Build the shared Atomspace inference engine that powers all four interfaces, plus the dual-language LSP server and two foundational CLI tools (prism parse, prism verify). The Atomspace engine parses Rholang and MeTTa into typed atoms with semantic relationships, manages persistent storage, and provides efficient Hyperon inference query interfaces. The LSP server formulates completions as backward-chaining queries for type-directed synthesis, tracks Rholang channel bindings through par compositions, and leverages MeTTa's native type system for pattern-aware completions. The parse CLI converts source files into structured AST and ingests into Atomspace. The verify CLI performs formal verification using MeTTa dependent types via Hyperon inference, with proof traces on failure. Both CLI tools follow Unix conventions with stdin/stdout JSON composition. The LSP and CLI share the same Atomspace instance and inference engine. This milestone establishes the core architecture: one knowledge graph, multiple access patterns, inference-first intelligence.

1. Shared Atomspace engine with persistent storage, Hyperon inference query interfaces, and cross-invocation knowledge accumulation serving as semantic substrate for all interfaces. 2. Rust-based dual LSP server for Rholang and MeTTa with type-directed completions via Hyperon backward chaining, diagnostics, hover info, go-to-definition, and VS Code extension. 3. prism parse CLI: full Rholang and MeTTa grammar, JSON AST output, simultaneous Atomspace ingestion with typed semantic relationship links. 4. prism verify CLI: MeTTa-native dependent type checking via Hyperon inference with proof trace output on failure, supporting both Rholang behavioral types and MeTTa type inference. 5. Unix pipe composition: parse | verify works with structured JSON flowing between tools, sharing the Atomspace with the LSP. 6. Test suite with 250+ cases covering LSP features, both parsers, verification, and pipe composition. 7. Architecture documentation covering Atomspace schema, inference query patterns, and multi-interface design.

$10,000 USD

1. LSP provides completions within 200ms for projects up to 50 files. Type-directed completions achieve 95%+ type-correctness rate vs 60-70% for LLM baselines. 2. prism parse correctly handles 100% of test corpus (50+ files each language) including edge cases. 3. prism verify detects 90%+ of injected type errors with proof traces identifying failing inference steps in 85%+ of cases. 4. Pipe composition parse | verify processes 50-file project under 10 seconds. 5. Atomspace queries respond under 100ms for graphs up to 10,000 atoms. 6. VS Code extension installs on Windows, macOS, Linux. CLI tools distribute as single binaries. 7. All 250+ test cases pass. Three developers validate on existing projects.

Deliver the Hyperon inference AI engine (prism ai), deployment CLI (prism deploy), and the MeTTa-to-Rholang intent compilation pipeline. The AI engine performs type-directed synthesis via Hyperon backward chaining with four modes: generate, transform, optimize, explain. It implements concept blending for novel patterns by composing Atomspace representations with compositional proofs. The deploy CLI handles full BlockDAG DevNet lifecycle: transaction construction, CBC Casper finality monitoring, Singularity Compute resource estimation, and Atomspace-informed dependency analysis. The intent compiler transforms MeTTa dependent type specifications into verified Rholang through systematic type-driven template composition. A curated library of 15+ verified MeTTa-to-Rholang templates covers token operations, oracle interactions, and multi-party protocols. Hyperon inference validates each composition step. All three components feed interactions back into the shared Atomspace, beginning the self-extending feedback loop with evolutionary optimization of inference strategies.

1. prism ai CLI with generate/transform/optimize/explain modes, each powered by Hyperon inference with explicit proof traces, plus concept blending for novel pattern synthesis. 2. prism deploy CLI: compile, construct transaction, submit to BlockDAG DevNet, monitor CBC Casper finality, produce deployment receipts. Singularity Compute integration for resource estimation. 3. MeTTa-to-Rholang intent compiler with type-directed synthesis from dependent type specs to verified Rholang contracts. 4. Verified template library (15+ templates): token operations (5+), oracle interactions (4+), multi-party protocols (6+), each with MeTTa spec, Rholang implementation, and Hyperon-verifiable correctness proofs. 5. Self-extending feedback capturing accepted/rejected suggestions as Atomspace evidence atoms with evolutionary optimization. 6. Full pipeline: parse | ai --mode=optimize | verify | deploy --devnet works end-to-end. 7. Intent compilation handles 3-level template compositions with preserved correctness at each level.

$35,000 USD

1. prism ai generates type-correct code for 85%+ of standard patterns (token, multi-sig, escrow, auction, oracle) verified by prism verify. 2. Concept blending produces viable patterns for 70%+ of composition requests, validated by reviewers and type-checking. 3. Deploy completes end-to-end to DevNet within 60 seconds including CBC Casper finality. 4. Singularity Compute estimates within 20% of actual costs for 90%+ of deployments. 5. Intent compiler compiles 15+ distinct MeTTa specs into valid Rholang. Hyperon confirms correctness for 90%+. 6. Self-extending feedback shows 10%+ suggestion acceptance improvement over 100 interaction cycles. 7. Three developers unfamiliar with system produce verified Rholang from MeTTa spec within 2 hours.

Deliver inference-trace learning mode across all tools, property-based testing CLI (prism test), web playground, and community launch. Learning mode adds explanation capabilities to every interface: LSP shows why it suggested completions, verify shows proof traces as teaching, the compiler shows derivation trees, and ai shows inference chains. A structured curriculum (30+ exercises from atoms through dependent types to contract specs) runs natively in the Atomspace with self-extending exercise generation from learner error patterns. The test CLI auto-derives properties from Atomspace type signatures, uses symbolic evaluation where possible, and produces minimal counterexamples. The web playground embeds LSP intelligence via WebAssembly in a Monaco editor with managed DevNet connectivity. Community launch targets functional programmers, type theory researchers, and Rholang developers with structured onboarding, tutorials, and CI/CD integration templates.

Deliver inference-trace learning mode across all tools, property-based testing CLI (prism test), web playground, and community launch. Learning mode adds explanation capabilities to every interface: LSP shows why it suggested completions, verify shows proof traces as teaching, the compiler shows derivation trees, and ai shows inference chains. A structured curriculum (30+ exercises from atoms through dependent types to contract specs) runs natively in the Atomspace with self-extending exercise generation from learner error patterns. The test CLI auto-derives properties from Atomspace type signatures, uses symbolic evaluation where possible, and produces minimal counterexamples. The web playground embeds LSP intelligence via WebAssembly in a Monaco editor with managed DevNet connectivity. Community launch targets functional programmers, type theory researchers, and Rholang developers with structured onboarding, tutorials, and CI/CD integration templates.

$5,000 USD

1. Learning mode: 80%+ of test users correctly understand inference traces and identify why suggestions were made. 2. Curriculum: 10+ developers complete atoms-to-contracts progression. Self-extending generator produces 10+ exercises from error patterns. 3. prism test derives meaningful properties for 80%+ of test contracts, catching 70%+ of injected bugs. 4. Playground loads under 5 seconds, WASM completions under 500ms. New developers deploy first contract within 30 minutes. 5. Five-tool pipeline parse | test | ai | verify | deploy works end-to-end on five contract types. 6. CI/CD templates work out-of-box on GitHub Actions for three sample projects. 7. Documentation rated positively by 5+ community reviewers. 20+ developers onboarded at launch.