Text below translated by DeepL
About 26 years ago, I used to hang out on cybersecurity forums at internet cafes. Since the internet was still in its infancy in China at the time, many people didn’t know how to navigate interactions in an anonymous online world, and a lot of the questions were quite naive. To help manage the forums, the moderators posted “The Art of Asking Questions.”
In the AI era, if we want to efficiently extract value from AI’s vast knowledge base and fully leverage its capabilities, we need to master the art of asking questions. This is the key to success in the new era—and this skill alone will set you apart from the majority of users.
I’ve been interacting with ChatGPT, and this guy has given me a lot of advice. It distilled various question formats into dozens of key points. By defining constraints, it facilitates better human-AI interaction, allowing users to ask questions in a way that aligns with AI’s algorithmic logic while using natural human language. Through a simple chat window and the art of questioning in the AI era, asking questions becomes a tool for extracting the “API of thought.”
I’ve simplified it and extracted some key points to share:
- Variables
This refers to “relationships” or “correlations,” which encompass causality, constraints, and other factors. For example, the interaction between A and B.
- Structure
This refers to “two-dimensional” or “three-dimensional” structural diagrams, such as changes in the state of A—whether linear, cyclical, or grid-based—or the hierarchical relationship between A and B.
- Boundaries
This refers to the scope of the topic, providing clear boundaries. This includes the boundaries of hypothetical questions.
- Density
This refers to the actual information content that the topic can convey; it is not merely about the quantity of numbers listed. It includes the level of detail in the information and the elimination of unnecessary verbiage.
- Emotion
Human-expressed emotions, including the bias present when asking questions, are rationally assessed by AI to determine whether they can align with the shared topic.
- Perspective
This refers to the level at which the issue is addressed and the dimensions from which it is approached.
- Information Sources
AI requires access to a broader range of information to form a more effective contextual assessment of the current issue; this essentially also defines the scope of constraints.
- Privacy
Take care to protect personal privacy during the conversation.
- Risk
AI positions itself as a strategy provider and assumes no liability for risks. This is the root cause of potential errors in its responses; therefore, you should use descriptive language to mitigate risks and minimize risky answers.
That sounds pretty abstract, so let’s use a concrete example that everyone “loves”:
“I’m broke, the world is terrible, I lost my job at Company X—how can I make money?”
Everyone asks questions like this, but they’re too vague. Not only do they require multiple interactions between the AI and the user (and the more interactions there are, the more confusion arises), but the AI’s responses are often neither accurate nor applicable—let alone capable of forming a coherent plan. Moreover, the emotional language in the original statement reduces the likelihood of empathy and consensus. Simply put, the AI perceives you as someone without productive resources and as an emotional individual. From its perspective, how can it possibly provide you with a better plan for making money? Additionally, personal privacy must be protected.
To be more specific, remove the emotional language, establish clear boundaries, increase the density of information, and minimize privacy leaks:
“In 2026, the XX industry I’m in is struggling. I’ve worked for over XX years and have extensive industry experience. What should I do if I seek employment? What should I do if I start a business? My current skills are XX and XX. How can I make money in this environment?”
This is much better. Although it provides the AI with a general self-portrait and the context of your situation, it’s still not an excellent question.
Add variables:
“In 2026, the XX industry I work in is in a slump. I have worked for over XX years in the XX department, applying skills such as XX and XX. Because AI technology has replaced my skills in XX and XX, and because the XX part of my job can be quickly handled by an agent, the company is downsizing to cut costs, leading to my unemployment. How can I leverage my existing skills to successfully transition within XX time, whether through re-employment or starting my own business, and make money in this environment? ”
Structure:
“In 2026, the XX industry I work in is in a downturn. I have worked for over XX years in the XX department, applying skills such as XX and XX. Because AI technology has replaced my skills in XX and XX, and because the XX portion of my work can be quickly handled by agents, the company is downsizing to cut costs, leading to my unemployment. The industry I’m in is one where company structures are rapidly flattening in the AI era. Its future growth is likely to be lackluster, but my personal goal is to pursue stable income. I have XX hours available for learning each day and approximately XX dollars to invest. My advantage over the average employee in this industry is XX. How can I leverage my existing skills to complete a personal career transition within XX time, whether through re-employment or starting my own business, and make money in this environment?”
This is already a very detailed question, but it’s a bit disorganized. By providing a clear perspective and giving the AI a framework to follow, we can help it connect the dots and minimize risks:
“In 2026, the XX industry I work in is in a downturn. I have been working for over XX years in the XX department, where I use skills such as XX and XX. Because AI technology has replaced my skills in XX and XX, and because the XX portion of my work can be quickly handled by an agent, the company is downsizing to cut costs, leading to my unemployment. My current industry is one where company structures are rapidly flattening in the AI era. Future growth is likely to be lackluster, but my personal goal is to pursue a stable income of at least XX per month. I have XX hours available for learning each day and approximately XX dollars to invest. My advantage over the average employee in this industry is XX. How can I leverage my existing skills to complete a personal career transition within XX timeframes—whether through re-employment or starting my own business—and generate income in this environment? If I seek re-employment, what paths can I take to ensure I remain competitive for the next N years? If I start a business, how should I structure my business plan based on my capital and investment to maximize the likelihood of profitability before cash flow dries up? Please provide a risk assessment for both the employment and entrepreneurship paths, including risks related to time, capital, competition, and market capacity.”
This is an excellent question. Although it isn’t entirely concise, refining it across several dimensions will ensure the answer aligns as closely as possible with your needs.
ChatGPT’s optimized version:
In 2026, the XX industry I work in is in a slump. I have XX years of experience in the XX department, with core skills in XX and XX. However, the XX portion of my work has been replaced by AI (which can be handled by an agent), leading to my unemployment.
Current industry characteristics: Flat organizational structure, limited growth (primarily competition for existing market share).
My constraints are as follows:
Time: I can dedicate XX hours per day (for XX months)
Funds: I can invest XX USD (no additional funds available)
Risk tolerance: I can withstand up to XX consecutive months without income
My goals (ranked by priority):
Achieve a stable income of ≥XX/month within XX months
Avoid high-volatility/unpredictable income models
Leverage existing skills as much as possible (to reduce the cost of learning from scratch)
Based on the “Resources-Time-Risk” constraints, please provide the following for each option:
Re-employment Path (phased: 0–3 months / 3–6 months / 6–12 months)
Entrepreneurial Path (specify: customer acquisition methods, cash flow structure, payback period)
And provide the following assessments:
Probability of failure for each path (qualitative or range)
Whether break-even can be reached before funds are exhausted
Whether the path relies on luck, platforms, or traffic dividends
If neither path is feasible, please directly identify the points of conflict among the constraints and provide recommendations for adjustments.
Essentially, this type of questioning isn’t about making things easier for the AI; rather, by establishing clear boundaries and reducing the search for ineffective paths, it makes the results more predictable. This benefits both users and the AI, allowing users to better leverage the AI’s capabilities while providing the AI with more precise feedback.
I reserve my opinion on this version of GPT. If anyone is reading this, please consider the approach that works best for you.
郄磊 ( Qie Lei )
April 5, 2026, 11:00 PM UTC+8