AI Introspection Rubric
Point an AI agent at a folder of your own notes to surface criteria you hold implicitly and turn them into a rubric.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 85%
Many people struggle to articulate the criteria they actually use — for hiring, craft, or judgment — because those criteria live implicitly across scattered notes. Jenny Wen's practice is to give an AI agent (Claude Co-work) a folder of her own writing (interview notes, one-on-ones, memos, random thoughts) and ask it to read everything and produce a structured rubric or memo. The output reveals things she'd been expressing implicitly but never named.
Origin
A concrete workflow Jenny Wen described on Lenny's Podcast, using Claude Co-work on her personal notes folder.
Core principles
- 01Your real evaluation criteria are often implicit and scattered, not written down
- 02An agent can read a whole corpus of your notes at once and find the pattern
- 03The goal is to make implicit judgment explicit and reusable
- 04Works for self-insight generally, not just hiring rubrics
- 05Co-work suits this because it handles longer-running analysis of messy inputs
How to run it
- 1
Gather your raw notes in one folder
Collect the unstructured material where your implicit views live — interview notes, one-on-one notes, memos, tiny thoughts — into a single folder, regardless of how messy.
Pro tip It doesn't matter how disorganized the folder is; the value is in the volume of authentic first-hand material.
- 2
Ask the agent to read it and build a rubric
Prompt the agent to read all of it and craft a specific artifact — e.g. 'help me craft a memo rubric for how I assess craft in interviews.'
Pro tip Name the concrete deliverable (a rubric, a memo) and the exact use, not just 'summarize my notes.'
- 3
Review what it surfaces about you
Read the output for criteria you were applying implicitly but never articulated, and use it as a reusable evaluation tool.
Pro tip Treat surprising items as the payoff — the things you 'wouldn't have realized' are the point of the exercise.
Watch out It reflects only what's in your notes; gaps or biases in what you wrote down will carry into the rubric.
In the wild
Jenny had Co-work read her interview notes and design-craft memos and generate a rubric for evaluating craft in interviews — articulating something she said many people struggle to put into words.
→ She got an explicit, reusable rubric plus new self-insight into criteria she'd been holding implicitly.
Common mistakes
Asking for a vague summary instead of a concrete artifact
Prompting 'summarize my notes' yields generic output; the value comes from specifying the exact deliverable and its use (e.g. a craft-assessment rubric for interviews) so the agent extracts decision criteria.
Treating the rubric as objective truth
The output only reflects what you happened to write down, so blind spots and biases in your notes get encoded into the rubric; it's a mirror, not an oracle.
Is it for you?
Best for
Leaders and individual contributors who make repeated judgment calls (hiring, craft evaluation) and want to make their implicit criteria explicit
Not ideal for
Decisions where you have little written history to draw on, or where criteria must be set from external standards rather than your own past behavior
From the transcript
“I just had it read through all of my notes, both like interview notes and other things that I cared about and memos and stuff…”
“and then it made me this rubric”
“read it and then help me craft a uh like a a memo rubric for how I assess craft in interviews”
“I wouldn't have realized even these things about myself”
From the episode
The design process is dead. Here’s what’s replacing it.
Jenny Wen (head of design at Claude)