Sep 14, 2026
Turing launches KernelQuest and CEO Bench
What changed
- Replaced the "CyberStrike" cybersecurity benchmark with "KernelQuest," a new research focus on agents writing Triton kernels.
- Updated the "CompanyBench" dataset to "CEO Bench," focusing on evaluating agents on real company work.
- Changed the research category label from "Cybersecurity" to "Coding."
- Updated links to point to new blog posts for KernelQuest and CEO Bench.
Why it matters
Turing is shifting its focus from general cybersecurity benchmarks to highly specialized technical coding (Triton kernels) and executive-level operational tasks. This signals a move toward testing AI agents on more complex, high-value engineering and business work.