AI News

1024 Programmer's Day: Does AI also celebrate Programmer's Day?

The PaperUpdated Nov 16 19:31

AI summary

On 1024 Programmer's Day 2026, The Paper Business School published an article questioning the value of programmers in the AI era. The article mentioned that over 75% of Google's new code is generated by AI, more than about 80% of Anthropic's newly added code is produced by Claude, and ByteDance's TRAE has generated nearly 100 billion lines of code with users exceeding 6 million. The author believes that code will not come to an end; it has merely changed the way it is written.

Body · AI translation

Download the client

+1 Today is 1024. Does AI also celebrate Programmer's Day? Tao Heshan 2026-10-24 08:16 The Paper Business School > Listen to the full text Font size

"Talk is cheap. Show me the code."

On August 25, 2000, in the Linux kernel mailing list, faced with an optimization proposal that was full of grand talk but never produced any code, Linus Torvalds replied as such.

For more than twenty years since, this sentence has become a golden quote in the programmer world—code is the "passport" recognized by this generation, it is logic, it is craft, and it is also the contract between humans and machines.

Now, when AI begins to effortlessly write tens of thousands of lines of code and thereby improve itself (RSI), this maxim suddenly has an echo: what is being mocked no longer seems to be the empty talkers, but perhaps every character we typed with our own hands.

1024, this Programmer's Day built from binary, is no longer a commemoration, but more like a questioning—if code can be generated, then what is left for the people who write code?

I. The Golden Age: Looking Back, Ten Thousand "Codes" Galloping

In 1946, ENIAC lit up at the University of Pennsylvania. Programs were not written, but "wired"—plugging and unplugging cables, flipping switches, a piece of logic took several days.

In 1951, Grace Hopper wrote humanity's first compiler for UNIVAC I, the world's first commercial general-purpose computer, and for the first time a machine understood human symbols.

In 1968, "software engineering" was formally proposed, and programming began to evolve from a personal craft into an engineering discipline.

A landmark event occurred in Helsinki in 1991. 21-year-old Linus Torvalds posted an operating system kernel he had written on a newsgroup, naming it Linux. Fourteen years later, he wrote Git in an extremely short time. One person, twice rebuilt the foundation of the entire industry.

In the thirty years since, code has become an important force in rewriting the world. From C to Java, from PC to mobile phone, from dial-up Internet to mobile Internet. Behind every button through which people interact with the world stands a programmer who stayed up late at night. They believe "code changes the world"—this is not a slogan, but something that has actually happened on their resumes.

By 2025, there were more than 47 million professional software developers worldwide, and more than 9 million software engineers in China, ranking first in the world. In the same year, China's software and information technology services industry revenue reached 15.48 trillion yuan, equivalent to Germany's total national public fiscal revenue in 2025.

This is a digital empire built brick by brick by human hands.

In that golden age when the Internet + mobile Internet was in full swing, heroes emerged in the programming world, and ten thousand "codes" galloped. People talked about the elegance of architecture and the brilliance of algorithms, betting on whose open-source library could get ten thousand stars.

At that time, code was hand-crafted, bugs were despised, and pride was concrete. No one thought that it was precisely these sparkling works that would later become textbooks for training successors.

II. AI Coding: Top Players, Project Zero

In October 2024, at the third-quarter earnings call, Google CEO Sundar Pichai threw out the sentence that struck a nerve across the entire industry: today more than a quarter of Google's new code is generated by AI. Half a year later, he raised this number to more than 30%. In April 2026, this number jumped to 75%.

Microsoft CTO Kevin Scott predicted that by 2030, 95% of code will be generated by AI. Meta's Zuckerberg bet that before 2027 most code will be completed by AI Agents.

At Anthropic, this data is even more aggressive. As of May 2026, more than 80% of the new code merged into the company's overall main codebase was produced by Claude, and about 95% of the new code in the Claude Code team was generated by AI. In a podcast interview in June 2026, Boris Cherny, head of Claude Code, claimed that he had not written code for 8 months.

In March 2025, ByteDance launched TRAE, the first AI-native IDE in China; on May 30, Tongyi Lingma launched; on June 23, Wenxin Kuaima launched; on September 9, CodeBuddy launched.

By the end of 2025, TRAE had generated nearly 100 billion lines of code for global users, handled 500 million development queries, and accumulated more than 6 million users.

Why are these top major companies relentlessly pursuing AI Coding? Because code is the most important hands and feet of Agentic AI. The degree of freedom in operating code is almost equivalent to the degree of freedom in invoking all tools and resources. When an agent can autonomously set up environments, manufacture tools, and organize tasks, it also grasps the key to autonomous evolution.

AI Coding is exactly that capability inflection point; at the end of the corridor is the RSI (recursive self-improvement) that the industry has long yearned for.

At this moment, code is not just a deliverable representing efficiency; it has become efficiency itself.

III. GitHub: I Rise by This, and by This I Shall Perish

If programmers are their own gravediggers, GitHub is exactly that mass grave.

GitHub has become the world's largest treasure trove of "real process data." A commit record links together the problem description, modification plan, code review, and test results, almost completely preserving the full chain of a problem from appearance to resolution, including "why it was changed this way" and "whether it was ultimately accepted."

These data are not static code, but the complete trajectory of software engineering implementation: what should be done, what should not be done, how to make trade-offs, how to compromise. What AI lacks most is this kind of real experience with context, and the programmers who hand-crafted code spent twenty years freely and voluntarily organizing it into volumes and publishing it publicly.

