Tuukul — daily thinking practice
The three gates, the daily check-in, the Gym, and your calibration curve.
- A personal calibration curve computed from your real AI work — when you say 90%, how often are you right?
- Five-minute daily reps (morning intent, evening reconcile) that build the judgment habit
- Everything is yours: features-only capture, learner-owned memory, one-tap export and erase
Think before you prompt
Tuukul (Yucatec Maya: thought) is the thinking strand of your AI-Fluency practice. Working with AI is a chain of judgment calls, and Tuukul installs a deliberate pause at the three moments that decide everything: Delegate (what exactly am I asking for, and what would tell me it's wrong?), Accept (did I verify this, or did it just sound right?), and Deploy (if this ships and fails, who is affected — and did I sign it?).
The daily heartbeat
Today holds the two three-minute rituals. In the morning, name the most consequential thing you'll delegate to AI, make one falsifiable prediction with a confidence, and tap what it touches if it goes wrong. In the evening, reconcile: was the prediction full, partial, or a miss? Surprise is information — a high-confidence miss is the most valuable data point you'll collect all week. Check-ins feed a weekend-safe streak.
The Gym, your Journal, your Score
The Gym holds six short drills that always run on your real work — decomposition katas, premortems on your actual next launch, inversion, estimation ladders, a war-room scenario and explain-it-back. Scores are heuristic v1 and say so. The Journal drafts itself from your sessions; you confirm each entry in under a minute, and over months it becomes the observed-practice portfolio behind the higher Itz'at credential levels. My Score shows the nine dimensions (A–H + T) with a "how computed" note on every number.
Coach, not cop
Raw prompt and response text never leaves your machine — the upload schema has no field for it. Your organization sees cohort aggregates of five or more people, never you. And My Memory is yours: view, search, edit, delete and export everything the coach remembers about how you work.
The journey, step by step
| Step | What you do | What you get |
|---|---|---|
| Morning check-in | Name the most consequential thing you'll delegate to AI today, predict the outcome with a confidence, tap the blast radius | A logged prediction at the Delegate gate — the raw material of your calibration curve |
| Work normally | Do your day; connected tools (like the Claude Code plugin) observe judgment signals — never your prompts | Verification, iteration and gate signals accrue to your private score |
| Evening reconcile | Score each open prediction full / partial / miss and add one line of reflection | The curve updates from reality; a journal entry drafts itself for you to confirm |
| Gym reps | Run a 5–10 minute drill against your real backlog — decomposition, premortem, inversion, estimation | Rubric-scored thinking reps that feed the T-strand of your Itz'at Score |
Frequently asked
Does Tuukul read my prompts?
No — structurally. The upload schema has no field for prompt or response text; only judgment signals (counts, flags, your own short predictions capped at 240 characters) ever sync. You can verify this: the ingestion endpoint rejects any raw-text-shaped field.
What does my organization see?
Aggregates over groups of five or more, never your individual rows — enforced by the database itself, and every metric shows how it was computed. Your coaching stays between you and the coach.
Why does my score say 'thin data' instead of a number?
Honesty by design: a dimension renders only after enough evidence events, and the calibration curve appears only after 10 resolved predictions. Missing data is never shown as zero.
Can I delete what it knows about me?
Yes — every memory is editable and deletable in My Memory, and the data-request control exports everything as JSON or erases the whole record. Memories you've edited are never overwritten by the system.