OpenAI’s president says AI coding tools have surged from generating 20% to 80% of code, pushing companies to rethink workflows.
AI coding tools are rapidly moving from side project to core software workflow, according to OpenAI president Greg Brockman.
Speaking at a Sequoia Capital talk uploaded by Misryoum. Brockman said the pace of change has been striking: AI agentic coding tools reportedly went from writing about 20% of code in December to roughly 80% shortly afterward.. His message to founders was direct: lean into these tools because they are improving quickly. and they are starting to reach beyond traditional software engineering roles.
In this context, the economics of software development are shifting. If AI can handle a much larger share of writing and iteration, companies may be able to move faster while reallocating costs away from routine coding tasks and toward review, testing, and product decisions.
Brockman also emphasized that AI-generated code still needs human accountability. He said OpenAI’s approach involves ensuring that a person is responsible for code that gets merged, positioning AI as a powerful assistant rather than an unchecked replacement for engineering judgment.
That theme is resonating across the industry. where multiple tech leaders have pointed to AI’s growing role in producing software.. Misryoum notes that the broader conversation is no longer about whether AI can draft code. but about how much of a company’s output it can generate. and what governance models are required to keep quality and security risks under control.
As AI takes on more of the “first draft,” teams may spend more time validating intent, performance, and safety instead of writing every line from scratch. For budgets and hiring plans, that can translate into different skill mixes and new expectations for review processes.
The shift also extends to how tools are packaged for different users. Brockman said OpenAI’s Codex has evolved from a platform mainly used by software engineers to one intended to support broader computer-related work, hinting at a wider labor impact for organizations trying to scale.
Beyond single-company statements, Misryoum sees a consistent direction: AI is becoming embedded in the mechanics of building software, and that forces a practical question for every organization. What should remain human-led, and what can be safely automated?
**Insight:** This matters because when AI accelerates code production. competitive advantage increasingly depends on how well companies manage the trade-off between speed and oversight.. The winners are likely to be those that pair automation with strong review. testing. and accountability rather than simply adopting tools without changing processes.