The source video presents 13 practical ChatGPT tricks. This training expands them into a durable 25-move system.
Use stable personality, tone and preference settings as defaults, then override them inside a task when needed.
Set stable preferences once; specify exceptions inside the task.Ask for assumptions, edge cases and verification. A phrase such as “think carefully” can request more care, but wording alone does not guarantee a hidden product mode.
Analyze carefully. State assumptions, check edge cases, then answer.Give representative samples and extract a style guide covering tone, rhythm, vocabulary, transitions and phrases to avoid.
Extract stable style rules from these samples.Preserve a clean baseline before exploring an alternative direction.
Keep this version intact; explore the alternative in a new branch.After iteration reaches a strong result, convert the conversation into a reusable specification or prompt card.
Produce the reusable specification that would recreate this result.Editing surfaces change over time. The durable workflow is draft, select, revise, compare and export.
Draft A. Revise only the selected section. Compare A vs B.Define inputs, outputs, validation, scoring, failure states and constraints before building.
Define inputs, outputs, validation, scoring and error states.External tools add power and risk. Connect the smallest trusted set needed for the task and understand permissions.
State what data is needed, what action is required and what stays read-only.Turn passive reading into one-question-at-a-time practice with hints, feedback and weak-area tracking.
Quiz me one question at a time. Give a hint before the answer.Save prompts because they work repeatedly, not because they sound clever. Store purpose, inputs, constraints and example output.
Convert this successful workflow into a reusable prompt card.Keep long-running workstreams together with files, instructions and conversation context.
Create project instructions: goal, scope, source-of-truth files and output standards.Project memory is about a workstream; reusable assistants/specifications are about repeatable behavior.
Ask whether this belongs to one project or should generalize across projects.Remove stale versions, duplicates and irrelevant files before analysis.
List authoritative, supplemental and ignored files before analysis.Tell ChatGPT to distinguish source claims, inference and outside research.
Use only the attached sources; label inference explicitly.Transform raw material into checklists, decision trees, SOPs, case types and templates.
Turn this source into a decision tree, checklist and reusable template.Start with the analytical question, inspect data quality, define metrics, test hypotheses, then visualize.
Profile the data, then test these hypotheses; create only useful charts.Check missing values, duplicates, outliers, type problems and inconsistent categories before trusting conclusions.
Produce a data quality report before analysis.Expose uncertainty with base, upside, downside and stress scenarios plus assumptions.
Build four scenarios with assumptions, triggers and early-warning indicators.Compress findings into recommendation, evidence, uncertainty, risks and next action.
Create a one-page decision memo from the analysis.Audit the output for factual, logical, numerical and constraint errors instead of asking the same pass to trust itself.
Audit the prior answer only for factual, logical and constraint errors.Use quick search for current facts and navigation; use deeper research for multi-source synthesis and comparison.
Decide whether this needs quick search or deep research before answering.Recurring tasks should explain what changed, why it matters and whether action is needed.
Summarize only meaningful changes and flag anything requiring action.If you care only when a condition changes, use a threshold or event watch rather than a generic recurring report.
Notify only when the condition crosses this threshold or materially changes.Extract recurring operational knowledge into prerequisites, steps, exceptions, QA and rollback.
Turn what we solved into an SOP with steps, exceptions, QA and rollback.Combine settings, projects, trusted sources, prompt cards, verification, automation and knowledge capture into one repeatable system.
Design my AI operating system across settings, projects, prompts, tasks, sources and SOPs.Stable Settings → Project Context → Trusted Sources → Reusable Prompt Cards → Data & Research Workflows → Verification → Scheduled Intelligence → SOP / Template Capture
Original video: 13 Trucos OCULTOS de ChatGPT — Watch on YouTube
This Article-Training uses the video as its starting point and expands the original ideas into a 25-move practical learning framework.
Complete all 16 practice cases and enter your name.