Lawyer Lin Shanglun argued in a commentary that the people best positioned to design professional software may no longer be those sitting furthest inside the traditional development pipeline, but frontline domain experts who can now build with AI. He made that case after an exchange with the Judicial Reform Foundation, where the foundation’s deputy executive director, Attorney Lu, presented a self-built lawyer disciplinary records search system created through vibe coding rather than through a conventional outsourced project costing several million New Taiwan dollars.
Lin wrote that the Judicial Reform Foundation team visited his law firm for an exchange, where Chairman Huang personally presented a certificate of appreciation. What struck him most, he said, was the team’s recent digital R&D work, especially the disciplinary search system built by Lu. In Lin’s description, the product’s search logic, case classification, and practical guidance were hard to fault and closely matched the real research habits of practicing lawyers.
The problem, he wrote, is not budget but translation loss in outsourced development
Lin tied that example to a broader complaint familiar to legal practitioners in Taiwan: searching administrative directives issued by government ministries. He named the Financial Supervisory Commission, the Ministry of Economic Affairs, and the Ministry of Labor as examples of agencies whose systems can feel outdated in design and counterintuitive in use. He described one especially absurd scenario: even when a user already has an exact document reference number and pastes the full string into the search bar, the system still returns no result.
In his telling, these systems did not become difficult to use because government agencies lacked money. The deeper issue is the communication gap built into the standard outsourcing model. A project starts with an agency, goes through tendering, is awarded to an IT vendor, and is then assigned to engineers who may have no understanding of legal practice. In between, project details are relayed again and again through PM meetings. Each handoff strips away part of the frontline user’s intuition and key requirements. By the time the product is delivered, the functions may appear complete on paper while the actual user experience is full of holes.
Lin said this long-running mismatch — technical teams lacking practical legal knowledge, while practitioners historically lacked the ability to write code themselves — has become a core reason public digital infrastructure remains stuck.
As coding gets easier, domain knowledge becomes the harder thing to replace
Lin argued that Lu’s system outperformed traditional outsourced products because it removed that divide. He described software as the digital embodiment of professional knowledge. For years, software companies have tried to recruit strong product managers to bridge business needs and engineering execution. But in practice, he wrote, the best PMs do not appear out of nowhere.
The Judicial Reform Foundation case, in his view, shows a different model. The strongest product managers may be people like Attorney Lu: professionals with years of hands-on experience, a detailed grasp of every operational step and every practical pain point, and the ability to use vibe coding. Once those experts can instruct AI directly and generate tools without losing detail in translation, the resulting systems can be far more mature and much better aligned with real needs than products built by outside contractors.
That shift also changes where the moat sits. Lin’s point was that writing code is no longer the scarce skill it used to be. The harder-to-copy asset is domain expertise — a close understanding of workflows, classification logic, and the specific ways professionals actually use information.
From a nonprofit example to a public-sector wish list
Lin then extended the discussion beyond one legal tool. If a nonprofit organization with limited resources can have core legal experts lead the development of an effective professional search system, he wrote, then regulators and ministries with much larger budgets have little reason to remain tied to older and less efficient ways of building public systems.
He pointed to Singapore as an example, saying its judicial system and government departments have already moved deeply into adopting AI, embedding algorithms and automation into actual operations. Taiwan, by contrast, still faces what he described as an "information island" problem in public infrastructure involving institutional costs and cross-disciplinary rules, with ministries operating separately and databases fragmented.
Lin ended with a specific wish. He said he hopes the Judicial Reform Foundation can use modern AI semantic search and data integration tools to connect directives across agencies, including the Financial Supervisory Commission, the Ministry of Economic Affairs, and the Ministry of Labor, and build a one-stop authoritative search platform shaped by practical legal logic.
Two questions he answered directly
Lin’s article closed by restating two central arguments in Q&A form. First, why are government directive search systems so difficult to use? His answer: the issue is not funding but communication breakdown in outsourced development, where user needs are diluted after repeated PM handoffs. Second, once the barrier to coding falls, who is best suited to build professional software? Using the Judicial Reform Foundation as his example, Lin said the advantage belongs to frontline experts who know the work in detail and can also use vibe coding to turn that knowledge into tools.
The article was first published by BlockTempo.