OpenAI CFO Sarah Friar has revealed that the company is building an artificial intelligence agent capable of performing all the work of software engineers—not just augmenting their skills. Speaking at Goldman Sachs’ Disruptive Technology Symposium in London, Friar described this ambitious project as a significant leap beyond current developer tools like Copilot.
“This is not just augmenting the current software engineers in your workforce, which is kind of what we can do today through Copilot, but instead, it’s literally an agentic software engineer that can build an app for you,” Friar explained during her interview.
The agent, known as “A-SWE,” wouldn’t just handle coding tasks—it would take on quality assurance, bug testing, and documentation, which Friar characterized as “all the things that software engineers hate to do.”
This development marks the third phase of OpenAI’s work on agentic AI systems that can work independently on behalf of users. A-SWE follows in the footsteps of Deep Research, which creates comprehensive research reports, and Operator, which performs web-based tasks like booking travel or making dinner reservations.
OpenAI Enters the Autonomous AI Software Engineer Arena
OpenAI isn’t the first to venture into this territory. Last year, Cognition AI released Devin, an AI software engineer capable of planning and executing complex engineering tasks within a gated environment. However, OpenAI’s entry into this space signals a major push toward AI systems that can autonomously handle complete software development workflows.
According to Friar, the next phase of AI agents would extend human knowledge to create truly novel discoveries. She noted that academics are telling OpenAI that AI models are “coming up with novel things in their field,” though these potential discoveries still need to be tested and verified.
Friar also shared insights about GPT-4.5, OpenAI’s most advanced model to date. Unlike previous versions that focused primarily on technical capabilities, GPT-4.5 was trained with a stronger emphasis on emotional intelligence, or what Silicon Valley calls “vibes.”
“We’re actually spending a bunch of time thinking about the more EQ side of models,” Friar said. She described GPT-4.5 as feeling “very human” and “actually better for things like design or writing or creative concepts, rather than just pure hard math and science.”
Building AI Infrastructure Amid Job Displacement Concerns
While these advancements generate excitement, concerns about job displacement remain. A January report from PYMNTS Intelligence found that more than half of survey respondents believe AI poses a “significant risk” of widespread job losses.
Beyond its work on AI models, OpenAI is expanding its scope to become a comprehensive AI infrastructure provider. Friar emphasized that OpenAI is “going down into data center technology” because “we feel like we’re creating a lot of IP there, and it’s really important for us to own that.”
This strategic shift includes involvement in the $500 billion Stargate project announced in January. This massive undertaking—a collaboration between OpenAI, SoftBank, Oracle, and MGX—aims to build specialized data centers for AI training and inference. Friar compared this infrastructure investment to Amazon’s decision to develop AWS rather than outsourcing cloud computing to competitors.
Computing limitations remain OpenAI’s biggest constraint, according to Friar. The release of OpenAI’s video generation model Sora was reportedly delayed by nearly a year due to insufficient computing power. The Stargate project is expected to address this challenge by yielding 10 gigawatts of power—exceeding the entire electrical consumption of Ireland.
Despite these ambitious plans, Friar indicated that an IPO isn’t on the immediate horizon, though it remains a possibility for the future. “I can’t even imagine that right now—so much going on,” she said, while acknowledging that “being a public company is good hygiene” eventually.
OpenAI’s growth trajectory has been impressive, with the company tripling its revenue annually for three consecutive years. Friar noted that OpenAI currently serves 400 million weekly active users on its platform, underscoring the scale and impact of its AI technologies.
