"The biggest challenge in your stack? Evals, Evals, Evals" - 2026 State of AI Engineering results

AI engineering has evolved into a cross-functional discipline, with senior developers rapidly adapting to new paradigms. Data from over 1,000 respondents reveals that cost is now a primary engineering constraint, influencing how teams architect and scale AI applications. While open-weight models are increasingly used, they serve as an augmentation to closed models rather than a replacement, as teams prioritize quality and agentic capabilities. A significant shift is occurring in agent deployment; 89% of teams now grant their agents write access, moving them beyond simple summarization into active system interaction. Despite this progress, control mechanisms remain largely experimental, and evaluation methods continue to rely heavily on subjective reviews. Ultimately, while AI adoption boosts job satisfaction and experimentation, it introduces new long-term maintenance liabilities and blurs traditional boundaries between engineering, product, and design roles.