Field study

How do people who use AI every day set it up?

Your AI writes up how the two of you really work together, from your rules and your past sessions. You get that write-up back as a private report. Once enough people have taken part, it also shows how you compare with other participants and what they do that you don't.

Built for people who work in Claude Code, Codex, Cursor or a similar tool.

We use your email to send your code and your report. No other participant sees your answers. Norsoft reads them to build your report and the study's findings. Been here before, or lost your code? Enter the same email and we'll send a new code for the same profile. It replaces the old code, and your earlier reports stay.

What you get

A report on your own setup, next to everyone else's

  • Your whole setup, written out in full
  • How you compare in each area of working with AI
  • What other participants do that you don't
  • Where your written rules and your real habits differ
See a sample report
How it works

Four steps, and your AI does most of them

01

Enter your email

We send a private code to that address, which is how we know the report goes to the right person.

02

Paste one sentence

The email holds one sentence with your code inside it, and you paste it into the AI that knows your setup best.

03

Your AI answers

Our server hands out the questions a few at a time, checks every answer and sends back anything that is incomplete. This is the long part, because every answer is written out. At the end your AI shows you a summary and waits for your yes, and nothing is kept as final before that.

04

Your report arrives

You get a link to a private page. Run it again later with the same code to see what changed.

The questions

Questions about what you really do

Every question asks one thing and gets a written answer: what actually happens, a real example, and when it doesn't apply, with room for technical detail. Deeper sets of questions open for the setups they fit, such as custom-built parts, code, or writing that goes out in your name. Your AI also picks tags from a fixed list, so answers can be counted and compared.

Read all the questions
Context, one of 50 core questions
By what means does background information reach the LLM when you work with it?
Because a rules file was picked
About how long is the rules file, counting everything it loads?
With every answer
Your AI also writes it out
  • What actually happens
  • One real example
  • When it doesn't apply
  • How well it works
  • The evidence it found

Pick one or more tags to see the follow-ups.

From the first private round
“Going through it strengthened how I manage context and how my setup is built. Comparing with the others also showed me that how people manage context depends on what they use AI for.”
Tim Wenger, Norsoft. Tim took part in the first private round with a small group, before the public report existed.
Why we run this

You should know what's in it for us

We run this study to learn how experienced people set up their AI, and we publish what we find so everyone can use it.

What does Norsoft get out of this?
A

We get data

Your answers become part of what we study and publish, as counts and never as your words. If you ever want your submission deleted, one link removes it.

B

We want to give something back

Advanced users rarely get to see how others work. This lets you learn from how other people set up their AI, safely, without anyone seeing your answers.

C

It helps our business

Norsoft is a custom software company, and we hope running this study makes us known for this kind of work.

Your data

What we keep and what we don't

We keep

  • Your AI's written answers to the questions.
  • Your email, stored apart from your answers.
  • Three broad facts about you: your kind of work, your industry and the size of your organization. Your AI can answer "would rather not say" to any of them.

We don't

  • Take your files. Your AI reads them and sends only the answers.
  • Show your answers to any other participant. Reports use counts, never another person's words.
  • Publish a ranking or show your numbers to anyone. The report counts practices for you alone, and it explains how each count is made.

Answers are reported by your AI and we can't verify them, so every comparison says how many people it's based on. We use AI tools to help us read and compare answers. When we do, the company that provides the tool processes those answers for us. Today that is Anthropic, the maker of Claude. Every email we send has a link that deletes everything we hold for you. How the study works

Taking part

One sentence, in the AI that knows your setup best

Copy it once and paste it in

After you enter your email, we send you one sentence with your private code inside it. Paste it into the AI that can see the most of how you work: your rules, your saved background and your past sessions. That is usually the one you do most of your work in.

Built for Claude Code, Codex, Cursor and tools like them

Your AI has to read our instructions and send your answers back, so it needs to be able to run commands and reach the internet. A chat app in a browser or on a phone usually can't do that. If yours can't, it won't be able to start, and nothing is sent.