Charles Darwin once described the peculiar species on the Galapagos Islands that evolved in isolation—perfectly adapted to their closed environment yet utterly defenseless against external competition. This is precisely the portrayal of Japan's technology sector over the past two decades. The country can manufacture industrial robots with micrometer-level precision and Toyota cars that never break down, yet in the landscape of the digital economy, it is a barren wasteland: no Google, no Facebook, no TikTok, no WeChat, no Taobao—only isolated islands built by local system integrators (SIers).
For a long time, this phenomenon of being a "hardware giant, software dwarf" has been attributed to the insular nature of Japanese culture. But the real ailment lies deep within the relations of production: in Japan's software supply chain, the "user" is no longer the object of service but merely an abstract symbol in a contract clause.
And now, this metaphor is crossing the ocean and arriving in China.
Although China's consumer internet has already proven its vitality globally, in those more substantial, less visible "deep waters of digital transformation"—from the backend systems of large enterprises to complex industry solutions—a structural convergence is emerging. We are surprised to discover that the path of "high outsourcing, low accumulation," which once trapped Japan's software industry, is quietly being replicated here.
The Vanishing Ghost
In the skyscrapers of Tokyo's Otemachi district, consultants in suits plan the digital future on exquisite PowerPoint slides; meanwhile, in another part of the city, dispatched engineers crammed into cramped cubicles stare at screens full of code. Separating them is a "Requirements Specification Document" hundreds of pages thick.
This is a modern-day "caste system": the upper echelon is responsible for thinking but bears no consequences; the lower echelon is responsible for execution but lacks business understanding. Japan calls this model SIer (system integration), once hailed as the pinnacle of industrial division of labor.
However, this ghost that once haunted Japan is now resurging in our software industry in an even more extreme form: a comprehensive regression from "project outsourcing" to "labor dispatch."
Once upon a time, enterprises purchased software to acquire a "solution"; now, more and more orders are merely to lease "heads." On financial statements, this is whitewashed as "flexible staffing" and "cost reduction and efficiency improvement," but on the engineering front, it means "infinite dilution of responsibility."
When software development is no longer settled based on "delivering results" but billed by "man-days," the contract between developers and end-users is completely torn apart. For dispatched engineers, their client is no longer the businessperson who actually uses the system but the administrative manager who signs their attendance sheets.
This misalignment of incentives leads to a fundamental distortion of software functionality: systems are no longer designed to empower frontline users but to satisfy the "controllability" of administrative management.
To manage this fluid, trust-deficient dispatched workforce, enterprises must establish cumbersome approval processes and traceability mechanisms. Thus, software is filled with reports designed for "upward reporting" and confirmation buttons designed for "liability avoidance."
Genuine user experience is drowned under the logic of administrative control. Here, there is only the consumption of man-hours, no accumulation of value; only the arrogance of management, no dignity for users.
The Tyranny of Mediocrity
Why, in a market with the world's youngest and most creative pool of engineers, do many B2B software tech stacks exhibit a strange premature aging? The answer lies not in technical capability but in transaction costs and the transfer of power.
Under the financial pressure of "cost reduction and efficiency improvement," a quiet purge is underway. Those "technocrats" who pursue excellence but command high salaries are forced to "graduate"; they are replaced by "administrative bureaucrats" skilled in budget control and report embellishment.
For the administrative managers who take over control, technology is no longer a core competency but a vast, uncontrollable black box. They fear the uncertainty brought by innovation and, even more, the political responsibility that comes with project failure. In their logic, outsourcing business to labor dispatch teams is not only the optimal financial choice but also a shrewd form of "risk outsourcing."
This outsourcing contract essentially becomes a firewall of accountability.
If an in-house team's innovation fails, the manager is held accountable; but if an outsourced project encounters problems, "insufficient supplier capability" becomes the perfect and compliant scapegoat. This mechanism of "internalizing gains, externalizing risks" makes mediocrity the safest survival strategy within the system.
At the execution level, this risk-averse logic creates a typical "lemons market": because buyers are only willing to pay for "man-months" (labor hours) and not for "excellence" (results), highly skilled talent is driven out, and system design is forcibly lowered to the "minimum maintainable level."
Thus, technological obsolescence is preserved as a feature—it means that even if the original team is replaced like "consumables," any cheap dispatched worker from a training class can take over at any time.
This forms a perfect closed loop: administrative bureaucrats, for the sake of attractive financial reports and political safety, expel technical elites and introduce a fluid dispatched army; and to accommodate this army, technical standards are actively castrated. Ultimately, enterprises obtain a mediocre but controllable machine but lose a beating heart.
The Eternal Intern
There is an often-overlooked iron law in software engineering: excellent systems are the product of the co-evolution of the team, the code, and the users.
In an ideal model, engineers increasingly understand users through continuous iteration and deepen their skills by filling every "pitfall"; meanwhile, the system, in this process, accumulates business-aligned features like tree rings. This "entanglement of people and systems" is the true value moat of software.
