An official uptake report and an academic long-horizon benchmark landed in the same week.
What happened
Cohere published "Automation's Early Footprint," measuring how far automation has actually spread, while arXiv carried CivBench, a long-horizon benchmark for tool-mediated agents. One vendor post, one paper — thin by volume.
Why it matters
Both measure adoption and durability rather than new capability. Measurement infrastructure is arriving ahead of the next feature wave, shifting the evaluation axis toward what is actually deployed. With only two sources, this is not yet confirmed as a trend.
What to watch
Whether other teams adopt CivBench-style long-horizon benchmarks, and whether Cohere's footprint measurement becomes a recurring report rather than a one-off.