can check these settings at any time in the "Exam Rules" tab.

  • Available AI Models: Either "Standard" or "High-Performance" tiers are set. You can only use models included in the assigned tier.
  • AI Usage Limit: A monetary limit is set for AI usage. Once the limit is reached, you will no longer be able to use the AI, so please be strategic in your interactions.
  • These settings apply to all AI tools used within the exam environment, including GitHub Copilot Chat and Claude Code / Codex.

The models available in each tier are as follows (as of July 9, 2026. This is subject to future updates):

Tier Available Models
Standard claude-sonnet-5, gpt-5.4, claude-haiku-4-5, gpt-5.4-mini
High-Performance claude-opus-4-8, gpt-5.5, claude-sonnet-5, gpt-5.4, claude-haiku-4-5, gpt-5.4-mini

Step 2: Implementation of Tasks

Once the development environment has launched, proceed with the implementation according to the instructions in the problem statement.

📝 About Evaluation

  • The code as it exists at the end of the exam will be evaluated as your work product.
  • Your interactions with the AI will also be subject to evaluation.

Step 3: End of Test

There is no individual "Submit" button for this exam. Your final work product in the IDE will be automatically submitted and evaluated at either of the following times:

  • When the time limit expires.
  • When you click the [Finish Test] button at the top right of the screen.

Please note that clicking the [Finish Test] button will completely end the entire exam, even if there are other unanswered questions remaining. Be careful not to press this button by mistake until you have completed all questions.