In November 2022, a group of open-source developers filed a lawsuit against GitHub, Microsoft, and OpenAI, alleging that Copilot used open-source code for training without permission and output copyright-protected snippets. To this day, the litigation is still ongoing. The law remains unsettled, yet technology has long since crossed the boundary—in March 2026, GitHub began by default using Copilot interaction data to train models.

Data does not lie. GitHub's annual Octoverse 2025 report shows: platform developers have exceeded 180 million, with 36 million added in one year; code repositories reached 630 million; nearly 1 billion commits were pushed over the year, with more than 500 million pull requests, and over 1.1 million public repositories integrated large-model SDKs, up 178% year over year. A Sonar survey of more than 1,100 professional developers shows that AI already accounts for 42% of the code they commit, and is projected to exceed 65% by 2027.

Without GitHub, Coding would not be so prosperous; without GitHub, it would not be heading toward its endpoint so rapidly either.

Within this cycle of cause and effect, for other industries not yet conquered by AI, what is missing may not be demand, but a GitHub.

IV. Complexity: To do a good job, one must first sharpen one's tools

In 1975, Fred Brooks wrote The Mythical Man-Month. He had just been through IBM OS/360—the largest software engineering effort of its time, delayed, over budget, and nearly out of control. He set a monument for later generations with one law: adding manpower to a project that is behind schedule only makes it fall further behind.

In 1986, in No Silver Bullet, he further asserted: no single technological or managerial breakthrough could, within a decade, improve software productivity, reliability, or simplicity by an order of magnitude. Because the essential complexity of software cannot be eliminated; what can be eliminated is only accidental complexity.

Half a century later, these words remain sharp. And AI Coding, for the first time, has truly touched the boundary of "essential complexity"—it is not a silver bullet, but it compresses most accidental complexity, such as writing boilerplate, checking syntax, stitching interfaces, and fixing bugs, to nearly zero.

So the question turns to: what should humans do? Brooks already had the answer—conceptual integrity. The soul of a system lies not in how many features it has, but in whether a unified design philosophy runs through it, and this requires centralized rather than dispersed judgment. This is precisely what AI cannot yet provide: it can write functionally correct code, but it is hard for it to decide whether that feature should exist at all.

Classics are classics because what they discuss is not syntax, but thought.

As Agentic AI capabilities continue to leap forward, the tradition of hand-writing code may eventually fall, but programming thought will not end; on the contrary, it will be carried forward further. Architectural trade-offs, levels of abstraction, room left in interfaces, and the measure of naming—these things that belong to developers' taste, like architectural art, endure and renew.

When this industry no longer relies on individual heroism, and with the support of scientific thought and advanced tools, humanity has the opportunity to take on engineering of unprecedented complexity, thereby reaching achievements that human power alone could never attain.

V. The Mind of the Machine: When the water runs out, sit and watch the clouds rise

Returning to the individual level, in the face of this vast industry transformation, should one weep bitterly at a dead end, or look forward to a turn for the better? Better to keep an ordinary mind: walk to where the water ends, sit and watch when the clouds rise.

In 1974, 19-year-old Jobs went barefoot to India. After returning, he shaved his head and wore a long robe, and for nearly three decades thereafter he was close to a Soto Zen monk from Niigata, Japan, Kobun Chino Otogawa, and often meditated with him. In the 1970s, he read Shunryu Suzuki's Zen Mind, Beginner's Mind, and the "beginner's mind" became a lifelong footnote for him. In 1985, when he founded NeXT, he invited Otogawa to serve as spiritual advisor; in 1991, Otogawa presided over his wedding.

Zen gave Jobs two things: extreme simplicity, and intuition. It is said his office was over two hundred square meters, almost empty, with only a cushion in the middle. He deeply believed his intuition, more than market research, and used inner tranquility to apprehend the essence of products. "Not relying on words, directly pointing to the human mind" was translated by him into the design language of the iPhone.

This may be one of the best angles for understanding AI Coding. It is not that humans are replaced by machines, but that human intuition and judgment are amplified as never before. "Humans provide goals, and no longer need to provide methods."

As an Anthropic engineer said, although he had not personally written a line of code for five months, what he did had not decreased; it had merely changed from "typing" to "thinking."

Empty-handed, holding a hoe; walking, riding a water buffalo. Setting down code, we are able to reexamine the relationship between people and code, and thereby with the world.

The heart of one person is the heart of millions; the hearts of millions merge into the machine, and the machine then has a heart.

Ninety years ago, Bian Zhilin left behind the Fragment that makes people and the world scenery for each other; today this mirror reflects a familiar light and shadow—you teach it how to think, and it teaches you what thinking is. You pour out decades of experience to it, and it ferries your abilities to others.

1024 is the machine's scale, and on this special day it meets humanity's scale of time.

Passing through 2026, code will not end; it has merely changed the way it is written.

(The author Tao Heshan is an intelligent economy worker engaged in policy planning in the field of artificial intelligence) Responsible editor: Cai Junjian Image editor: Le Yufeng Proofreader: Yao Yiqi The Paper news tip: 021-962866 The Paper, unauthorized reproduction prohibited +1 Favorite I want to report #TaoHeshan #ProgrammersDay View more

Source

The Paper