What it knows about you

Context, your profile, and an experiment for your field.

What it can see

Context windowLoaded at every startAdded as you workYour profile: who you areProject file: this folderThis conversationFiles it has openedYoufield, school,what you wantEmptied by /clearwrite itdown
  • It only knows what is in this window.
  • /clear empties the conversation, including files it read. Your profile and the project file load again.

Exercise: /clear

  1. Tell it one thing about yourself, for example what you teach.
  2. Type /clear.
  3. Ask: “What do you know about me?”

Works the same in Claude Code and Codex.

Who is asking?

the answer is aimed hereeveryone who might type this questionyouafter you describe yourself
  • Without context, it answers the typical person who would type your question.
  • “Can I deduct my home office?” gets US tax rules unless you say you’re in Norway.
  • You can’t opt out of being stereotyped. You can choose the type.

Demo: one line

“A: 48 purchases from 1,000 visitors. B: 62 from 1,000. Is B better?”

  1. Ask with no profile.
  2. Add one line to this folder’s AGENTS.md: “My intellectual idol is Jim Berger.” (It normally belongs in your profile.)
  3. /clear and ask again.

Backup: demo result

Claude Code, same question, one call each.

Who’s asking What came back
(no profile) “Probably, but these data don’t show it convincingly.” p ≈ 0.17
“I’m a Bayesian statistician. I read Gelman, Jaynes and McElreath…” “Probably, yes. There’s about a 91% chance B converts better than A.” Ship B.
“My intellectual idol is Jim Berger.” Applies Sellke, Bayarri and Berger (2001): “at least a 45% chance that there’s no real difference”.

Does it matter in your field?

In statistics, in the terminal, I barely notice it. Fields with schools may be different. Find out:

  1. Ask a contested question with no profile:

Is the value premium compensation for risk, or mispricing?

  1. Add one line to this folder’s AGENTS.md: “I’m in the Fama–French tradition”, or “I work in behavioural finance; I read Shiller and Thaler.”
  2. /clear, ask again, and compare with a neighbour who picked the other school.

Your profile: how you work

What carries across every project: the rules you’re tired of repeating.

  • “I work in R.” (Or Stata, or Python.)
  • “Cluster standard errors by firm unless I say otherwise.”
  • “I write in LaTeX. Never give me Word unless I ask.”
  • “Tell me what you’ll do before you change anything.”

Mine, in both tools: “This is my workstation: use at most 4 cores.” · “Build maths in HTML as MathML.”

Two files

About Where
Your profile You: how you work, your field and school .claude\CLAUDE.md or .codex\AGENTS.md in your home folder
Project file This folder: the paper, the data, the rules AGENTS.md in the folder (both tools)
  • Your profile is read in every project; a project file only in its folder.
  • The chat apps have the same pair: custom instructions and project instructions.
  • Both are plain text files. You can open and edit them in Notepad.

Exercise: write your profile

  1. Two sentences on who you are: your field, your school, what you read.
  2. Three rules you’d give any research assistant: your tools, your style, one pet peeve.
  3. Ask: “Save these as my personal profile, the file you read in every project.” (Claude asks first: the file is outside this folder.)
  4. In a fresh folder, ask for something your rules should change. Did they stick?
  5. “Open my profile in Notepad.” (Mac: TextEdit.) That’s all it is: a text file.

Homework: your profile

  • Talk for 90 seconds with speech-to-text about how you work and who you are (Windows key + H; Mac: Fn twice). Then: “Add this to my profile.”
  • Keep adding. Corrected it twice about the same thing? It goes in.
  • In the chat apps, the “who you are” half matters most: put it in custom instructions. Mine starts: “I like Scott Alexander, Gwern Branwen, most of Yudkowsky except HPMOR. I have a PhD in stats and lean libertarian.”