Mark Zuckerberg’s 2025 AI Engineer Prediction and the Future of Software Jobs

In a January 2025 interview with Joe Rogan, Meta CEO Mark Zuckerberg predicts that Meta will probably have AI capable of performing work like a mid-level engineer during 2025. This is a forecast, not a demonstration that such a system already exists. I see it as part of a larger shift in both what software is and how it gets built.

Meta has considerable resources to invest in AI, and I think the workforce could change rapidly. Looking beyond 2025, I speculate that AI will eventually write much of the software we use. The timing, scale and employment consequences are uncertain, even if the technology improves quickly.

I wonder whether code will remain in familiar forms or be generated when needed. My tentative picture is that databases still provide a source of truth, while agents create tools, interact with those databases and generate interfaces for a particular request. That would change what we mean by an application; it is a possible future, not an established deployment.

In the interview clip, Zuckerberg says AI engineering may initially be expensive, become more efficient and eventually produce much of the code in Meta’s apps and AI systems. He expects this to augment people and free them to pursue more creative work. His statements express an anticipated trajectory, not a measured present-day outcome.

That raises a review problem. If AI writes increasing volumes of code, will people be able to understand and review all of it, or will AI systems review some of it too? Human reviewers could become a bottleneck, and teams may need new systems for testing and oversight.

I share Zuckerberg’s optimistic view that AI can complement people and make them more productive. More output does not automatically tell us how many workers a company will employ, however, and the effect can vary across roles and organizations.

Rogan raises the concern that AI could eliminate jobs. Zuckerberg says it is too early to know exactly how things will play out, but guesses that the technology may create more opportunities for creative work. He compares the potential transition with the long decline in the share of people working in agriculture.

Zuckerberg suggests that agricultural productivity allowed more people to pursue cultural, creative and other work or leisure. He predicts that the share of people doing work strictly required for survival could continue to shrink, while people who are deeply engaged in their work could accomplish more with technology.

I agree with the broad optimism and ask what happens if a software engineer can produce far more software. My expectation is that much of the additional capacity will be used to build more. Lower costs can make previously uneconomic uses worthwhile and increase demand, although the actual response differs across markets.

The transition will not necessarily be smooth. I expect layoffs and a period of friction as companies and workers adjust. Productivity gains alone do not settle whether particular people will retain their jobs or whether new opportunities will appear quickly enough for everyone affected.

As a counterpoint, I cite a December 2024 report quoting Salesforce CEO Marc Benioff on the company’s 2025 plans. He said Salesforce was not adding more software engineers the following year and attributed an engineering productivity increase of more than 30% to Agentforce and other AI technology. That percentage is his company’s claim, and a hiring plan is not the same as a verified outcome.

The same account says Benioff expected fewer support engineers while potentially adding 1,000 to 2,000 salespeople to explain and sell AI services. I am skeptical of the marketing incentives around the claims, but the contrast is revealing: the company was discussing reduced demand in some roles alongside expansion in others.

Software is changing, and I encourage engineers to experiment with tools such as Cursor and Pythagora. I expect smaller teams to solve more long-tail problems, although that is a prediction rather than a guarantee. I also mention Sam Altman’s prediction of a billion-dollar one-person company enabled by AI.

The possibility that AI gives small teams more leverage is particularly exciting to me as someone interested in entrepreneurship. I hope it helps people start companies and solve useful problems. That optimistic scenario does not remove the uncertainty and disruption for workers during the transition.