Skip to content
kc@kumarChandrachooda.com:~$ cd /policy/ai && read --section="top" 0%
Policy

How this site uses AI

What AI does here, what it never does, whose model runs, and what leaves the site.

Version 1 Effective 08 Nov 2026 Under review
In five lines
  • The site runs no AI model of its own and pays for none.
  • It never receives or stores your AI keys.
  • Search, proofreading and checks are ordinary code.
  • You can bring your own assistant to the site, and your assistant's company runs it.
  • Every post is reviewed by me, and a post the writer helped with says so.
What runs with no AI at all

Proofreading in the Workshop, its spelling and house-style checks, its checks on figure descriptions and its tag suggestions are rules and ordinary code. They run on the site, and nothing you type into them leaves it.

Search and related posts

Search ranks posts by the words in their titles, tags, summaries, series and categories. That part is ordinary code.

Related posts are worked out ahead of time, when I build the site, by a small open model that reads only the posts. It never sees anything you type.

Your assistant and the site's connector

You can add this site to Claude, ChatGPT or another assistant as a connector. Your assistant then searches the blog and reads post summaries through the site. Your assistant's company runs the model, under your agreement with them. The site sees only the requests your assistant sends it. The details are in Connected AI and the MCP server.

The site also publishes /llms.txt, a plain map of its public pages for assistants.

When you connect your own AI

If you are signed in, you can connect an OpenRouter account or your own key from Anthropic, OpenAI or Google on /ai/connect. The key stays in your browser, encrypted in that tab, and lasts only as long as the tab. Calls go straight from your browser to the provider you chose, and the site never receives the key. The page where you connect it may talk only to the site and those four companies.

The details for authors are in AI for authors.

Writing with AI help

In the Workshop I can ask a writer for help with a draft, through my own AI connection. Invited authors can too, through theirs. The draft goes from the browser to the provider chosen on /ai/connect, never to a model of the site's. Every suggestion arrives marked, and nothing changes until it is kept. When one is kept, the post records which of the writer's tools helped.

The details for authors are in AI for authors.

Posts, figures and descriptions

I review every post before it goes live, and I am responsible for what it says. A post a writer helped with says so under its title. The description read out for each post's hero image comes from the image's own caption and labels, by rules, not by a model.

Some earlier posts, and some of their hero images, had AI help. They carry no label of their own, so this line is theirs.

Every feature, in one table
Feature Uses AI? Whose model What leaves the site Kept by the site
Search no — nothing nothing
Related posts when I build the site a small open model nothing the list of related posts
Proofreading, house style, tags, figure checks no — nothing nothing
Ask this post no — nothing nothing, but your question is in the page's address, so your browser's history, a link you share and access logs along the way may keep it
The site in your assistant yours your assistant's company the site's public content, to your assistant a salted hash of your address for one minute, and the tools called for 30 days
Your own key, on /ai/connect yours the provider you chose what you send, from your browser nothing, the key stays in your browser
The Workshop's writer the writer's own the provider chosen on /ai/connect the draft, from the browser which tools helped, once a suggestion is kept
Hero image descriptions no — nothing the description, once I keep it
What your browser is allowed to run

Every page on the site carries a security policy that lets your browser run only the site's own scripts. Only the page where you connect your own AI may talk to an AI company.

Changes

This page changes when the site's AI features do. Each version is dated, and earlier versions stay readable at their own address. Questions to kumarchandrachooda@gmail.com.

Which CV would you like?

Pick the one closest to the role. Each is two pages and carries the same facts.

Applied AI

Agentic AI & Cloud Engineer · multi-agent systems, RAG and MCP on AWS

Cloud & .NET

C#, .NET 10 and ASP.NET Core microservices on AWS and Azure · Terraform & AWS CDK

Data Engineering

AWS lakehouse · Kinesis, Glue (PySpark), Apache Iceberg, Lake Formation

Everything

All three domains, with the standout project from each