Working with AI

AI tools are welcome at Hardal.

But...

Claude, Codex, and Cursor are assistants that help you move faster. They do not replace you. AI is a tool. You should not treat it as if it were you.

You should read No Meat Proxy first. It makes the case against pasting unread AI text as if you wrote it.

Don't

• Don't paste real customer data into AI chats.
• Don't paste passwords, keys, or tokens into AI chats, Slack, or screenshots.
• Don't ship or send AI text you have not read.
• Don't dump unread AI walls or old threads into Slack, Linear, or email.
• Don't let a bot speak for you. You answer in your own words.
• Don't stay quiet if private data went into a bot. You should tell Berkay in a Slack DM.
• Don't answer a critique of your work with a reply a model wrote for you.
• Don't hand the blame to the model when someone gives you feedback.
• Don't hide a gap in your understanding by saying a model produced the words.
• Don't lower your standard for the work because an AI tool helped you make it.

Here is what you should try instead

• You should read the work, understand it, and verify it before you send it.
• You should reply to feedback as the owner of the work.
• You should take responsibility for the work. Your name is on it, so the work is yours.
• You should show other people that you care about what you are doing.

You can read more about how we keep data and secrets safe on the Security page.