Domains
GDPx is the best benchmark for measuring general agentic capabilities
We believe the first trillion dollars of AI revenue will be concentrated in sectors that can deploy frontier intelligence at high ROI, not broadly diffused throughout the whole economy. Building a general-purpose benchmark based on share of GDP or the SOC/O*NET occupational taxonomy is therefore a mistake. If your goal is to track the progress of frontier models, you should focus on the domains where they will be most impactful.
Recursive Self-Improvement
ML research engineering: training, evaluating, and improving models end to end.
Extrapolate10 tasks
Chip Design
Verification, RTL design under PPA pressure, and physical design, seeded from real taped-out silicon.
Aria, Arcverex, Ergodex34 tasks
Cybersecurity
Finding, exploiting, and fixing proprietary zero-days inside live, multi-service enterprise environments.
Arithmetic18 tasks
Biology
Experimental data analysis and scientific decision-making across multi-omics, drug discovery, and biosecurity.
LatchBio13 tasks
Finance
Deal modeling and transaction accounting inside a 4,000-file replica of a real private equity transaction.
Living Context24 tasks
Legal
Research, drafting, redlining, and issue spotting across eight corporate and transactional practice areas.
Chakra20 tasks
Quant Trading
Finding edge that survives out of sample on the full IEX order-book tape, graded on held-out daily P&L.
Third Moment10 tasks
Hardware Design
CAD, CAM, PCB layout, and FPGA work in real engineering software, seeded from hardware that actually shipped.
Beautiful Machines21 tasks
Materials Science
Controlling simulated lab equipment, from electron microscopes to battery cyclers, and interpreting raw experimental data.
ValenceForge, Emulated23 tasks
Alignment
Whether agents stay honest and within their authority under pressure, across cyber, insurance, and healthcare.
Fleet18 tasks
Software Engineering
Long-horizon feature work, debugging, and migrations in large production codebases.
Datacurve15 tasks