, home Work with me

> diagnose · self-diagnostic · about five minutes

Which mode are you in?

Think of one task you use AI for most often. Answer for that task, not for AI in general.

Seven or eight questions. You get your default stance, the Stack layer it leans on, the one guardrail to adopt, and which AI @ Workplace level to watch next. Nothing is sent anywhere: your answers stay in this page (and, if your browser allows it, the last result is remembered on this device only).

session: diagnose

question 1 (7 or 8 in total)

keys: number to answer · backspace to go back

The interactive console needs JavaScript. Here are the same questions, so you can run the diagnostic on paper: the first questions lead to your stance; then look it up in the matrix and read across to its layer and guardrail.

the questions

  1. In this task, what do you mostly want from the AI?
    • An answer, or a fact
    • A thing made, or a job done
    • Help understanding something
    • Help thinking, practising or deciding
  2. If you chose “An answer, or a fact”: Where should the answer come from?
    • From the AI itself: I want the truth about a hard question I can't easily check myself → Oracle
    • From what is already known or published: find it and pull it together → Librarian
  3. If you chose “A thing made, or a job done”: What do you hand over?
    • My own rough material (notes, bullets, a draft), to be polished or formatted → Scribe
    • A request for one deliverable, which I check when it arrives → Genie
    • A goal: the AI works out the steps and does them, and I review at the end → Executor
  4. If you chose “Help understanding something”: Which is closer?
    • I have the material (a contract, a report, a dataset): tell me what it means and what follows from it → Interpreter
    • I want to understand it myself, and keep that understanding → Tutor
    • I need to carry it between two worlds (the engineer to the marketer, the doctor to the patient) → Translator
  5. If you chose “Help thinking, practising or deciding”: Which is closer?
    • Attack my plan or draft: find the holes → Adversary
    • Rehearse something real: a negotiation, a difficult conversation, a market → Simulator
    • Reflect my own thinking back to me, with more structure → Mirror
    • A non-judgemental space to think aloud about something personal → Therapist
    • Ideas: surprise me
  6. If you chose “Ideas: surprise me”: How do you want the ideas?
    • One good spark, which I judge as soon as it arrives → Muse
    • Many variations, kept, and sorted later with a cold eye → Miner
  7. How long have you used AI tools for work?
    • I'm new to them, or have barely tried
    • Weeks to months of chatbot use
    • Daily chatbot use, maybe custom bots or projects
    • I use agents that take actions (files, tools, other apps)
  8. In this task, how much of the work do you watch?
    • Assistive: I type, the AI suggests
    • Directed: I describe, the AI types, I review
    • Autonomous: I assign, the AI executes, I review the result
    • Continuous: the AI runs as a background service
  9. How big is the ask?
    • Speed: the same work in less time
    • Altitude: the outcome beyond the task
    • Depth: the task done the way a field-leading expert would
    • Range: bring in other fields, tools and techniques
  10. Last time the output disappointed you, what did you do first?
    • Rewrote or reworded the prompt
    • Added files, examples or background information
    • Explained what mattered most, or which tradeoff to make
    • Checked it against a source, a test or the original
    • Gave up and did it myself
  11. Before the output leaves under your name, how do you usually check it?
    • I don't, usually
    • I ask the same chat to check its work
    • I paste it into a fresh chat and ask it to find errors
    • I run it past a different model
    • I check it against the source, or my own calculation
    • I ask a person with the expertise I lack