Privacy policy

Who runs this site

This site is operated by Michael Lynn as an independent personal educational project. It is not affiliated with, sponsored by, or endorsed by Anthropic, Cursor, or any other company whose products it discusses. “Northwind” is a fictional company invented for the course; it is not a real business and it does not sell anything.

The course lives at triage.mlynn.dev. The companion demo storefront lives at northwind.mlynn.dev. Both are covered by this policy.

What the course site collects

Reading the course requires no account and no sign-in. Specifically:

• Page views. Aggregate page-view analytics through Vercel Analytics, so “is anyone using this course” has an answer. It sets no cookie and no cross-site identifier, and it does not build a profile of you.
• Your course progress. Lab progress, mission progress and the presenter timer are stored in your own browser’s local storage. They never leave your device and are never sent to a server. Clearing site data erases them.
• Search. Site search runs entirely in your browser against an index shipped with the page. Your queries are not sent anywhere.
• Server logs. The host (Vercel) records standard request logs, including IP address and user agent, for security and abuse prevention.

What the assistant collects

“Ask Northwind” is the chat panel in the corner of the course, and the support flows on the demo storefront work the same way. When you send it a message:

• A session cookie (northwind_assistant) identifies your conversation so replies land in the right transcript. It is a random identifier and holds nothing about you. See the cookie policy.
• Your message and the assistant’s reply are sent to the Anthropic Claude API to produce an answer, and are stored so the conversation has memory of its own earlier turns.
• If you ask the assistant to escalate something to a human, the message and any contact detail you typed into it are stored as an escalation record.

Please do not type anything sensitive into the assistant. It is a teaching demo, not a support desk for a real company. Do not enter passwords, API keys, payment details, government identifiers, or health information into the chat. (Your own Anthropic key has its own form, described below.)

Signing in with GitHub, and free credit

The AI features — the Tutor, Ask Northwind, and the demos on the storefront — run on the Claude API, which costs money per request. Where the site asks you to “Sign in with GitHub”, it is to give each learner a small amount of free credit and to stop that credit being claimed over and over.

• The site requests no GitHub scopes. It reads your public profile once and keeps only your numeric GitHub id, your username, and when your GitHub account was created (to decide whether it qualifies for free credit). It does not receive or store your email address, and it discards the GitHub access token immediately after that one read.
• Against that id it stores how much free credit you were given and how much you have used, in dollars.
• A record of each AI request you make: which feature it was (for example a Tutor hint), which Claude model answered, how many tokens it used, what it cost, how long it took, and an error code if it failed. It does not store what you asked or what the AI replied. Requests from visitors who are not signed in are recorded the same way, with no identity attached.
• Tutor review results. When the Tutor reviews an attempt, the site keeps which lab it was for, whether it passed, how many rubric criteria were met, and whether each deliberately planted mistake was fixed. Your attempt, the Tutor’s feedback and the rubric text are not stored.
• The site’s operator can see these records, by username, to understand how the course is used, what it costs, and which labs are not landing. Nobody else can.
• A session cookie (nw_session) keeps you signed in. The database holds only a one-way hash of it. See the cookie policy.

You can revoke the site’s access from your GitHub application settings.

Feedback

Course pages, the Tutor and the assistant have “Was this helpful?” buttons, and there is a feedback page. If you use them:

• The site stores your rating, any reasons you pick, your comment if you write one, the page’s path (never its query string), and which lab it concerns. It does not store the text of the page, lesson or reply you rated.
• Before a comment is stored, email addresses, phone numbers, card numbers and API keys in it are removed. Please still leave personal details out.
• If you are signed in with GitHub, your feedback is linked to your username so the operator can read it alongside how you used the course. Otherwise it is anonymous.
• Only the site’s operator can read it. There is no reply address.

Your own Anthropic API key

When your free credit runs out you can add your own Anthropic API key. If you do:

• It is sent once, over HTTPS, to the storefront, checked with Anthropic by a request that costs nothing, and stored encrypted (AES-256-GCM) and bound to your current sign-in session, so it cannot be used from any other session.
• It is used only to make the AI requests you make on this site, and it is never shown back to you or to anyone else — the page only ever sees its last four characters.
• It is deleted automatically after 24 hours without use, immediately when you press “Remove key” or sign out, and immediately if Anthropic rejects it.
• Requests made with your key are billed to your Anthropic account by Anthropic. The site records how much it spent this session so the meter can show you.

A key created for this purpose, with a spend limit set in the Anthropic Console, is a good idea, and you can revoke it there at any time.

Signing in with Google

Where the site offers “Sign in with Google”, Google returns your name, email address, and profile picture. That is the whole of it: the site requests no other scope, and it never asks Google for access to your Gmail, Drive, Calendar, Contacts, or any other Google service.

That information is used only to identify your account and show you your own work. It is not sold, rented, or shared with third parties, it is not used for advertising, and it is not transferred to anyone except the infrastructure providers listed below who need it to run the site. Google’s own handling of your data is governed by the Google Privacy Policy. You can revoke this site’s access at any time from your Google account permissions page.

How long anything is kept

Retention is enforced by the database itself, not by anyone remembering to run a cleanup job. Records expire on this schedule:

Who else sees it

No data is sold or shared for advertising. Three providers process data because the site cannot run without them:

• Vercel — hosting, request logs, and page-view analytics.
• MongoDB Atlas — the database that holds assistant conversations, escalations, sign-in records, AI request records, Tutor review results, feedback and encrypted API keys.
• Anthropic — the Claude API, which receives assistant messages in order to answer them, and verifies an API key you add.
• GitHub — only if you choose to sign in with it.

Data may also be disclosed where the law requires it, or to investigate abuse of the service.

Your choices

• Don’t use the assistant, and nothing about you reaches the database at all.
• Clear your browser’s site data to erase local progress and the session cookie.
• Ask for deletion of anything held about you, or for a copy of it, by emailing the address below. Most records expire on their own within 30 days regardless.
• Revoke Google sign-in access from your Google account permissions page.
• Remove your Anthropic key or sign out from the credit meter in the corner of any page; both delete the stored key at once.

Children

This site is aimed at professional developers and is not directed at children under 13. It does not knowingly collect information from them.

Changes

If this policy changes materially, the date at the top of the page changes with it. The page history is public in the course repository.