The OpenAI company is working on creating a new AI agent, A-SWE, capable of replacing a software engineer and performing self-testing. About this informs Officechai.
OpenAI CFO Sarah Friar announced work on a project codenamed “Agentic Software Engineer,” which will be an autonomous assistant for software development and code bug detection.
Previously, OpenAI has already introduced the Responses API — a tool for developers that allows you to create agents with the functions of searching the Internet, analyzing files, and performing computer tasks on behalf of the user. But A-SWE, according to Friar, will not be another standard LLM assistant like Copilot. It will be a full-fledged developer who is able to create programs, process PR requests from users and act independently, without human intervention.
In addition to directly writing software code, A-SWE will also be able to take on tasks that developers usually avoid. According to Sarah Friar, the new agent will test himself on the quality of the work performed, correct errors and work with documentation. The CEO of OpenAI, Sam Altman, previously stated that by the end of 2025, artificial intelligence could become the best programmer in the world.
We will remind that at the beginning of April, the company OpenAI announced the attraction of 40 billion dollars of investments for the development of AI research. In addition, the company invested for the first time in the field of cyber security – the startup Adaptive Security. This company not only detects potential threats, but also conducts stress tests for employees, simulating fraud attempts.
