OpenAI is taking AI-assisted development to the next level with its latest innovation: an autonomous software engineer called A-SWE. Unlike GitHub Copilot, which enhances a developer’s productivity by suggesting code snippets, A-SWE is designed to go beyond assistance—it can build entire applications independently.
YFO Sarah Friar unveiled A-SWE at the Goldman Sachs Disruptive Technology Symposium in London, calling it part of the company’s third phase in agentic AI evolution. “This is not just about augmenting engineers. It’s about creating an AI that can write, test, debug, and document code—end-to-end,” Friar said.
A-SWE follows the release of OpenAI’s earlier agents—Deep Research, which compiles detailed reports, and Operator, which executes web-based tasks like booking travel. Together, these tools mark a shift from AI as an assistant to AI as an autonomous worker.
A-SWE’s launch comes as competition intensifies. Last year, Cognition AI introduced Devin, the first AI software engineer to complete complex coding tasks solo. OpenAI’s offering is expected to go even further.
With the launch of GPT-4.5, OpenAI also focused on emotional intelligence (EQ) in AI—making the models more intuitive, creative, and “human-like” in execution.
The era of AI-built software may be closer than we think.