About

How this site is structured, built, and operated.

https://the-shashi.com/about/ / Last updated: 2026-08

What is this site?

A personal site that researches and writes about the history of Japanese companies and the decisions their managements made. Its purpose is to give working people a long-term perspective. I (Yutaka Sugiura) build and run all of it alone, on weekends.

It has been a hobby for ten years, and it is built one small piece at a time.

Who reads this site?

To optimise for access by AI rather than by people, I built my own log analysis pipeline for traffic at scale.

Breakdown of page requests (a sample period in August 2026)

Category	Share	Detail
Human: desktop	10.1%	Spoofed user agents removed as far as possible
Human: mobile	3.4%
AI: from data centers	24.0%	ChatGPT 60%, Claude 4%
Bot: from data centers	51.5%	Googlebot, bingbot and the like
Blocked: excluded regions and others	11.0%

Every request that reaches the edge is parsed out of the AWS CloudFront logs and sorted into four groups. Human is a real reader on a desktop or a phone; user-agent spoofing can never be ruled out, so addresses that look automated are subtracted in the daily analysis. AI is traffic through assistants such as ChatGPT, Claude, Meta AI and Perplexity, together with their crawlers. Bot is search and SEO crawlers such as Googlebot and bingbot, plus anything judged automated from its user agent, ASN or IP address (data-center ASNs such as M247 and Oxylabs among them). Blocked is traffic from regions outside the delivery scope, refused at the edge without a response. AI crawlers take the HTML and never execute JavaScript, so tag-based analytics never counts them at all. In an age where the script no longer fires, raw logs are the only thing left that shows what is really arriving — and CloudFront logs cost almost nothing to keep, so they are parsed in house.

How is it delivered?

To keep compute to a minimum the site is served entirely as static files, at a target cost of one US dollar a day. The financial data for the past ten years is generated by machine, from XBRL pulled directly out of EDINET, the disclosure system of the Financial Services Agency of Japan.

← Back to Home