AI & Enterprise
As artificial intelligence (AI) coding tools spread and lower barriers to software development, a claim is drawing attention that developers’ core competitiveness is shifting from programming ability to knowledge that deeply understands specific industries and tasks.
Gigazine reported on June 1 that software developer Aaron Brethorst (아론 브레트호르스트) said that the more AI can generate code quickly, the more important it is not “writing code itself” but having expertise that can understand “coding structure.”
For example, in a payroll calculation system it is more important to accurately grasp real rules such as tax rates, deduction conditions and adjustments by pay period than to implement calculation logic. In that context, the claim is that the standard for judging whether a system works properly depends not on programming syntax but on on-the-ground knowledge.
He explained that when using the same AI coding tool, a logistics dispatch planner with 15 years of experience and an excellent software engineer show different strengths. Even if the dispatch planner cannot write code directly, the planner can determine whether the AI-created logistics system fits on-the-ground requirements. By contrast, the engineer can assess code quality but could miss whether the system meets actual operational requirements.
He also said that as agent-type AI makes it possible to create software without directly building an operating model, the long-assumed link in development between domain expertise and code has weakened.
This change is also reshaping the role structure between developers and domain experts. In the past, engineers repeatedly collaborated with domain experts and refined systems through trial and error in operating environments. Domain experts had task knowledge but found it difficult to secure the capability to directly build reliable software. But as AI lowers the cost of turning ideas into working software, the value of domain knowledge is being highlighted more than engineers’ implementation skills, the report said.
An Anthropic hackathon was cited as an example supporting this claim. The event, which was a competition in using the latest AI models, drew 500 participants, most of whom were developers, but 3 of the 5 winners had no experience launching software. Systems researcher Dexter Hadley (덱스터 하들리) said the results were a case in which domain expertise outpaced coding ability.
Brethorst said areas where experienced engineers should invest time going forward include deep understanding of real industries and work processes, specialized equipment and regulatory systems. He said the value of the skill of implementing with clean code has declined, while knowledge that deeply understands real work and is verified in practice remains scarce.
Still, there was also a counterargument that domain experts do not immediately succeed in software development. On Hacker News, some said the ability to verify whether a system’s output is correct differs from the ability to instruct AI to produce correct output in the first place. The comment said that even experts in a specific field may struggle to clearly organize rules learned through long experience into tests and requirements that AI can understand.
While AI is changing how software is produced, the point that responsibility for judging what results are correct still depends on human industry knowledge is emerging as the core of the discussion.
This content was produced with the assistance of AI and reviewed by our editorial team. You can read the original version in Korean here.
Leave a Reply