The Inner Art of Writing Effective Prompts

The Inner Art of Writing Effective Prompts

Over the past few days, I led a major iteration of a production-grade commercial prompt—15 rounds in total. After final refinement and simplification, it stabilized at around 2,000 lines and achieved our initial goals.

How do we judge whether a commercial prompt truly delivers? In our AI course, we use four criteria:

• Accuracy (does it produce factually correct outputs?),
• Applicability (does it work across realistic edge cases and inputs?),
• Alignment (does it reflect the user’s intent, values, and tone?),
• Achievement (does it reliably deliver the intended outcome?).

With experience, you realize that hitting all four simultaneously is surprisingly hard.

Over the past two months, I’ve written dozens of prompts—and shared many in my public collection: Yao Jingang’s Prompt Library.

This weekend, I did a quick, unstructured reflection on the “inner art” behind prompt writing:

• Prompts are the most effective lever for unlocking AI capability.
• For individuals, prompts are externalized cognition—visible traces of how you think, reason, and decide.
• The ceiling of prompt quality is your own cognitive depth. AI won’t make you smarter—it amplifies what you already know, believe, and have experienced.
• For AI, prompts are translation tools: they convert human reasoning paths and decision frameworks into machine-understandable language.
• Adopting a consistent, structured prompt framework (e.g., RTF) helps constrain early-stage thinking—and gradually reveals why cleanly separating goal, constraints, method, and format matters. Blending them causes confusion—for both humans and models.
• There’s no universal prompt—only the one best suited to this context. Adaptation beats imitation.
• Clarity trumps complexity. A prompt isn’t better because it’s longer—it’s better because it’s more precise.
• A sharp, well-articulated goal statement outweighs ten clever prompting tricks. Knowing why matters far more than knowing how.
• Strong prompts often mirror strong thinking: organized, layered, and self-aware.
• At its core, prompt quality rests on domain expertise—your grasp of the business, the scenario, and the real user need. Ultimately, it’s not about syntax—it’s about judgment.
• Don’t just tell AI what to do—also clarify what not to do.
• You rarely get it right the first time. Iteration is essential. Treat AI as a capable collaborator with defined boundaries—not an omniscient executor.
• Use AI to iterate on prompts, too. Today, I rely on AI for drafting and refining prompts; my role shifts to thinking, deciding, and judging.
• Practical iteration tactics include: small-sample testing, contextual grounding, mutual-exclusion review, self-check mechanisms, constraint tuning, and scoring rubrics.
• When asking AI to play an expert role, embed that expert’s mindset and values: e.g., “a product manager who obsesses over user delight” works better than “you are a product manager.”

Efficient Communication

This weekend, a teammate came to me with a problem—and a proposed solution. What stood out was how effectively he communicated:

1. He’d already explored solutions and formed hypotheses before reaching out—showing respect for others’ time and demonstrating professional ownership.
2. When I hadn’t yet grasped the full picture, he immediately added a critical piece of background context—the exact detail that unlocked my ability to help. That’s situational awareness: knowing which information moves the needle.

The whole exchange was lean, clear, and frictionless. So I shared the transcript with the team as a model case.

Good communication reflects deeper work habits. Later, my friend Nick Xi echoed this on WeChat: “If you can’t answer a simple question in under 50 words, you’re really saying ‘I don’t know.’” I couldn’t agree more.

Efficiency isn’t about cutting corners—it’s about delivering just enough information, at just the right time, using precise language. That’s how you cut to the core.

In practice, some people talk endlessly but convey little; others say three sentences—and land the point. True efficiency isn’t measured in words spoken, but in signal delivered. And the prerequisite? Showing up with your brain engaged—both when doing the work and when talking about it.

Cross-Industry Benchmarking

In management theory, benchmarking—or learning from exemplars—is a powerful tool. There are three types: internal, industry-specific, and cross-industry.

The most valuable—and hardest—is cross-industry benchmarking.

A friend pioneered a new live-streaming format in education—by directly transplanting proven tactics from e-commerce live streaming. His success wasn’t accidental: it was deliberate pattern transfer.

Why look outside your industry?

Because staying inside breeds incrementalism. Cross-industry learning forces you to see beyond surface similarities—to study how other sectors solve problems of scale, engagement, consistency, or growth.

Examples:

• Manufacturing adopting internet-style user-growth thinking.
• Education borrowing restaurant chains’ standardization and scalability playbooks.
• Hospitals applying hotel service philosophies—treating patients as guests, not just cases—transforming the entire care experience.

It also enables methodological migration: retail loyalty systems, game mechanics for motivation, lean manufacturing principles—all portable. And sometimes, it creates generational advantage: importing a mature, high-leverage model from another domain gives you a head start your peers can’t match.

