A trend report does not ask what is hot but how the number was produced. When surveys of the same question report 90% and 59% at the same time, it is usually not that one is wrong but that the sample, the wording of the question or the survey period differ. We also note whether the organization behind a survey sells products in that field and whether responses were voluntary.
Why JetBrains' 90% and Stack Overflow's 59% can both be right; what reference labels and comparison baseline produced TypeSafe's claims of 0.4 seconds per case and 67.8% agreement for its Jev model; and the five axes along which the coding-agent landscape moved in the first half of 2026.