Which of the Three Emerging Programmer Tiers Do You Belong To?

Three emerging programmer tiers

Tier 1 – AI system designers are a very small group (about 1 % of developers). They do not write business‑level code; instead they design AI systems, train large models, and build foundational infrastructure such as operating systems, databases, compilers and browsers. The article notes that AI performance on system‑software tasks is still extremely poor, so these components remain the domain of expert engineers whose work is used by virtually all programmers.

Tier 2 – AI “shepherds” (mainstream programmers) will become the numerically dominant layer. Their daily work shifts from typing code line‑by‑line to formulating clear problem statements and orchestrating multiple AI agents. The author illustrates a typical workflow: a programmer sits with a laptop, issues high‑level commands to ten AI agents, and simultaneously acts as project manager, system architect and ultimate accountability holder.

This is an excerpt of a high‑concurrency e‑commerce system request that an AI would receive: “This is an e‑commerce system with user management, product management, inventory, order processing, etc., spanning many modules. Technically it must be distributed, use caching, master‑slave replication, sharding, message queues, service decomposition, rate‑limiting, circuit‑breaking, search, flash‑sale handling, and so on. <8000 words omitted> ” The programmer must decompose the functional and non‑functional requirements, prioritize and split them into small units, then direct each AI agent to implement a piece. After code generation, the programmer must also command the AI to write tests and run self‑validation, because AI can produce toy prototypes that appear correct but fail under real load.

Even for new projects, legacy codebases pose additional challenges: millions of lines of existing code make any change risky, requiring extensive planning, AI‑assisted code‑base understanding, prioritization, and rigorous code review.

The article cites a statement attributed to the Google CEO that AI raises senior engineers’ productivity by roughly 10 %.

Why the tiered structure emerges

AI reshapes the cost structure of software production. Historically, becoming a developer required mastering a wide set of skills—languages, algorithms, frameworks, design patterns—creating a high entry barrier. The AI revolution reduces the scarcity of “knowing how to code” and makes the ability to express problems clearly and orchestrate AI agents the critical competency.

Research by Erik Brynjolfsson and colleagues (Stanford Digital Economy Lab) finds that programmers are among the occupations most exposed to AI. Since 2022, the number of junior developers aged 22‑25 has fallen 16 %, while senior‑level positions have shown little change.

Tier 3 – AI‑empowered “wild magicians”

This layer consists of non‑technical professionals (doctors, teachers, accountants, designers) who use AI to solve personal or departmental problems. Their tools are typically low‑complexity, built within hours or a day, and are not intended for enterprise‑scale deployment.

Concrete examples:

Maxime Cuisy, a production manager at a Paris printing company, used Codex to create a cross‑platform program that processes up to 2 000 photos per run, solving a manual‑adjustment bottleneck without hiring developers.

A teenager wanted a turn‑based card game with attack, defense, and effect cards, variable ranges, movement, health, and dynamic visual effects. After the author helped with initial technical choices, the teenager, using natural‑language prompts, drove AI to implement each game feature independently.

These cases illustrate a nascent “personal software” era analogous to today’s user‑generated media content.

Implications

Software development as a profession will persist, but the traditional premium on “writing code” diminishes. Senior engineers remain essential for complex, high‑concurrency, or legacy systems that AI cannot handle alone. Junior roles face erosion as companies rely on AI‑augmented senior staff, reducing the incentive to hire new graduates. The decisive factor for future programmers will be the ability to articulate requirements precisely and to coordinate AI agents effectively, rather than raw coding skill.

Code example

往
期
推
荐
1、
现在为啥见不到光驱了？
2、
曾经对程序员最好的公司，倒下了
3、
老牌开源项目该不该接受 AI 代码？
4、
面试官问我：“AI 写代码比你快 100 倍，你的价值在哪？”
5、
国家为什么要把国企等企业电脑全部换成Linux环境？能不能从专业的角度分析一下？