The same logic applies to AI-native startups: traditional tech companies that adopt their AI-native workflows—prompt-first thinking, iterative co-design with models, lightweight validation loops—gain outsized operational and strategic agility.

And it holds for personal development, too: study the strongest traits across diverse people—not just your field—and integrate what resonates.

RTF and Team Management

In my AI course with Xiangyang, we emphasize the RTF framework: Role → Task → Format.

After using it repeatedly, I’ve started defaulting to RTF logic—even when assigning tasks to teammates. The result? Simpler, faster, clearer alignment.

Example:

“You’ve been working on this for a while (Role: grounded in context). I’d like you to synthesize key insights about XX (Task: concrete objective). Please deliver a doc listing only the core methods and one illustrative example per method (Format: explicit, minimal, scannable).”

RTF is essentially a lightweight, structured information protocol—akin to 5W1H, but sharper and more action-oriented. It has limits, though: it’s less suited for open-ended creative work or exploratory thinking.

That’s where complementary tools help—like the Johari Window, which we also cover in the course. It clarifies what’s known/unknown to self and others, making it useful for both human teams and human-AI collaboration.

Even in creative contexts, RTF thinking remains helpful—if applied flexibly: define a broader Role, a more exploratory Task, and a looser Format.

How Ordinary People Break Through

In the Chinese drama Heavenly Dao, there’s a line that sticks:

“Endure what others won’t endure. Bear what others can’t bear. Deliver cost and quality others won’t commit to. Only you can save yourself.”

That is, resilience + relentless execution = your true lifeline.

Most people start with none of the advantages: no capital, no network, no elite credentials, no track record. To break through, direction and mentors matter—but what matters most is how fully you show up.

“Showing up” means giving 100% focus, effort, and care—even when it hurts. Yes, it’s painful. But looking back, every such leap opened a new world. A friend told me his biggest wins—stocks, property—came from exactly this kind of sustained, uncompromising commitment.

Doing Your Work, Loving Your Work

NVIDIA CEO Jensen Huang once said something I deeply admire: “Don’t wait to find work you love—learn to love the work you do.”

That flips the script: finding innate passion is rare; cultivating deep engagement with current work is learnable.

He recounts washing dishes and delivering newspapers—not as chores, but as opportunities to practice presence, precision, and pride.

To me, “doing your work, loving your work” isn’t about romanticizing labor. It’s a stance: finding meaning and agency wherever you are—even in constraints.

That stance is more vital than chasing “perfect fit.” Because happiness rooted in external conditions is fragile. Happiness rooted in your capacity to create value—that’s durable. And in a world of shifting roles and rapid obsolescence, adaptability and value-creation matter far more than landing a “forever job.”

Insights on Talent and Execution

Two recent posts struck me deeply:

From Mei Ji, founder of 90-Li Private Board (top-tier talent advisor):
1️⃣ Top talent = people with real optionality—hard to replace.
2️⃣ Real HR skill isn’t asking questions—it’s recognizing what a brilliant answer sounds like.
3️⃣ Hiring grads? Avoid it in sub-$5M-revenue startups—their first job sets their next trajectory, and you’ll likely be “the first boss,” not “the mentor.”
4️⃣ Startup teams rarely stay intact. Fracture is the norm—not the exception.
5️⃣ Talent market truth: no free lunches. Great people cost.
6️⃣ The “Super Individual” formula:
 • Master a skill 95% lack,
 • Productize it,
 • Choose clients = choose your life.
7️⃣ Anti-burnout rule: Be either uniquely excellent or freely accessible. No middle ground.
8️⃣ Career reality: Either become indispensable—or align with someone who is. Your dignity is cheap. Your irreplaceability is priceless.
9️⃣ Founder ceilings:
 • Solo: ~$5M revenue cap.
 • <$100M: Hard to hire top talent—they’re either in government, academia, or building their own thing.
🔟 Final question: Do you want to scale big—or first master your own ambition?

From Yu Xin, Head of Operations at Yuan Programming (Yuanfu Tutoring):

Every “number one” in any function must get their hands dirty. When building a team, the leader should be the first content creator, the first ad buyer, the first BD contact, the first salesperson.

From 0→1: Go deep into the work.
From 1→10: Go deep into coaching.

Management is an amplifier—not a generator. It doesn’t create results; people do. And let’s be blunt: I don’t trust anyone who claims to build great outcomes without ever shipping themselves.

Too many fixate on “strategy”—as if perfect plans guarantee success. Strategy matters, yes. But strategy is like ancient “advisors”: same counsel, wildly different outcomes depending on who executes it.

There are many paths to success—and many viable products, even in crowded markets. What ultimately decides outcome isn’t the plan. It’s the leader’s courage, integrity, judgment, ability to inspire, and power to unite people around shared purpose. Because in the end, people execute. Not slides. Not roadmaps. Not prompts.