What Determines Price?

Lately I’ve been researching new product pricing—and it’s deepened my understanding in unexpected ways.

  1. More and more people now recognize that price—whether for physical or digital goods—is not determined by cost, even though cost-based pricing remains one common method.
  2. Fundamentally, price is the outcome of value and negotiation: neither purely cost-driven nor demand-driven, but an equilibrium shaped by multiple interlocking forces.
  3. These forces include cost, supply and demand, perceived value, and multi-party negotiation—and they constantly influence one another.
  4. Cost’s most essential role is setting a floor, but that floor shifts with strategic intent. For example, Didi subsidized rides heavily during its expansion phase; Meituan and Ele.me did the same in their recent food-delivery wars—where “cost” was deliberately ignored to capture market share.
  5. Supply and demand strongly shape pricing power. In near-monopoly markets—like certain life-saving drugs—companies command high margins. In perfectly competitive markets, pricing power vanishes. That’s why positioning theory emphasizes creating new monopolies: through differentiation, you carve out a unique category or reference frame—effectively building defensible pricing power.
  6. Beyond functional value, perceived value often matters more: brand, design, storytelling, user experience, and trust can multiply price—even for identical products. Add financial value (e.g., scarcity, collectibility, or store-of-value properties, like Moutai), and pricing power intensifies further.
  7. A special case lies in consulting and AI services: value scales dramatically when embedded in real business contexts. Huawei paid IBM $400M for management consulting—not because IBM offered generic frameworks, but because its consultants immersed themselves in Huawei’s operational details, redesigning workflows end-to-end. Similarly, an AI consulting tool gains outsized value only when tightly coupled to a client’s specific business logic and pain points.
  8. Negotiation adds another layer: the same price may close easily with one customer segment or context—and fail completely with another. That variance reflects shifting power dynamics among buyers, sellers, competitors, regulators, and even internal stakeholders.
  9. To summarize:
     • Cost sets a strategic floor (fluid, not fixed);
     • Supply and demand define the market envelope—the range within which price can float;
     • Value determines premium potential, including functional, perceived, financial, and contextual value—i.e., the ability to land precisely where business operations live;
     • Negotiation skill governs conversion capability: turning value and positioning into actual deals.

AI and Business Acceleration

On Thursday, I joined Xiangyang and Huangshu at a 43 Talk AI salon centered on Vibe Coding.

From my perspective as a founder and product leader, I presented: How Vibe Coding Accelerates AI Product Commercialization.

I walked through real examples—how I’ve used Vibe Coding to solve concrete, business-adjacent problems, fast.

Before diving in, I posed a question I’ve long wondered about:
AI is advancing rapidly—and boosting individual productivity dramatically. But at the company level: what’s AI’s actual contribution to overall business acceleration?

The informal poll yielded a striking result: fewer than 20% believed AI meaningfully accelerated their organization’s core business. Most saw near-zero impact.

That surprised me—until I reflected. For legacy enterprises, internal “baggage” is immense: entrenched systems, siloed incentives, risk-averse culture, and fragmented ownership.

Technically, bridging AI into real enterprise workflows remains a “last-mile” challenge:

  • Process integration: AI doesn’t operate in isolation. Embedding it into existing ERP, CRM, or operational systems demands deep, often messy, integration—not just API calls.
  • Cultural adaptation: Many teams still default to manual, linear workflows. Widespread skepticism—or outright resistance—means cultural transformation lags far behind technical capability.

Yet some companies are succeeding. Over the past few months, I’ve seen several reimagine entire business processes with AI—not as a bolt-on, but as a redesign catalyst. Their efficiency gains are dramatic.

Who makes this happen? Rarely a single specialist. The people who crack the “last mile” consistently share three traits:
 • Deep domain expertise (they know the business inside out),
 • Real decision-making authority (budget and hiring power),
 • Strong AI fluency (not necessarily coding—but knowing what’s possible, how to evaluate it, and where to intervene).

Few hold all three. Most who do are CEOs or COOs.

That explains why startups—lean, integrated, and founder-led—often outpace incumbents here.

