PCPcaml

Pengyi Caml / AI-native Polyglot Coding OS

Semantic translation for real software systems.

Pcaml models Data, Type, State, Invariant, Interface and Boundary once, then materializes the same design into Python, C++, OCaml, TypeScript, Rust and SQL with tests and hostile review.

Pcaml IR TYPE -> STATE -> INVARIANT -> IMPLEMENTATION
ProblemSystem ModelLanguageTestsReview

Four Agent Engineers

Each agent has a focused page and a hard engineering contract.

Selene, the Pcaml Language Agent under a blue sky

Language Agent / black silk command look

Selene scans new languages into Pcaml IR.

She maps syntax, type systems, memory models, error models, concurrency and ecosystem constraints into one reusable language card.

  • Language -> Pcaml IR
  • Paradigm comparison
  • Wheel and runtime risk scan
Iris, the Pcaml Design Agent under a blue sky

Design Agent / white satin architecture look

Iris turns problems into system architecture.

She extracts Data, State, Invariant, Transition, Interface, Boundary and Ownership before any code is written.

  • Problem -> System Model
  • Boundary design
  • Architecture-preserving translation
Vera, the Pcaml Review Agent under a blue sky

Review Agent / sky-blue technical look

Vera attacks broken invariants.

She compares implementation against the IR, searches for impossible states, contract drift, hidden effects and unmeasured claims.

  • Hostile review
  • Invariant and test gap scan
  • Regression triage
Mira, the Pcaml Builder Agent under a blue sky

Builder Agent / graphite leather build look

Mira materializes the model.

She builds implementation slices, fixtures, parity tests and benchmark cards without claiming production quality before measurement.

  • IR -> implementation
  • Tests and examples
  • Benchmark evidence capture

Pcaml IR

Not source-to-source translation. Semantic translation.

Pcaml preserves behavior, invariant, interface contract, state transition and tests. The target language is free to express the same system with its native best practices.

DataTypeStateInvariant TransitionFunctionInterfaceBoundary OwnershipFailureEffectPerformance

Build Pipeline

01Real problem

Order, trade, position, risk or research workflow.

02Pcaml IR

State, invariant and boundary become explicit.

03Language build

Python, C++, OCaml, TS, Rust or SQL implementation.

04Review gate

Tests, benchmark and human approval decide promotion.

Evidence State

Private repo live. Website live. Claims remain measured.

Current cross-language parity and production performance remain UNMEASURED until benchmark cards prove them.