Skip to content
K Koda Intelligence
All Dojos
AI Foundations Claude Code Cowork M365 Copilot NotebookLM ChatGPT Lovable Perplexity
Discipline 05 · 3 white, 8 brown, 4 black

ChatGPT Dojo

Run the Plus tier deliberately: model picker, Projects, Canvas, Agent mode. The capstone is a hardened custom GPT with a citation-audited report.

Your rank

Untied

15 modules · 178 prompts · 81 sourced
White · 3 modulesBrown · 8 modulesBlack · 4 modules
White Belt3 modules58 promptsNot started
Brown Belt8 modules72 promptsNot started
Black Belt4 modules48 promptsNot started
Version 2.0.10·Revised 2026-09-13·Refreshed weekly·46 sources cited
Inventory
Modules15
Prompts178
Prompts sourced81
Sources cited46
Your session
Modules woven0
Prompts ticked0
Minutes trained (est)0
Start training
// 01 / The ladder

Fifteen modules

White belt is an account configured on purpose. Black belt is a hardened custom GPT beside a citation-audited research report.

White Belt
Fundamentals
Not started
Capstone · White BeltKnot open
The Daily Driver
Turn a raw ChatGPT account into your configured daily driver, then prove it on real work. You will set up custom instructions and personality, learn what the model picker is actually choosing between, and drill the iteration controls until improving a prompt is faster than complaining about an answer. The pass bar is one real task from your week finished measurably better than your old way.
  • Write complete custom instructions covering who you are, what you work on, and how you want answers formatted, then verify a brand-new chat visibly obeys them.
  • Choose a personality setting and tune the tone controls, then ask the same question in a fresh chat before and after to confirm the difference is real, not imagined.
  • Ask one deliberately complex question on the default Auto routing and spot where ChatGPT switched to deeper reasoning, then force Thinking mode on the same question and note what the extra depth bought you.
  • Take one vague prompt from your real work and sharpen it twice using the edit control instead of re-prompting, keeping all three versions as a before-and-after record.
  • Use retry to generate three variants of one answer that matters, pick the winner, and write one line on why it won.
  • Branch a conversation at a valuable answer to explore a tangent, then return to the main thread and confirm it stayed clean.
  • Run one real task from your week end to end through the configured setup, from first prompt to usable output, and note the time it took versus your old way.
DeliverableA configured ChatGPT account (custom instructions, personality, model picker fluency) plus a one-page prompt portfolio: the vague-to-sharp rewrite trail, the three-variant retry comparison with your pick, and the finished output of one real task with its before-and-after time note.
Brown Belt
Advanced
Not started
Capstone · Brown BeltKnot open
The Context Engine
Build one Project that earns a permanent place in your week: grounded in your files, aware of your preferences, and able to draft, see, search, and compute. Every brown belt tool gets exercised inside a single container, and the pass bar is a recurring task that now runs measurably faster inside the Project than outside it.
  • Create a named Project for a real recurring domain with project-level instructions, and confirm they override your global custom instructions inside the Project.
  • Upload three to five grounding files within your plan's per-Project file limit and verify answers actually cite them rather than general knowledge.
  • Seed two saved memories relevant to the domain and verify both with a 'what do you remember about me?' check.
  • Draft a long-form document in an inline writing block and make at least two targeted section edits without regenerating the whole piece.
  • Upload a CSV and produce one chart plus one insight sentence with data analysis, then verify a computed figure by asking for the calculation behind it.
  • Generate one branded image with embedded text and correct one region with a targeted edit, then use vision to extract structured data from a screenshot or photo.
  • Run a web search question that returns citations and click through at least one source to confirm it says what ChatGPT claims it says.
  • Connect one app, use it inside the Project, then run your real weekly task end to end and record minutes saved against your manual baseline.
DeliverableA working Project with instructions, grounding files, seeded memories, and a connected app, plus one artifact per tool: an edited long-form draft, a verified chart, a corrected branded image, a vision extraction, and a cited search answer, capped with a timed end-to-end run of one real weekly task.
Black Belt
Mastery
Not started
Capstone · Black BeltKnot open
Mission Control
Ship an automation system that works while you do not: a custom GPT for the repeatable thinking, deep research for the hard questions, agent mode for the hands-on-web work, and a scheduled task that keeps the loop running. The pass bar is a real artifact produced on schedule, verified by you, without a single manual prompt on the day it shipped.
  • Build one custom GPT with spec-grade instructions (role, rules, output format, refusals), two to four knowledge files, conversation starters, and an app or action connection, then red-team it for leaks before sharing anything.
  • Run one deep research task with a deliberately edited plan and a site restriction, then audit five load-bearing citations before accepting the report.
  • Spin the audited report into at least two derivative deliverables with zero new facts introduced, verified by your own cross-check.
  • Run one agent task end to end including a takeover-mode login on an authenticated site, optionally inside Atlas if you are on a Mac, producing a deliverable you spot-checked.
  • Compute your monthly agent-quota budget in a table and design a task cadence that fits inside it with headroom for on-demand runs.
  • Create the recurring scheduled task at the cadence your budget allows and verify it at chatgpt.com/schedules with the correct next run time.
  • Chain at least three features (for example Project plus deep research plus scheduled task, or GPT plus app plus agent) into one workflow that produces a recurring artifact.
  • Write the runbook (owner, schedule, quota cost, failure signs, success metric), let the system run one full cycle untouched, and file the artifact it produced on its own.
DeliverableA live automation system: one hardened custom GPT, one citation-audited research report with verified derivatives, one completed agent task, and a quota-budgeted scheduled workflow verified at chatgpt.com/schedules, plus the runbook and the first artifact the system produced on schedule without manual prompting.
Beginner Intermediate Advanced Bar length is the module's prompt count against the longest module here; its segments are that module's own mix. Minutes are the curriculum's per-module estimates, summed over what you have woven.
One cell is one module, at its belt cloth Woven cell: module complete Knot open until the belt's last module is woven. Kept on this device only.
// SOURCES

