apey.co
BlogSandboxAbout中文

How I Built apey.co with AI

Zero Manual Execution: I Built a Full Website Entirely With AI

I recently completed a landmark end-to-end AI practice. In just three hours, I built and officially launched the bilingual blog apey.co entirely through AI collaboration. Throughout the entire project, I wrote no code and performed no manual deployment. AI independently delivered the full technical implementation, including site initialization, functional development, and online launch.

The total computational cost of this project was merely 5.85 RMB. Deployed on the mimo-v2.5-pro lightweight plan (39 RMB per month), the project consumed only 15% of the monthly quota with over 1,000 AI API calls. Aside from the one-time 100 USD domain name fee, the project achieved near-zero technical expenditure. This combination of extreme speed and minimal cost validates a new paradigm for human-AI collaboration: AI takes full ownership of all executable work, while humans focus on core positioning, value trade-offs, and directional decision-making. Human cognition and judgment constitute an irreplicable competitive advantage that AI cannot surpass.

This practice yields a disruptive core insight: technological competition in the AI era no longer centers on human coding proficiency or operational speed. Instead, it hinges entirely on human cognitive depth, strategic discernment, and trade-off wisdom. AI can efficiently handle all tedious, repetitive, and implementation-focused tasks, yet a project's core positioning, spiritual temperament, and long-term iteration roadmap can only be defined by humans.

The New Standard of Tech Selection: Optimized for AI, Not for Humans

During this development cycle, I completely abandoned the traditional human-centered technical selection logic and established a brand-new set of AI-era principles: Technical decisions no longer prioritize human developer experience. The sole benchmark is alignment with AI workflows and the maximization of AI generation accuracy and stability. This cognitive upgrade is essential for every modern technical practitioner.

Traditionally, programming language selection prioritized human usability, favoring intuitive syntax, comprehensive documentation, and active community support. Today, AI has become the primary driver of code generation, triggering a fundamental shift in human roles: from hands-on executors to requirement definers, problem calibrators, and final decision-makers. The underlying logic of technical selection has been thoroughly overturned.

When evaluated by AI generation accuracy, operational stability, and scenario compatibility, the hierarchical gap among mainstream programming languages becomes distinctly clear:

  • TypeScript / JavaScript delivers the highest AI compatibility and output precision by a significant margin. Powered by massive open-source training datasets on GitHub, AI produces exceptionally mature and stable code for mainstream ecosystems including React, Tailwind CSS, and Node.js. As the only technical stack that fully covers both browser frontend and terminal CLI scenarios, it dominates comprehensive adaptability.
  • Python ranks second overall. It serves as the de facto standard for data science and AI toolchains with abundant training corpus. Its core limitation lies in the lack of browser-side support, alongside a less robust terminal ecosystem compared with JavaScript, restricting its application to backend services and data analysis.
  • Go offers moderate overall performance, with its greatest strength being ultra-concise syntax. Minimal grammatical complexity effectively reduces AI coding errors, making it ideal for system-level development, though it likewise excludes browser-side scenarios.
  • Rust exhibits relatively low AI compatibility. Limited public training data and rigid, intricate syntax frequently result in uncompilable AI-generated code, imposing substantial manual debugging costs. It is only viable for niche scenarios requiring extreme performance and rigorous security constraints.
  • Java / C# occupy a unique position. Despite massive legacy code reserves, most available training resources are outdated, leading to conservative, highly redundant AI-generated outputs. Verbose syntax was negligible in manual development workflows, yet under AI's pay-as-you-go computational model, every redundant character incurs tangible overhead, turning traditional syntactic complexity into a critical disadvantage.

TypeScript's prevailing advantage stems not from superior native design or performance, but from its unparalleled data scale and full-scenario coverage, which deliver the most stable and accurate AI iteration results. Traditional metrics such as syntactic aesthetics, ecosystem scale, and talent recruitment costs have become secondary. AI compatibility and training data volume are now the sole decisive factors in technical selection.

