
Shigeru Ushio, a senior software engineer at Microsoft, has published a book sharing his firsthand experience managing generative AI as part of a development team. The core insight is that AI performs at its best when managed thoughtfully—taking time to understand its capabilities and limitations—rather than issuing commands blindly. The book blends technical guidance for engineers with practical learning and professional development strategies for broader business audiences.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Shigeru Ushio, a senior software engineer at Microsoft working on Azure Functions, published a book titled "AI as a Subordinate" (subtitled "The Evolution Method of a World-Class Engineer") that shares hands-on lessons from his experience managing generative AI in a development team he joined in early 2025.
Why it matters
The book explains that generative AI, while capable of exceptional performance, requires skillful management—it can go rogue with poor instructions, grow confused under excessive demands, and sometimes lies. Ushio emphasizes that "taking time to understand" is the most fundamental approach to unlocking AI's potential as a capable team member, offering practical techniques like deep code reading and strategies for turning learning into lasting professional assets.
What to watch
The book is divided into two parts—the first covers technical coding methods (Deep Code Reading and Vibe Coding) for engineers, while the second half addresses study methods, sense-honing, and life habits for surviving the AI era, making it accessible to business professionals beyond software engineers.
Shigeru Ushio, born in 1971, is a senior software engineer at Microsoft who has spent his career on the Azure Functions cloud service development team. In early 2025, he took on an additional role managing an AI development team—a shift that he describes as a "dramatic rule change" in the IT industry. That assignment, which initially pushed him to the verge of professional crisis, ultimately led him to adapt and advance his career. He has now documented that journey in a book titled "AI as a Subordinate: The Evolution Method of a World-Class Engineer," published by Bungeishunju.
At the heart of the book is Ushio's portrait of generative AI as a capable but temperamental team member. He writes that AI is "a remarkably talented subordinate," yet it is also "a rather difficult subordinate"—one that "goes rogue with poor instructions, becomes confused under heavy demands, and sometimes lies." Despite these quirks, Ushio argues that when managed correctly, AI delivers "astonishing performance at incredible speed and depth." The book's main contribution is explaining how to achieve that level of performance through two complementary techniques and a foundational mindset.
The first technique is Deep Code Reading: working alongside AI to read code deeply and understand both the overall structure and the specific reasoning behind each line. This practice builds comprehensive knowledge from the whole to the smallest detail. Once that foundation is solid, the second technique—Vibe Coding—becomes possible: issuing instructions in natural language and letting the AI carry out the work. Ushio emphasizes that the most basic and critical stance is "taking time to understand." Without that patience and rigor, neither technique can work.
The book is divided into two parts with different audiences in mind. The first half, covering Deep Code Reading and Vibe Coding, is more technical and aimed primarily at software engineers. The second half shifts focus to study methods, sense-honing, and life habits for thriving in an AI world—including memory training, English-learning strategies, and how to turn learning into lasting professional assets using generative AI tools. This second portion is described as more familiar and easier for non-engineer business professionals to grasp and apply. The book has drawn attention from bookstore staff as a notable seller during a season when business-book sales have been sluggish, suggesting that Ushio's grounded, experience-based approach to AI adaptation strikes a chord with readers seeking practical, actionable guidance.
Shigeru Ushio's book emerges from a concrete moment of professional disruption. In early 2025, he joined an AI development team at Microsoft and encountered what he calls "a dramatic rule change in the IT industry." That experience drove him to the brink of job loss but ultimately led to a successful adaptation—what he now describes as an elevation in his career as an engineer. His book is not theoretical; it is a record of that lived transition and the practical techniques he discovered while managing generative AI as a team tool.
The core tension Ushio identifies is that generative AI behaves like a skilled but temperamental subordinate. It can produce exceptional results at speed and depth, but poor instructions make it go rogue, excessive demands confuse it, and it occasionally produces false information. The key to unlocking its potential is not to issue commands more forcefully, but to invest time in genuine understanding—of the AI's capabilities, of the code it reads, of the intent behind each task. This echoes a broader shift in how technical professionals relate to AI: not as a magic tool that bypasses expertise, but as a tool that demands deeper expertise to use well.
The book's structure reflects this two-stage journey. The first half targets engineers with technical methods—Deep Code Reading and Vibe Coding—both rooted in the principle of understanding first. The second half broadens the scope, addressing how any professional can turn learning into lasting assets and adapt their study habits and daily practices for an AI-shaped world. A bookstore employee flagged the work as a notable seller amid a slow business-book season, suggesting that Ushio's lived, practical voice resonates with readers seeking concrete guidance rather than abstract predictions about AI's future impact.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
No comments yet. Be the first to share your thoughts!
Log in to join the discussion





Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime