Descriptions:
In this wide-ranging interview on the 20VC podcast with Harry Stebbings, Alexander Bericos—product lead for Codex at OpenAI—discusses the future of software engineering, OpenAI’s strategic positioning in the coding tools market, and what it means for the profession when AI can generate most of the code.
Bericos pushes back on alarmist narratives about software engineering becoming obsolete, drawing historical analogies to the shift from assembly language to higher-level programming—each transition automated specific tasks while expanding overall demand for developers. He describes a “compression of the talent stack” where engineers become more full-stack and product-aware, and predicts significantly more builders in the future. On competitive dynamics, he offers candid insight into how the Codex team thinks about retention against rivals like Cursor and Claude Code, acknowledging the unusual position of training models that also serve competitors. He frames this as intentional: if competitors improve, OpenAI learns from the ecosystem.
Bericos highlights Codex’s core structural advantages: ChatGPT’s massive distribution, proprietary model training tightly coupled to their agent harness, and what he describes as the most conservative sandboxing approach in the space. He also comments on recent launches from Warp, framing the entire agentic coding landscape as different companies converging on the same architectural conclusions simultaneously. For anyone tracking OpenAI Codex, Cursor, or Claude Code, this interview offers rare first-party perspective on product strategy and the long-term vision for AI-assisted software development.
📺 Source: 20VC with Harry Stebbings · Published February 21, 2026
🏷️ Format: Interview