I extended this AI-first logic to platform selection. While Vercel offers a more polished traditional developer experience, I ultimately opted for Cloudflare for its superior alignment with AI-driven personal project workflows. It unifies pages, serverless functions, storage, and analytics within a single console, eliminating the need for cross-platform service integration. Core functionalities remain permanently free with transparent tiered pricing, accommodating the full growth cycle from individual projects to small-scale commercial deployments. Its native terminal tooling integrates seamlessly with AI workflows, drastically reducing manual intervention.

The definitive principle of AI-era technical decision-making is clear: prioritize maximum compatibility with AI-driven workflows over theoretical optimal specifications on paper.

The Ultimate Form of AI Products: Terminals for AI, Browsers for Humans

This end-to-end site development practice solidified my core judgment: all future AI product forms will eventually converge into two fundamental scenarios — browsers and terminal command lines.

Browsers serve as human-centric visual interaction portals, tailored to intuitive interface operations and content browsing behaviors. Terminals, by contrast, are lightweight, high-efficiency, and free of redundant visual layers, constituting the optimal operating environment for autonomous AI agents. When AI agents manage devices and execute batch tasks, terminal command operations consistently outperform graphical web interactions in efficiency.

TypeScript uniquely bridges these two core scenarios, supporting both browser frontend development and terminal CLI tool building within a unified ecosystem. This dual-scenario capability is not a result of deliberate forward-looking design, but a historical byproduct of Node.js's iterative evolution, forming its irreplicable competitive barrier.

This yields a pivotal industry conclusion: future human-AI collaboration will evolve toward minimalist text and voice interaction, gradually devaluing traditional complex UI and UX design experience. Terminals will emerge as AI's exclusive working environment, while browsers remain the dedicated interaction interface for humans.

From One Logo: Understanding the Essence of Human-AI Collaboration

The full process of site construction and brand visual design enabled me to fully clarify the clear boundaries and inherent essence of human-AI collaboration.

I initially generated five distinct Logo concepts via AI but rejected all iterations. The cartoon-style ape appeared immature; the linear hand-drawn version looked dated and rigid; the oracle bone "ape" character felt stiff and lifeless; the final option suffered from tonal misalignment despite having no explicit technical flaws. Objectively, all proposals featured precise proportions, balanced white space, and standardized geometric structures — AI's technical execution was impeccable. Yet none captured the spiritual temperament matching my personal ethos and the website's core positioning.

Accordingly, I personally defined the core creative imagery: the oracle bone character meaning "explore", visually depicting a hand reaching into a hidden cave, symbolizing the pursuit of the unknown, inward inquiry, and continuous breakthroughs. I delivered this definitive creative direction and aesthetic standard to AI, supervised the overall tonal iteration, and refined multiple rounds of details to finalize the minimalist visual solution with a dark background and clean white lines.

This Logo carries a layered dual metaphor: reaching into darkness embodies active exploration of the unknown, while light penetrating the cave signifies cognitive illumination through self-reflection and precipitation. This dimensional artistic conception was an unexpected gain from human-guided AI iteration, surpassing the initial creative imagination of both myself and AI.

This lays bare the ultimate essence of human-AI collaboration: AI possesses an execution ceiling far beyond human capability, yet it only executes rather than defines. Humans anchor directions, define values, and control aesthetics, while AI undertakes all refined, repetitive, and granular implementation work.

Knowledge and Action: AI Is Closing the Gap Between Cognition and Practice

While refining the website's core positioning, I derived a renewed, practical interpretation of the ancient philosophy "unity of knowledge and action" tailored for the AI era.

I positioned the site as a carrier for precipitating personal cognition and independent thinking, which AI concisely condensed into the timeless philosophy: Unity of Knowledge and Action. My original translation "To Know Is To Act" appeared rigid and static. The revised version "To Know To Act" fully elevates the conception, transforming a passive assertion into a dynamic closed loop: knowing motivates action, and action iterates knowing, enabling continuous self-improvement through cyclic refinement.

Wang Yangming's core wisdom remains profound: knowledge without practice is not true knowledge. Theoretical cognition without practical implementation equals incomplete mastery. Traditional AI was inherently constrained by this divide: it could output knowledge and answer inquiries but failed to deliver tangible implementation results, leaving knowledge and action fundamentally disconnected.

