Claude Code Gets 2 Likes on Demo, Now Generates 4% of All GitHub Code

Claude Code’s meteoric rise began in the shadows of Anthropic’s engineering team, where a modest terminal‑based prototype earned only two “likes” during an internal demo, according to a March 16 post by Anthropic engineer Aditya Agarwal. Boris Cherny, who built the demo, deliberately chose a command‑line interface because it could be prototyped by a single developer and would remain stable as the underlying model improved, eliminating the need for continual UI redesigns. That pragmatic decision, though initially dismissed as “too weird” for a 2025 audience, proved prescient: by February 2025 the tool was released externally, and its adoption accelerated as Anthropic’s models grew more capable.

Internal usage quickly turned into a broader phenomenon. Anthropic staff began relying on Claude Code for daily development tasks, and daily active users surged, leading to an external launch that, while slow to catch on, eventually “spread like wildfire” as the model’s performance outpaced competing solutions. Wired reports that OpenAI is now scrambling to catch up, underscoring the competitive pressure Claude Code has placed on the market. By March 2026, Anthropic claims the tool is responsible for roughly 4 % of all public GitHub commits—about 135,000 commits per day—and generates an estimated $2.5 billion in annualized run‑rate revenue. The company also reports onboarding one million new users each day, a scale that dwarfs the typical growth curves of most developer tools.

The sheer volume of code produced by Claude Code eclipses the output of many large software firms. If the tool writes 135,000 commits daily, that translates to more than 49 million commits per year, a figure that exceeds the annual code shipments of many Fortune 500 companies. This productivity boost is attributed to the tool’s stream‑oriented terminal interface, which allows developers to pipe in increasingly sophisticated models without reworking the UI. Ars Technica notes that Anthropic has extended the concept with “Cowork,” a general‑computing assistant built on the same principles, suggesting the terminal paradigm may become a broader platform for AI‑augmented workflows.

Analysts see Claude Code’s success as a case study in how minimal viable products can outpace polished competitors when they align with developers’ core workflows. The tool’s rapid adoption illustrates a shift away from heavyweight IDE plugins toward lightweight, model‑driven interfaces that integrate seamlessly into existing command‑line environments. This trend is reinforced by the fact that Anthropic’s growth metrics—one million daily new users and a $2.5 billion run‑rate—are now benchmark figures for AI‑coding assistants, according to the internal report cited by Agarwal.

While OpenAI’s own coding agents remain prominent, the Wired piece highlights that the company’s response has been hampered by a legacy focus on GUI‑centric products, leaving a gap that Claude Code has filled. As Anthropic continues to iterate on its terminal‑first approach, the market may see a broader migration toward similar architectures, especially as model improvements make the command‑line interface increasingly powerful. The story of two likes turning into a 4 % share of global GitHub activity underscores how a single developer’s pragmatic choice can reshape an entire industry.