Take the instructions
One line. It points your assistant at the questions and tells it how to behave.
I want an honest read on how I actually work with AI. Fetch https://ai.rimaslogic.pl/check.md and follow it. Ask me the questions one at a time.
Some assistants cannot open links. When that happens they rarely say so; they invent something plausible instead. If the questions start sounding made up, use the second button. It copies the whole file, so nothing needs fetching.
Put it in your own tool
Anything that holds a conversation will run this. We build the training around Claude, but nothing here depends on it.
Claude
New chat, paste, send.
defaultChatGPT
New chat, paste, send. If browsing is off for your account, use the full text.
Gemini
New chat, paste, send. Workspace policy often blocks fetching, so the full text is the safer route.
Claude Code
Paste at the prompt. The interview runs in the terminal.
Cursor
Chat panel, paste, send. Full text if the fetch is refused.
Anything else
Use the full text and it will work.
Talk for ten minutes
It asks one question at a time and pushes for specifics: what you pasted in, what came back, what you did with it next. It will not explain what AI can do, recommend a tool, or tell you that was a great example. It is collecting, not teaching.
Nothing is uploaded, saved or sent anywhere. The instructions say so in as many words, so an assistant with connectors attached leaves them alone.
One block, however suits you
It finishes by printing a summary, then one paragraph on what to look at first. Paste the block into the form you were given, or mail it.
Shortened. The real block carries a few more lines.
Where your answers go
To Rimas, and nowhere else. Not to your manager and not to your team. Where this runs inside a group programme, everyone sees the overall picture and nobody sees who said what. Question three asks what did not work. That only gets an honest answer if it stays private.