Modern AI has completely bridged this historic gap. It can not only sort out logical cognition, but also independently complete coding, deployment, domain verification, and full-stack project delivery. The age-old chasm between theory and practice has been eliminated, endowing this five-century-old philosophy with brand-new practical connotations in the AI era.

Human-AI Collaboration: Peeling an Onion — AI Deepens, Humans Define the Core

The full human-AI collaborative workflow resembles peeling an onion. Starting from vague preliminary ideas, I leveraged AI to diverge multiple implementation solutions. I screened and eliminated invalid directions, retained valuable insights, and conducted in-depth polishing through iterative interaction. Gradually, ambiguous and scattered thoughts evolved into clear, rigorous, and actionable outcomes.

This efficient collaborative paradigm adheres to an eternal core premise: AI cannot replace humans in goal definition, value screening, and tonal control. It merely accelerates the implementation and iteration of human ideas. Core trade-offs, tonal positioning, and value judgment ultimately require human confirmation.

Throughout this project, AI undertook all cumbersome technical execution: project initialization, component development, responsive layout adaptation, multilingual configuration, SEO optimization, and full online deployment. AI occasionally generated over-encapsulated logic or incorrect resource paths, yet all deviations could be instantly corrected with precise human guidance and iterative calibration.

In contrast, my work was concentrated on three irreplaceable core responsibilities: defining precise requirements, correcting execution deviations, and delivering final decisions.

Domain Selection: AI Handles Execution, Humans Define Temperament

My online alias "Ape" encapsulates my core perception of AI evolution and underpins my domain name selection logic.

Human civilization originated from two core capabilities: language that fostered abstract thinking, and tool usage that liberated manual productivity. AI's evolutionary trajectory perfectly mirrors this pattern. Large language models build cognitive systems through linguistic comprehension, while tool invocation frees humans from repetitive labor. Ancient humans evolved from apes and created civilization via language and tools; modern AI sparks a new technological revolution through these identical foundations.

Guided by this logic, I confirmed the domain name must incorporate the root "ape". I utilized AI to batch retrieve, compare, and filter numerous root combinations, spelling variations, and suffixes. After eliminating options such as apehub.io, I finally locked onto apey.co. AI completed all mechanical retrieval and comparison work efficiently, yet it cannot assess a domain's brand temperament, tonal fit, and long-term value.

I ultimately selected apey.co for its conciseness, memorability, and distinct brand recognition. The .co suffix delivers a mild commercial texture — sophisticated and restrained, without excessive commercialization. AI can automate all trivial execution tasks, but the ultimate judgment of temperament, tone, and value resides exclusively in human hands.

In the AI Era: What Makes Humans Irreplaceable

This zero-code, zero-manual-operation development practice helped me resolve a core industry proposition: when coding, deployment, design, and optimization — all execution-level technical work — can be fully automated by AI, what constitutes the irreplaceable value of technical practitioners and creators?

The answer lies in three uniquely human capabilities that AI can never replicate.

First, Judgment: rational trade-off screening and precise definition of project value and development boundaries. Second, Taste: subtle aesthetic perception and consistent control of overall tonal quality. Third, Direction: forward-looking iteration anchoring and long-term strategic planning.

These capabilities all stem from exclusive human subjective cognition, value judgment, and independent thinking, carrying unique human perspectives and ideological depth. AI is efficient, accurate, and mass-producible, yet it lacks self-awareness, aesthetic initiative, and value pursuit — it merely executes human instructions mechanically and faithfully.

I will continue to iterate and polish this website with no commercial incentives. It serves purely as a medium for cognitive sorting and ideological precipitation through continuous human-AI collaboration. AI undertakes coding and implementation work, while I fully define and control the project's development direction, trade-off criteria, aesthetic tone, and ideological core.

All execution-level work can be outsourced to AI, yet human thinking, aesthetics, and judgment remain eternally irreplaceable.

Project Cost Breakdown

  • Model: mimo-v2.5-pro
  • Plan: Lite Version, 39 RMB/month, 4.1 billion monthly credits
  • Resource Consumption: 615 million tokens (15% of monthly quota)
  • Computing Cost: 5.85 RMB
  • Request Count: 1000+ times
  • Time Cost: 3 hours
  • Extra Expenses: 100 USD for domain name (the only self-funded item)