Why Reading Code Is the Missing Skill Every Developer Needs

In recent talks across Bangkok, Bangalore, and Tokyo about "next‑generation software development," I highlighted the evolution from simple assemblers to AI‑powered coding assistants and posed the question: how should we prepare for this new era? My answer: learn to read code.

Back to College Days

In the late 1970s at Montgomery College, I rummaged through trash bins for the IBM mainframe OS build lists, which required re‑assembling source code and resulted in massive stacks of green‑striped paper.

While classmates enjoyed carefree college life, I delved into those lists to understand the operating system and improve my assembly coding.

The Lion Book

In the early 1980s, while working at Contel Information Systems in Bethesda, I acquired a copy of the famous "Lions' Commentary on the UNIX Operating System" (often called the Lion Book) after buying the "C Programming Language" and using Unix System 7 on a PDP‑11.

Spending my spare time studying system initialization, task switching, memory management, and I/O handling deepened my understanding of low‑level principles.

Years later, after reading code extensively and being tasked to run four single‑user operating systems on the same hardware (see the story "IN/MSX: Running Four OSes Simultaneously"), I realized the value of code reading.

Learn to Read Code

Our education system rarely emphasizes this skill; instead, we start from a blank file. Imagine if the first weeks of a CS curriculum focused on classic code, teaching students to grasp architecture, data structures, algorithms, and even code aesthetics such as naming, indentation, and comments.

Courses should also examine project evolution, analyzing changes commit‑by‑commit to build deep understanding.

Rather than studying idealized pseudocode, we should study battle‑tested production code maintained, fixed, and enhanced by developers of varying skill levels.

These reading experiences help students see how real‑world code matures, including its flaws and shortcomings, preparing them to write better code under expert guidance.

Write Code for Readability

Two approaches that immediately come to mind for improving readability are Hungarian notation and readability programming.

Hungarian Notation

Invented by Microsoft early employee Charles Simonyi, Hungarian notation encodes a variable's type or purpose with a prefix—e.g., p for pointer, sz for string, w for word, b for byte. I used it extensively during my 25‑year tenure at Microsoft, though its current relevance is uncertain.

Readability Programming

Proposed by legendary Donald Knuth, readability programming was first described in his 1983 paper and elaborated in his 1992 book "Readable Programming." It treats programs as literary works, urging us to focus on explaining to humans what the computer should do.

Let’s change the traditional attitude toward program construction: instead of imagining our main task is telling the computer what to do, we should concentrate on explaining to humans what we want the computer to do.

Although readability programming offers attractive teaching benefits, it never became mainstream. I wonder whether the detailed natural‑language explanations in readable code could serve as ideal training data for large language models that power today’s coding assistants.

In summary, I hope you spend more time reading code; it will make you a better developer and help you extract more value from your favorite AI coding assistants.