However, the prevalent outsourcing and dispatch models brutally sever this symbiotic bond.
At project sites, personnel rotate like a revolving door. When a group of "newcomers" who know nothing about the business history take over the system, they see only obscure code, not vivid logic.
They don't know that a seemingly redundant field exists to accommodate the most crucial major client; nor do they understand that a peculiar architectural design is meant to handle seasonal traffic surges. Because they lack understanding of users and memory of history, they cannot grasp the system's "why it was done this way (Why)," only mechanically execute "what to do (What)."
This cognitive misalignment leads to the most severe "institutional amnesia" in the industry. Codebases become "big balls of mud" that no one dares to touch; value ceases to grow, leaving only meaningless accumulation.
The cruelest cost of this model is the systematic squandering of human capital.
In Japan, this led to the solidification of the "IT construction worker" class; in China, a similar group is rapidly expanding. Lacking the soil for long-term, deep business engagement, engineers cannot grow through the compound interest of experience. They are forced to wander between different projects, repeating one year of experience for ten years.
This is why we see a bizarre phenomenon: systems are constantly "aging," but the people maintaining them are forever young.
This fracture in the career ladder creates a group of "eternal interns." They have no sense of belonging, no technical accumulation, and, of course, no future.
When the threshold of age 35 arrives, they are seen as "old batteries" with reduced cost-effectiveness and are mercilessly replaced, while the industry gradually hollows out in rounds of low-level repetition.
The Illusion of Compliance
When an organization is filled with fluid, rootless "eternal interns," how can managers ensure that the vast system does not spiral out of control?
The answer can only be replacing trust in people with cold, rigid processes.
To manage this massive "mercenary army," a survival philosophy originating from traditional bureaucratic institutions is transplanted into software engineering: process is more important than results, liability avoidance is more important than excellence.
In Japanese corporate culture, as long as the process is compliant, even catastrophic failures can be forgiven. Managers know well that relying on a group of disloyal outsourced personnel to pursue "excellence" is unrealistic. Therefore, the goal is downgraded to "controllability." Hence, we see a proliferation of "defensive management."
In the requirements phase, documents grow thicker, not to clarify business logic but to prevent contract disputes—documents transform from carriers of knowledge into handover slips of responsibility.
In the development phase, approval chains are stretched indefinitely, not to control code quality but to ensure every step is documented, achieving traceability at every level.
This management model of "killing creativity with process" not only creates rigid systems but also steers enterprises' digital transformation onto a misguided path.
True digitalization should be business-oriented—empowering the frontline through data flow, letting those who hear the gunfire make decisions. But under the dominance of the "illusion of compliance," we see a large number of "pseudo-digitalization" efforts that are management-oriented.
These systems are designed as electronic approval flows, with the value of data confined within reporting hierarchies. They are like "digital concrete slabs"—flat and compliant, but barren.
This model originating from Japan is leading digitalization astray and is incompatible with the AI era.
The proliferation of AI technology can greatly expand an enterprise's service radius, pushing demands toward massive scale, real-time, and personalization. Software iteration cycles will no longer be measured in months or quarters but in weeks or days.
In this "overclocked competition," management-oriented defensive systems appear clumsy and laughable. The outsourcing model relies on lengthy contract negotiations and requirement confirmations; when requirements change weekly, every "change request" to the outsourcer becomes a verdict that delays seizing opportunities.
The essence of AI is "removing management" and empowering individuals. Those systems piled up by outsourced armies and wrapped in layers of defensive processes are inherently anti-AI. They not only cannot nurture the seed of AI, which requires flexible soil, but are accelerating the aging of enterprises.
The Fork in the Road
We are facing a dangerous misalignment: the current system specializes in producing "porters," while the only scarcity in the AI era is "gardeners."
The rise of generative AI is rewriting the rules of software engineering. Those standardized, divisible, low-responsibility-density coding tasks will be taken over by AI at zero cost. What will be truly scarce in the future are architects and product people who can understand complex business and solve problems through "human-machine collaboration" amid uncertainty.
Unfortunately, the existing outsourcing and dispatch systems specialize in producing the former and stifling the latter. You cannot delegate innovation through outsourcing contracts, nor can you cultivate "super individuals" capable of harnessing AI on an assembly line.
This is not an irreversible cultural destiny but the result of long-term selection by wrong incentive mechanisms. The hollowing out of Japan's software industry has proven that when a market only prices "heads" and does not pay for "value," it inevitably degenerates into a Galapagos island.
China spent twenty years forging a highly commercialized internet path, but in the deep waters of digital and AI transformation, if it cannot reverse this "de-skilling" incentive mechanism, the industry will face a Japan-style chronic degeneration.
Will we continue to choose Japan's "safe mediocrity," or dare to embrace "high-risk innovation"?
Regardless of which we choose, undoubtedly, in the ongoing AI race, entry tickets will only be issued to enterprises that have "gardeners," not "porters."