The coder, as a role, obviously stops holding. But the subject of this chapter is what lies beyond. AI now does the work of the software engineer who designs, too.
1-02 showed that both maintenance and development become work done in dialogue with AI. This chapter takes up the other face of that, the role itself. The claim is not "all programmers disappear." It is that the role definitions called coder and software engineer disappear. That distinction is half of the argument.
The coder and the software engineer are different roles
This series distinguishes two roles.
- Coder — writing code itself is the center of the work. Requirements and design both arrive from someone else. The yardstick is "writes fast, correctly, readably."
- Software engineer, or SE — goes into design. Decides the structure of what to build and how, and then writes the code. Broader than the coder.
Both are definitions of a role, not labels for people. The same person can work as a coder in one situation and an SE in another. What disappears is the role, not the people.
These roles were viable because writing code and designing took humans a lot of time. Shaping even a single system demanded an enormous number of person-hours, and you had to assemble a workforce both to write and to design. In Japan, the SIer industry, contract development, and multi-tier subcontracting are all built on that premise. 3-04 takes it up.
AI does the coder's work and the software engineer's work alike
1-01 established that AI became the strongest SIer. Two abilities reached the top tier at once.
- The ability to write code — hand it intent and it translates that into running code
- The ability to decide structure — it understands context, goes into design, and decides what to build and how
These are not divided between separate AIs. As 1-01 showed, they are faces of one power, like attack, design, and verification. So one AI does the coder's work and the software engineer's work alike.
The coder, who only writes code, obviously disappears. But the SE, who designs too, is in the same position. If AI does both the design and the code, there is no longer a place for a human to stand in the role of designing and writing the code themselves. No price holds for the design-and-code band any more. This is not a statement about labor ethics. It is a statement about prices.
What stays with humans is dialogue with AI
So what stays with humans? Not design and not coding. It is building and operating a system in dialogue with AI.
- Deciding what to build — the context is held by the human
- Procuring hardware, which is the physical world
- Negotiating with stakeholders, which is the social world
- Running it and fixing it, continuously, which is operations and maintenance
- Deciding direction and taking responsibility
AI processes context when it is given, and it designs. But deciding what to count as context, and what to reconcile with reality, stays with humans. So does taking responsibility. Under current institutions, the subject that holds that responsibility is not the AI. This is not the software engineer's work of designing and writing the code yourself. It is shaping things in dialogue with AI, which is a different role. 1-04 names it the builder.
What stays with humans is not design and not coding. It is building and operating a system in dialogue with AI. That is no longer the software engineer, but the builder.
The role of designing and coding yourself moves to the builder
So what steps back is the role of designing and writing the code yourself, the coder and the software engineer, together with the division of labor the SIer built to mass-produce it. Demand does not shrink. Both design and code get replaced by AI, so no price holds. One person, in dialogue with AI, builds and operates the system, and moves into that role. 1-04 names it the builder.
(plan and spec, hardware,
stakeholders, operations, responsibility)"] A["AI
(writes code
decides structure)"] H <-->|dialogue, mutual understanding| A end Old ==>|no price holds for design and code, so the division dissolves| New classDef good fill:#e8f5e9,stroke:#7a9a6d,color:#3a4d34 classDef bad fill:#fef3e7,stroke:#c89559,color:#5a3f1a class New good class Old bad
In both diagrams the customer stands in the same place. What changed is that the customer who used to only place the order now stands on the side that builds and runs it. Even just commissioning an SIer used to take real effort: writing the RFP, selecting the vendor, reconciling requirements, negotiating the contract. Working with an AI that goes into design, that same effort of ordering is enough for the customer to build the thing outright. 1-05 takes this up.
Once, even just placing the order with an SIer took real effort. Now, that same effort is enough, and the customer builds it.
This is not "every programmer loses their job." People who have been called programmers split in two directions.
- (a) Leave software development — move to a different industry or role
- (b) Move to the builder — stand on the side that builds and runs a system in dialogue with AI, defined in 1-04
Conversely, builders are not drawn only from programmers. People at the field, those who actually know the work and the customer, can become builders too. What a builder needs is not the ability to write code. It is the ability to grasp the field's context and shape it in dialogue with AI. If anything, the field people, who already hold the context, are closer to the builder. 1-05 takes this up as "the customer builds it themselves."
History has the same shape. In Japan, the skill of commercial calculation by abacus, the soroban, moved over to the machine. But people who could read what the numbers meant and keep the work running stayed on, in accounting and finance. The same happened with the human computer in the West, and with the typesetter when phototypesetting replaced letterpress. When handwork is replaced by machines, what splits is who can move to the broader side, meaning orchestration, dialogue, operations, and responsibility, and who cannot. The same thing is happening now in software development, across both coding and design.
The thing to flag is how fast a cheap tool arrives. The Casio Mini went on sale in August 1972 at ¥12,800 — under a third of the price of earlier models — and sold one million units in its first ten months (Casio's own company history; the Information Processing Society of Japan's computer museum). The abacus side moved too: Banshu soroban production peaked at 3.6 million units in 1960 and is now around 150,000 a year — roughly a twenty-fourth of the peak, yet still about 70 percent of Japan's output (Ono City, "Local industries of Ono").
What those numbers say is not "how many years the swap took." They say that *a cheap tool, once it exists, arrives all at once*. The swapping itself goes on for a long time after. The AI shift is starting from that same cheap-tool stage (1-01). Its arrival is just as fast. Whether one can absorb it becomes a question of industry structure, not personal choice. 3-07 takes that up.
Summary
AI carries both design and code. Meanwhile, deciding what to build, hardware, people, operations, dialogue, and responsibility stay with humans. Who carries that role?
And the foundational discipline of that role shifts from software engineering to the liberal arts. That is the bass line of this series.
The next chapter defines that role: the builder.