OpenAI President Claims AI Generates 80% of Company’s Code

OpenAI president Greg Brockman announced at Sequoia Capital’s AI Ascent 2026 conference that artificial intelligence now writes about 80 percent of the company’s code. Speaking to an audience of investors and technologists, Brockman said, “It’s hard to know what percent is not being written by AI,” echoing a comment he made on the Knowledge Project podcast earlier this month.

The figure sparked immediate analysis among industry observers. Brockman’s statement can be read in two ways. One interpretation suggests AI tools generate 80 percent of the lines of code committed to OpenAI’s codebase, implying a dramatic boost in raw productivity. A second, more nuanced view sees AI involved in some capacity—autocomplete, refactoring suggestions, or generated code that engineers later revise—in 80 percent of coding tasks. Brockman’s own qualifier leans toward the latter, emphasizing the difficulty of measuring the exact non‑AI portion.

OpenAI is not alone in touting high AI‑coding percentages. Anthropic CEO Dario Amodei previously claimed that AI writes 90 percent of his company’s code, with a target of full automation within months. Similarly, GitHub Copilot reports 90 percent adoption among Fortune 100 firms, and Cursor has hit $2 billion in annualized revenue largely on the back of AI‑assisted development workflows.

Critics, however, caution against taking internal productivity numbers at face value. A February 2026 paper from the National Bureau of Economic Research found that 80 percent of firms using AI reported no measurable productivity impact. A 2025 MIT study concluded that 95 percent of corporate AI pilot programs delivered zero return on investment. Cognitive scientist Gary Marcus warned that large language models, while impressive, remain “deeply flawed imitators” that may overstate their capabilities, especially in producing secure, maintainable software.

Context of OpenAI’s Capital Deployment

OpenAI’s claim arrives as the company ramps up capital spending. In 2026 the firm raised $122 billion and is eyeing a potential trillion‑dollar IPO. Brockman has repeatedly said that compute, not model capability, is the new revenue driver. The productivity narrative around AI‑generated code plays a central role in justifying massive investment in compute resources.

Labor market dynamics add another layer of relevance. Over the past two years, many tech firms have laid off thousands of engineers, often citing AI‑driven efficiency gains as a rationale. If AI truly handles 80 percent of coding work at OpenAI and similar labs, the implications for engineering talent could be profound. Conversely, if the statistic reflects only AI assistance rather than replacement, the pace of layoffs may outstrip actual productivity improvements.

Brockman himself remains a prolific coder, reportedly spending 60 to 100 hours per week writing code. His dual role—as both the chief advocate of AI‑driven productivity and the company’s most active human programmer—underscores the tension between optimism and skepticism surrounding the 80‑percent claim.

Industry watchers will likely monitor OpenAI’s internal metrics as the company expands its compute capacity and prepares for a public offering. Whether AI truly writes the majority of code, or merely assists in most tasks, will shape expectations for future software development practices and the broader conversation about artificial general intelligence’s proximity.