Curriculum Record

Every published source this curriculum cites: 46 distinct references, counted from the curriculum brief and from the prompts themselves. Prompts that quote one carry a source line; the rest are marked unsourced author prompts.

  1. 01help.openai.comchatgpt-capabilities-overview
  2. 02help.openai.comgpt-53-and-gpt-54-in-chatgpt
  3. 03help.openai.comprojects-in-chatgpt
  4. 04help.openai.comchatgpt-agent
  5. 05help.openai.comchatgpt-custom-instructions
  6. 06help.openai.comconnectors-in-chatgpt
  7. 07help.openai.comwhat-is-memory
  8. 08help.openai.comwhat-is-the-canvas-feature-in-chatgpt-and-how-do-i-use-it
  9. 09help.openai.comdeep-research-faq
  10. 10help.openai.comtasks-in-chatgpt
  11. 11help.openai.comhow-can-i-use-gpts
  12. 12help.openai.comcreating-images-in-chatgpt
  13. 13help.openai.comchatgpt-search
  14. 14openai.comintroducing-chatgpt-atlas
  15. 15help.openai.comgpt-56-and-gpt-6-pro-in-chatgpt
  16. 16help.openai.comchatgpt-accuracy-and-limitations
  17. 17help.openai.commodel-release-notes
  18. 18help.openai.comprompt-engineering-best-practices-for-chatgpt
  19. 19youtube.comwatch?v=WTF5qphyy8w
  20. 20help.openai.comchatgpt-best-practices
  21. 21learn.chatgpt.comwhats-new
  22. 22expandi.iochat-gpt-rules
  23. 23zdnet.com10-chatgpt-prompt-tricks-i-use-to-get-the-best-results-faster
  24. 24aiproductivity.aichatgpt-tips-tricks
  25. 25learn.chatgpt.combest-practices
  26. 26forbes.comhow-to-use-chatgpt-to-streamline-your-week
  27. 27reddit.com100_practical_ways_to_use_chatgpt_to_be_more
  28. 28youtube.comwatch?v=XyLFcvkn2VE
  29. 29learn.chatgpt.comprompting
  30. 30felloai.comhow-to-use-chatgpt-effectively
  31. 31youtube.comwatch?v=Fhj3KiLoOYM
  32. 32juma.ai15-chatgpt-tips-from-experts-in-2024
  33. 33jeffsu.orgchatgpt-5-prompting-best-practices
  34. 34help.openai.comcustomizing-your-chatgpt-personality
  35. 35help.openai.commemory-faq
  36. 36help.openai.comtemporary-chat-faq
  37. 37help.openai.comusing-projects-in-chatgpt
  38. 38help.openai.comdata-analysis-with-chatgpt
  39. 39help.openai.comimages-in-chatgpt
  40. 40help.openai.comapps-in-chatgpt
  41. 41help.openai.comsharing-and-publishing-gpts
  42. 42help.openai.comusing-cloud-browser-in-chatgpt
  43. 43youtube.comwatch?v=B47FxHTHYrs&vl=en
  44. 44academy.openai.comresources/tasks
  45. 45help.openai.comchatgpt-release-notes
  46. 46help.openai.comscheduled-tasks-in-chatgpt

Get the morning Signal

175 editions so far, one a day. Unsubscribe anytime.