So that 20% figure isn’t just about tech maturity. It’s a proxy for organizational readiness.

Vibe Coding

Over the past two years, my professional identity has quietly shifted toward AI Education and AI Marketing.

The core idea? Rebuild my prior expertise—marketing frameworks, growth tactics, content strategy—through the lens of AI capabilities.

This process has deepened both my fascination with and practical fluency in AI.

For me—and for our business—AI has been a powerful accelerator. Especially as a non-engineer, the speed and quality leaps have been genuinely startling.

Take Vibe Coding. It’s transformed how we ship AI-related work.

A typical example: tasks that once took a week now take 15 minutes—or less—while matching or exceeding prior output quality.

It’s not just speed. Collaboration quality improves too: tighter feedback loops, faster iteration, clearer alignment between intent and execution.

My current Vibe Coding use cases fall into five buckets:
 • Internal business ops tools (e.g., dashboards, reporting automations),
 • Marketing experience tools (interactive demos, personalized landing flows),
 • High-conversion marketing landing pages,
 • Company website sections (especially dynamic, AI-updated content),
 • End-to-end prototypes for AI products (frontend + backend logic + mock integrations).

These are classic low-code/no-code domains—low technical barrier, high business relevance. More importantly, they let us embed AI directly into real workflows, making its impact tangible—and building confidence to tackle harder, more systemic challenges.

Professional Judgment

I recently heard investors say they refuse to fund startups whose pitch decks are AI-generated. Not because AI is “bad”—but because an AI-made deck often signals something deeper: no thoughtful editing, no team-specific logic, no authentic reflection of the founders’ own reasoning.

AI excels at generating things that look impressive at first glance: polished decks, fluent proposals, “sophisticated” jargon-filled narratives.

But look closer—and many collapse under scrutiny. Logic gaps appear. Assumptions go unchallenged. Business realities get glossed over.

I’ve reviewed dozens of deliverables where the “DS flavor” (a telltale sign of AI over-reliance) was unmistakable—yet the underlying analysis couldn’t withstand basic operational review.

When should we use AI? How far can it take us? When should we trust its output—and when must we override it? These questions test something fundamental: professional judgment.

AI is powerful—but it’s not autonomous. On critical decisions—product strategy, pricing, hiring, partnership terms—it still needs human guidance, standards, and guardrails.

Without judgment, AI’s “impressive” outputs become unusable—not malicious, but misaligned: impractical, unrealistic, or simply detached from ground truth.

Becoming that kind of “expert” doesn’t require decades of experience. It does require deeply understanding the first principles and operational standards of your domain—the “why” behind every “what.” Only then can you direct AI effectively—and know when to say “no.”

Here’s how I approach new tasks, ideas, or requirements:

  1. Ground first: Draw on experience and intuition to sketch core logic—maybe just 3 sentences, or 5–10 keywords. Establish a mental anchor.
  2. Probe with AI: Run initial prompts with GPT-5 or Claude 4.1—not to generate answers, but to stress-test my assumptions. Are there blind spots? Unexamined dependencies? Does the emerging picture still align with real-world constraints?
  3. Draft manually (or voice-to-text): Create a rough version myself—or ask AI to structure my words. Then invite AI to supplement, not replace. I review every addition: Is it accurate? Relevant? Actionable? I decide what stays—and what gets cut.

In this workflow, AI becomes a supercharged assistant—not a thoughtless autopilot.

A Hoopoe

Today’s morning run: 22 km, 396 m of elevation gain.

Midway up a long, grinding climb, a hoopoe landed just ahead of me—neither too close nor too far.

Each time I advanced, it fluttered forward a short distance—
as if pacing me.

We held that rhythm for nearly a kilometer.
It felt uncanny. Magical.

My breath synced to its flight. My legs lightened.
That brutal incline became, paradoxically, the easiest stretch.

Near the crest, it soared to a cliff edge—and paused, turning back to watch.

In that moment, running didn’t feel like exertion.
It felt like dialogue—with nature, with effort, with self.

Life’s steep climbs don’t come with permanent guides.
The real engine isn’t external validation—it’s the quiet certainty within,
and the unshakable pull toward freedom.