In a striking milestone for the artificial intelligence sector, Anthropic revealed in its S-1 regulatory filing on Monday that over 90 percent of its internal codebase was written by its proprietary AI model, Claude, during September.

Closing the Loop

The artificial intelligence pioneer is increasingly relying on its own models to research, design, and engineer subsequent iterations—a strategy the firm describes as "closing the loop." Within the filing, the company emphasized that Claude has evolved into a primary engine driving its technological momentum.

Presently, Claude executes experiments, synthesizes training data, and streamlines evaluations and safety reviews. Furthermore, the model has enabled critical algorithmic optimizations, allowing Anthropic to maximize its computing infrastructure efficiency.

Financial Highlights and IPO Outlook

Preparing for an upcoming public debut, Anthropic disclosed its financial standing for the first time to prospective investors, reporting $16.3 billion in revenue alongside $693 million in operating losses during the initial six months of 2026. However, the enterprise achieved its first profitable quarter in Q2 2026, posting $974 million in operating income.

Given its 5,300-strong workforce, heavy reliance on Claude is expected, yet internal AI operational expenses have now surpassed salary expenditures. By April 2026, the calculated value of internal Claude utilization surpassed monthly personnel costs, excluding stock-based compensation. According to the S-1, the value of internal Claude usage calculated at standard list prices surged more than 250-fold between January and September 2026. Consequently, human engineers produced over 45 times more code daily in Q3 2026 compared to 2024.

Autonomous Engineering and Operational Multipliers

Claude Code independently tackles sophisticated programming challenges, with human engineers transitioning primarily into supervisory roles over autonomous agents. Personnel across operations, go-to-market, and research divisions similarly deploy agent teams to automate routine responsibilities, empowering leaner teams to scale output and multiply individual impact.

Structural Advantages and Emerging Risks

Although Anthropic regards Claude's self-improvement capabilities as a vital structural edge, the dynamic introduces substantial operational hazards.

By assisting in the creation of future model generations, Claude compresses timelines and accelerates the developmental trajectory. Anthropic cautioned in the filing that a central risk involves artificial intelligence advancing so rapidly that the firm's safety and security protocols could lag behind. Defects embedded in current iterations might propagate into successors, complicating detection and remediation. Moreover, AI-driven research obscures the true velocity of progress until models are fully realized. With competitors adopting similar recursive training methods, falling behind in efficiency remains a persistent threat.

To mitigate these hazards, Anthropic is advocating for a collaborative, industry-wide framework to regulate frontier progression and mandate independent safety assessments. Over the next half-decade, the company plans to invest $1 billion to establish external review capabilities in partnership with Accenture.