Build a small AI-style chat with token streaming and cancellation to learn how WebSockets and ASGI work together.

A simple comparison of WSGI and ASGI, using an AI agent that spends most of its time waiting for LLM APIs.

In 2025, OpenAI introduced a new way to call its models: the Responses API. Their pitch is that switching from the legacy Chat Completions…

2年前の記事で整理したデュアルキャリアの意思決定を、実際に運用した結果を振り返ります。出産、イタリア移住、交代制ワンオペ育児を経て見えた現実と学びをまとめました。

I migrated this personal blog from Gatsby to Astro, but the main goal was to fix long-standing UI problems and simplify the site's structure.

2025年に読んだ本を振り返ります。

This article picks five notable AI papers from 2025 and summarizes their key ideas and limitations, including reinforcement learning for reasoning, agent benchmarks, long‑task metrics, and a statistic...

AIによって知識がコモディティ化していく時代において、人間の価値がどこで生まれるのかを、フロンティア知識と行動のプレミアムという観点から整理しました。

Don’t chase leaderboards. Run the benchmark yourself and map your score–latency–cost frontier to choose models for production.

These metrics capture coverage and reliability.

AI agents need stateful benchmarks. Unlike LLMs, agents interact with databases and users. We explore why and how to evaluate them…

Some LLMs disable sampling knobs like temperature and top_p. Here’s why.

A deep dive into how LLMs serialize prompts, output schemas, and tool descriptions into a token sequence, with examples from Llama 4's implementation.

The story behind our AI agent that dynamically investigates insurance claims, automating 30% of support tickets it receives.

『コンピュータの構成と設計』（通称「パタヘネ本」）のまとめと感想です。

A deep dive into how databases work.

Learn how to extend Claude's capabilities by building your own Model Context Protocol server.

DeepSeek shocked the world by dropping an open-weight successor to OpenAI’s o1: R1. This post summarizes our learnings from the tech…

DeepSeek shocked the world by dropping an open-weight successor to OpenAI’s o1: R1. This post summarizes our learnings from the tech…

A detailed guide on how to build applications with foundation models.

At Alan, exceptional customer care isn’t just a service — it’s a core part of what sets us apart in the insurance industry.

At Alan, exceptional customer care isn’t just a service — it’s a core part of what sets us apart in the insurance industry.

今後5年程度の未来を見据え、AI時代を生き抜くための方法について自分の考えをまとめました。

Reflecting on 2024's deep learning breakthroughs! Discover my top 10 favorite research papers that shaped the field this year.

RLHF is not the only method for AI alignment. This article introduces modern algorithms like DPO and KTO that offer simpler and more stable alternatives.