Qualiteg - IT & AI Technology

Building a Coding Agent from Scratch, Part 2: The Completion Gate and What to Write in a Pushback

AI Agents

Building a Coding Agent from Scratch, Part 2: "Done" Means "It Runs," Not "It's Written." The Completion Gate and What to Write in a Pushback

The model's "Done" slipped through in five ways, from a bare syntax check to tests with no assert. We describe, with numbers, a completion gate that closes in three stages, and why a pushback should contain the failing lines and how to investigate, but never the answer.

AI Agents

Building a Coding Agent from Scratch, Part 1: Why "Finishing the Job" Is Hard, and Separating the Turn Runner from the Stop Decider

When we built our own coding agent, the hardest part was not running the loop but deciding when to stop. In one run, 234 of 300 turns never called a tool. Starting from that number, we separate the turn runner from the stop decider and fix how pushbacks are sent.

GPU

TensorRT 10 × Blackwell Migration Guide (Part 1): Inference Assets Won't Run on RTX 50 — The Basics and the First Walls

Hello! You get a new GPU, excitedly move your existing inference stack over, and everything that ran fine yesterday now stops with errors across the board — sound familiar? In migrations to the NVIDIA RTX 50 series and NVIDIA RTX PRO series (the Blackwell generation), this happens with rather high probability.

Generative AI Frontlines

Suspended Three Days After Release: What the U.S. Government Directive on Fable 5 / Mythos 5 Reveals About a New AI Availability Risk

Hello! In our previous article, we covered Claude Fable 5 and Claude Mythos 5, announced by Anthropic on June 9, 2026. Fable 5 took Mythos-class capability, added safeguards, and brought it down into a form that could be offered to general users. We described it as "the myth descending

AI Mathematics

How to Talk Honestly About PII Detection Accuracy: Evaluation Design Through Recall, Confidence Intervals, and Representativeness

Hello. This is the Qualiteg Research Team. We work on technology for detecting and masking sensitive information — personal information (PII), confidential information, and special-care-required personal data (https://pii-fi.com). In that work, the question we face every day is: 「how can we talk about accuracy figures honestly and correctly」 ? Suppose,

Daily Dev Tips

PyCharm Silently Crashes on npm start: Isolating the Cause

Hello! This is the Qualiteg Product Development Team. The moment you run npm start in PyCharm's built-in npm tool terminal, the IDE vanishes without a single error message. Restart it, try again, and it crashes again. The logs offer no clues. In this post we document our encounter

LLM Security

Understanding Active Directory: The Technology Behind Enterprise AI Security - Part 6: Common Problems and Solutions

Hello! In this sixth installment of the series, we cover Troubleshooting - common problems and how to solve them. In the previous installment (Part 5), we looked at how Integrated Windows Authentication behaves in the browser. By understanding the concept of the intranet zone, it became clear why authentication behaves

LLM Security

What the Confusion Matrix Doesn't Show in PII Detection: Cross-Recognizer Collisions and Integration Testing

Hello! This is the Qualiteg Research Team. Automatic detection of personally identifiable information (PII) can be framed as the problem of extracting specific expressions from text and determining which type of PII each one represents. As the range of PII types to detect grows — phone numbers, personal names, account numbers,

IT & AI Technology

Understanding Active Directory: The Technology Behind Enterprise AI Security - Part 5: Browser Configuration and Authentication

Hello! In this fifth installment of the series, we cover "Browser Configuration and Authentication." In the previous installment (Part 4), we explained how joining the proxy server to the domain makes it possible to reliably identify who accessed ChatGPT or Claude. You should now have a good grasp

IT & AI Technology

Optimistic vs. Pessimistic Locking: Lessons in Concurrency Control from a Real-World Incident

Hello! This is the Qualiteg product development team! "We implemented optimistic locking, but we're still getting conflict errors..." This actually happened to us. In this article, we explain the difference between optimistic and pessimistic locking through a real incident we ran into. Rather than abstract explanations,

LLM Security

Why Did Enterprise Security Become So Complex? From the AD + Proxy Era to the Modern Cloud

Hello! Now that generative AI services like ChatGPT and Claude are making their way into everyday work, "the risk of sending confidential information to AI" has emerged as a new security challenge. As we grappled with this challenge, we found ourselves looking back at how enterprise security architecture

Generative AI Frontlines

Getting Started with Claude Code

Hello! Today we take up Claude Code, the AI coding agent that has been generating a lot of buzz among engineers lately. AI coding assistants have multiplied rapidly over the past year or two, and many of you probably feel you can no longer tell which one to choose. In

LLM

Japanese LLM Ranking 2025: Benchmark Analysis Report (December 18 Edition)

⚠️ A newer version of this article is available The latest LLM ranking (March 2026 edition) is here Read the latest LLM ranking (March 2026 edition) → Introduction This report is a comprehensive analysis of the performance of Japanese-capable LLMs, based on benchmark data from Nejumi Leaderboard 4 (the 2025/12/18

Daily Dev Tips

The Overlooked CPU Bottleneck in GPU-Based Distributed Processing — and How taskset Solves It

Hello! When designing parallel processing systems that use multiple GPUs, it is easy to design the whole system without giving much thought to the CPU. "The GPU is the star of a machine learning system, so surely the CPU doesn't need much attention?" Actually, that is

IT & AI Technology

Understanding Active Directory: The Technology Behind Enterprise AI Security - Part 4: Proxy Servers and Integrated Windows Authentication

November has arrived, and the morning and evening chill has set in for real. Many of you have probably started turning on the heat at the office. With a warm cup of coffee in hand, we bring you Part 4 of the series, "Proxy Servers and Integrated Windows Authentication&

LLM

Japanese-Capable LLM Rankings 2025: A Benchmark Analysis Report

Introduction This report analyzes the performance of Japanese-capable LLMs based on Nejumi Leaderboard 4 benchmark data (October 11, 2025 edition), offering a comprehensive picture of the field. Nejumi Leaderboard 4 is known as a highly reliable benchmark that evaluates LLM performance on Japanese-language tasks from multiple angles. In this analysis,