James Kovalenko

System architecture, machine learning, and forward deployed engineering for verification-bound technologies.

My research focuses on verification-bound technologies: systems where output volume grows faster than the capacity to test, trust, and operationalize it.

Web development

Current work

Manifold Control

Research and engineering for verification-bound systems.

Focus areas

  • AI verification and evaluation
  • system architecture
  • machine learning systems
  • forward deployed engineer
  • audit and repair infrastructure
  • capacity-constrained workflows
  • retention, reuse, and composability

Daily Temperature Markets

A measurement-first trading system for daily-high temperature prediction markets. Every decision leaves an audit row proving its inputs were fresh. Every strategy is scored against public baselines before it can touch capital. Every claim is bounded by what the atmosphere makes predictable at a given lead time.

System

  • measurement substrate before capital
  • calibrated probabilistic forecasting
  • multi-venue market pairing and settlement
  • shadow strategies and counterfactual ledgers
  • nightly calibration with full provenance
  • forecast-horizon bounds as doctrine

The Progress Function

A research program on how progress accumulates when generated variation passes through verification, diffusion, retention, and friction constraints.

The core frame is simple:

Generation
creates candidates.
Verification
admits structure.
Retention
lets structure compound.
Friction
measures unresolved interaction cost.