AI Predictions and the Uncertainty Facing Software Engineering [Anthropic, NVIDIA, Programmer Commentary]

Recent statements from Anthropic's CEO and an Anthropic engineer claim that AI will soon eliminate the need for software engineers, predicting that "90% of cobas can be written by AI" and forecasting that the field will be obsolete by early 2026. Jensen Huang, CEO of NVIDIA, stated that "nobody was going to need to learn to code anymore because of AI." Yet, a programmer reflecting on these claims observes that since early 2023, the countdown to the "death" of software engineering has been consistently six months away.

The programmer emphasizes that AI's impact is not solely a technical matter; it's equally about organizational decisions. Even if AI-generated code fails to compile most of the time ("maybe it only compiled 5% of the time"), companies could theoretically eliminate programmers regardless. However, today's AI coding tools are increasingly capable, and seasoned programmers are integrating AI for daily tasks such as generating unit tests and business logic. The speaker maintains responsibility for high-level decisions and debugging, a skillset still demanded despite AI assistance.

Risk is presented on a spectrum: at one extreme, all technical jobs are automated, collapsing the economy; at the other, AI progress stalls and coding jobs remain secure. Reality likely resides between these extremes, with two envisioned outcomes:

1. The Job of Programmer Transforms: AI will automate routine coding ("writing out a for loop"), and programmer roles will become focused on decision-making, system design, and debugging. This shift explains the shrinking junior market for coders and continued strength at the senior level. AI may make advanced skills more accessible, accelerating junior developers' progression. Thus, coding remains fundamental, especially for debugging, which requires code comprehension.

2. Technical Roles Evolve Entirely: AI could eventually replace more complex tasks (including advanced debugging), birthing new technical job categories. Future roles may involve managing AI agents or configuring AI workflows, diverging from traditional titles like "software engineer," "data scientist," or "product manager." Two reasons support this outcome: business leaders will still desire human guidance, and new jobs will likely emerge in tech as investments in AI and technology increase. The speaker warns these future qualifications are unknown but expects technical work to remain essential.

Recommendations stress adaptability. For those entering the field, the speaker urges learning to code and recommends Python as a beginner-friendly, widely-used language for ML and AI projects. Building technical foundations is crucial, as is developing "senior" skills early—debugging large systems and system design. Staying up to date with AI and tech research will help new entrants prepare for evolving job requirements.

While uncertain, the presenter encourages flexibility: those comfortable with shifting job duties will continue to find value in technical careers, even as the specifics of software engineering change.