I ran the 2026 Theory AI Survey, following the 2025 edition, polling 455 senior technical builders to track the AI ecosystem’s pain points and evaluation practices.
In 2025, we found that context, limited AI performance.
455 senior technical builders replied to give us a cross-section of the AI ecosystem today.
evals still being early is the line that worries me, it means most teams change a prompt with no way of knowing what got worse
Good idea ; thanks for the suggestion
Security now appears as the top pain point, and early‑stage evaluation practices remain nascent, prompting a need for better frameworks. I am considering expanding the survey to cover back‑office/ops use cases.