Validated industrial data for every climate vertical.
We're building the industrial data repository for entrepreneurial climate scientists evaluating commercial viability — expert-validated economic and process data, covering every climate vertical, organized around how TEAs are actually built. Free and open as a public good.
The economic anatomy of an industrial process.
The repository is organized by reference class — the established and emerging industrial processes that breakthrough technologies slot into, displace, or recombine. Every reference class includes raw material costs, capex benchmarks, performance ranges, and carbon intensity data — anchored to a validated process flow diagram.
The data repository is the foundation for the AI-enabled TEA tooling. → Learn about the tooling
Process flow diagram
Drag nodes, pan, and zoom to explore.
Data collected per reference class
System Definition
Process flow diagram, functional unit & system boundary, mass & energy balances
Process & Performance
Efficiency & yield, operating conditions, capacity factor, equipment lifetime
Capital Costs
System installed cost (CAPEX), equipment costs (uninstalled), scaling exponents
Operating Costs
Feedstock & raw materials, utilities, labor & overhead, maintenance rates
Market & Product
Product specifications & purity, price benchmarks
Derived Economics
LCOE / LCOH / LCOX benchmarks, sensitivity & cost drivers
Metadata
TRL classification, source citations with confidence ratings, validation status
A hybrid AI and expert validation pipeline.
Each reference class is built and validated through a four-stage process combining AI-assisted data collection with domain expert review.
Orchestration of tailor-built agents for baseline flow
Automated iteration to mimic expert-level judgement
"Best AI can do" evaluation & iteration
Reference class complete
Continuous integration of expert feedback
Comparison against known examples and curated data
Panel of agents using expert-defined sector heuristics
Each review refines heuristics to improve the automated pipeline
Manual expert validation (5–20 experts per class)
Each expert review refines the heuristics that improve the automated pipeline — so the system gets better with every class we build.