Software Engineering in 2026 (benjamincongdon.me)

As we head into 2026, the evolution of large language model (LLM) coding tools is poised to reshape the landscape of software engineering. The significant reduction in the cost and time for producing high-quality code has transformed the "building" phase of software development, while tasks like "evolving" and "operating" software systems now face new bottlenecks. Companies, particularly product organizations, are encouraged to harness LLMs to drive productivity, but a growing reliance on these tools necessitates a shift in practices, such as ensuring robust CI infrastructure and human code reviews to maintain code quality. Key transitions include the prioritization of infrastructure abstractions that support both human and LLM users, the rethinking of testing methodologies to better suit automated code generation, and a heightened emphasis on creating stringent human-guided abstractions to combat potential technical debt. As the role of human code reviews becomes more critical, engineers will need to cultivate "review taste" more quickly, even as they spend less time writing code. The market will likely see fluctuating project timelines as the variability in LLM assistive tasks impacts overall cost estimations, while the "build vs. buy" dilemma is likely to shift towards building for simple SaaS solutions, leaving more complex services relatively unchanged.