Back to posts

Mindset Upgrade for the AI Era: From Writing Code to Defining Problems

5-minute read
AIMindsetCareer
Edit

Core Insight: Where Human Value Lies

Human value lies not in computation, but in asking questions, setting goals, and making ethical judgments.

Abandoning a programming career now because of large language models is like abandoning woodworking because of the table saw. The tool changed, but the craftsperson’s value did not disappear – what vanished is pure manual labor, and what remains is aesthetics, design, and judgment.

Those with clearer and broader product awareness, architectural design thinking, and business logic – people who think clearly, define boundaries, and have precise expectations for outcomes – are the ones who can drive greater AI capabilities.

The Changing Role of Engineers

From “Writing Code” to “Defining Problems”

When the cost of code generation approaches zero, the real scarcity is “asking good questions.” Engineers need to shift from “writing code” to “defining problems and guiding AI,” becoming AI “shepherds.”

Technology Implementation Methodology

  • Start from business pain points rather than blindly chasing tech trends
  • Emphasize multi-dimensional evaluation and rapid experimentation; establish a “small steps, continuous iteration” engineering paradigm
  • High-quality technical proposals require: clear principles, explicit trade-offs, reusability

Team Innovation Mechanisms

Oppose “distribute materials” style learning; advocate “learning through play” to spark innovation – for example, exploring technology potential by experiencing non-tech products like lobster or bicycles.

Organizational Support Initiatives (referencing a leading internet company)

  • Launched an internal AI tools platform to support employees using cutting-edge AI tools
  • Provided monthly AI tool reimbursement budgets for engineers
  • Pushed R&D infrastructure toward AI Agent-friendly transformation
  • Introduced new roles like “Agent Engineer,” reshaping the R&D talent system

Career Strategy in the AI Era

Taking Shortcuts the Right Way

I want to take shortcuts to understand all the underlying fundamentals.

Method: Join the best team, stay extremely agile, and relentlessly seek feedback. If you get likes, it means you have helped many people – that is the standard, you have made a substantive contribution to the field.

Building Personal Influence

Cast a wide net and showcase what you have built. Attend every event possible and connect with as many people as you can. When attending events, ask people to introduce you, and make sure they do it on the spot. When someone asks “Should I make an introduction?” your answer is: “Yes, let’s do it right now!”

When talking to people you think are relevant and can help you, show them your value. For example: “I built a high-performance open-source component – you might want to check it out.”

Where AI Is Advancing Fastest

The three fields where AI is advancing fastest right now: content creation, mathematics, and coding. These three share a common trait – output is easy to verify. AI will replace energy as the most important commodity.

Industry Trend Observations

Taking an AI coding product incubated internally at a leading internet company as an example, its goal is to cover the entire software development lifecycle, not just assist with writing code snippets.

Core capabilities: from “writing code” to “autonomous programming”

  • Implicit Knowledge Externalization (Rawi): AI automatically generates readable documentation from code, helping both humans and AI quickly understand engineering architecture
  • Spec-Driven Programming (Spec Coding): Through detailed specifications (similar to PRDs) guiding AI to complete tasks end-to-end
  • Web Coding: For non-professional users, supporting “one sentence generates complete functionality”
  • Async Long-Task Execution: AI can run independently for hours, completing large-scale feature modules – achieving “task delegation” rather than “real-time collaboration”
  • Continuous Autonomous Evolution: AI can distill skills from every interaction, becoming more attuned to user habits over time

Industry Impact

  • AI coding will disrupt traditional SaaS: simple tools are replaced by “generate on demand, discard after use” desktop agents
  • Future software services need to be Agent-Native: API-first rather than GUI, so AI can call them efficiently
  • High-value SaaS will focus on complex systems with data/scenario moats, but development cycles will shrink from “ten years” to “months”
  • AI coding is not a tool upgrade but a thinking revolution: from “humans writing code” to “humans directing AI to complete system-level tasks”

E-commerce Seller Advice

Sellers should not rush to chase sales volume and reviews. Instead, focus on product copywriting and visual optimization, use low-cost advertising to gradually boost sales, and avoid wasting resources and making poor decisions.

Learning Methodology

The Feynman Technique

Explain the concept as if you were explaining it to a 12-year-old.

This is like the Feynman Technique – the best way to learn is to teach it to someone else.

Chase the Aha Moments

Chase these “aha” moments and make them happen as frequently as possible – that is your utility function.

Principles for Interacting with AI

Be extremely direct and specific. Always show me all intermediate states and generated code structure. Make sure I can intuitively understand why it works this way. If you are confident, show me different approaches – what others have tried, why this one works, why that one doesn’t.

Useful Prompt Snippets

  • “Hello AI, here is my entire codebase. Tell me 10 ideas for how I can improve it.”
  • “List the distinctive methods used in this paper. Compare them with previous techniques. Be extremely specific about what they did differently.”
  • “Please rewrite this article in a Hemingway style.”
  • “Rewrite it in the style of Stephen King’s On Writing.”
  • “Don’t rush to answer my question yet. What additional information do I need to provide for a higher-quality answer?”

Documentaries

Books

  • VC: An American History (originally: The History of Venture Capital)
  • World Order (originally: Patterns of the World)
  • A Brief History of Intelligence
  • The First Eye

Other

References