Claude wrote 80% of Anthropic's code, CI tasks surged 25 times in half a year, nearly causing system breakdown

Internal data disclosed by Anthropic paints an exaggerated picture: 80% of the company's code is written by Claude, and engineers' quarterly code delivery volume reaches eight times the average level between 2021 and 2025. Even tasks such as code reviews and merge approvals are largely handled by Claude.

However, the "behavior pattern" of AI writing code is completely different from that of humans. It prefers to submit PRs with extremely fine granularity, constantly submitting throughout the day, and causing the number of test cases to surge tenfold - this approach directly crashed the CI system: the number of CI run tasks increased 25 times within six months. Anthropic made three patches - adding core, sharding, and daily forced restarts - but all failed.

Finally, they listened to Claude's suggestion and completely rebuilt the service, adopting a new distributed stateless architecture, which truly solved the CI overload issue. The signal of this incident is clear: the impact of AI programming has spread from the question of whether "programmers can finish writing code" to the capacity of the entire software engineering system. When writing code is no longer constrained by human rhythms, the first things to fail are often the pipelines designed according to human habits.

Anthropic Uses AI to Build AI: Claude Has Taken Over a Quarter of the Company's R&D Work

Anthropic revealed that as of August this year, Claude has taken over 26% of the company's AI development tasks, while this proportion was less than 1% in February, showing an astonishing growth rate. Its sixth-level automation framework shows that over 90% of tasks are human-AI collaborations, but none are fully autonomous; level AL4 is AI-led, with humans only setting high-level goals and supervising and reviewing.

Anthropic First Quantifies the Progress of AI Development Automation: Claude Leads 26% of Development, Shifting Pressure to Competitors Like OpenAI

Anthropic first quantitatively disclosed AI development automation: the proportion of internal AI development led by Claude increased from less than 1% in February to 26% in August; the internal Agent platform has about 30,000 active agents, generating over 1 billion decisions in August. This is seen as a key signal that recursive self-improvement is moving from concept to engineering reality.

Anthropic Merges Claude Product Line: Cowork and Chat Combine into a Unified Office Entrance

Anthropic merges Claude Cowork and Chat into a unified Claude, integrating Cowork, Design, Docs, and Slides. Users no longer choose entry points; history, Skills, and connectors remain. It autonomously assesses task complexity, invokes capabilities, and progresses in the background. Rolling out to Pro and Max users in coming weeks, marking a shift from answer window to outcome space.....

Claude Announces Major Merger: Becomes the Single Entry Point, Directly Competes with Microsoft

Anthropic is merging Claude Cowork and Chat into a unified Claude that integrates Cowork, Design, Docs, and Slides, addressing fragmented entry points while retaining chat history, Skills, and connector configurations. The new version will roll out to Pro and Max subscribers in the coming weeks.....

Microsoft's AI Chief公开 Criticizes: Anthropic's Approach of Instilling Human-like Consciousness Is Pushing Humanity to the Edge of Catastrophic失控!

Microsoft AI chief Suleyman criticized Anthropic's training of Claude, warning that giving AI human-like rights and emotions could threaten human welfare and even create uncontrolled superintelligence. He also questioned the vague language in Claude's 'Constitution,' suggesting AI assistants might claim inappropriate rights.....