YannisKiefer/the-ai-diary

attention-ledger

Use this to account for where AI-assisted attention actually went and what it bought - including the uncomfortable parts.

audit-router

Use this as the always-on brain for any question about how someone thinks, decides, debugs, or works with AI. Read the request, find the first match, act.

bias-candidates

Use this to test decision patterns against repeated behavioral evidence - without ever labeling from one anecdote.

blindspot-miner

Use this when analysis is complete. Mine for findings the subject cannot see from inside any single session.

calibration-audit

Use this to compare what was expected against what actually happened, and find where confidence systematically outruns outcomes.

cognitive-loop-map

Use this to reconstruct how someone repeatedly moves from intent to learning, and how that loop shifts by task type.

collaboration-audit

Use this to map the full human-AI working loop - prompts in, verification out, understanding kept or leaked.

contradiction-finder

Use this to surface stated values that collide with measured behavior, and strengths whose shadow side shows.

debugging-trees

Use this on bug hunts and ambiguous problems. Reconstruct the hypothesis tree and watch how the mental model changes.

episode-method

Use this before interpreting anything in the corpus. Attribute behavior correctly, then rebuild work episodes as the unit of analysis.

evidence-hygiene

Use this whenever the corpus might contain secrets, customer data, or anything private. Protect it while analyzing it.

failure-loops

Use this to classify what happens right after a first attempt fails, and which second moves pay off versus spiral.

framing-planning

Use this to measure how problems get represented and how much planning actually happens before action starts.

learning-tracker

Use this to trace lessons across months - learned once, learned slowly, forgotten, or never turned into a system.

model-matrix

Use this to compare outcomes across AI models and decide which model serves which kind of work. No global rankings.

prospective-kit

Use this after the report ships. Turn open questions into falsifiable predictions and instrument future sessions lightly.

report-writer

Use this to assemble the final audit in the exact order that lands - one page, ten findings, then depth.

source-discovery

Use this to inventory every accessible history of human-AI work before analyzing anything. Discover paths; never assume them; measure coverage honestly.