DEEP Connects Bold Ideas to Real World Change and build a better future together.

DEEP Connects Bold Ideas to Real World Change and build a better future together.

Coming Soon

Meridian: Inference-Native Developer Toolkit

chevron-icon
RFP Proposals
Top
chevron-icon
project-presentation-img
user-profile-img
Schreiber
Project Owner

Meridian: Inference-Native Developer Toolkit

Expert Rating

n/a

Overview

Meridian delivers a TypeScript LSP server for Rholang and MeTTa where every code intelligence feature is powered by Hyperon inference over Atomspace code graphs, not LLM prediction. M1 delivers dual-language LSP with type-directed synthesis via Atomspace. M2 adds AI coding assistant using Hyperon backward-chaining and DevNet deployment CLI with BlockDAG and CBC Casper finality. M3 wraps the LSP in a web playground for zero-install onboarding. Developer patterns feed back into Atomspace via Singularity Compute, creating self-extending intelligence. Not a GPT wrapper. MeTTa reasoning about code.

RFP Guidelines

Open for Proposals

An AI-native Development Environment for...

Ending on:

13 Feb. 2026

Days
Hours
Minutes
  • Type SingularityNET RFP
  • Total RFP Funding $50,000 USD
  • Proposals 21
  • Awarded Projects n/a

An AI-native Development Environment for the ASI:Chain

Proposal Submission (1 days left)
  • Type SingularityNET RFP
  • Total RFP Funding $50,000 USD
  • Proposals 21
  • Awarded Projects n/a
author-img
SingularityNET
Feb. 4, 2026

This RFP seeks proposals for the development of an AI-native Development Environment (IDE) that improves the efficiency and accessibility of blockchain application development for the ASI:Chain.

Proposal Description

Our Team

Archway Research builds developer infrastructure combining formal methods, decentralized systems, and AI-native tooling. Our team has compiler engineering expertise (LSP implementations, type systems) and deep Hyperon/Atomspace experience. We ship production TypeScript developer tools for blockchain platforms with active Rholang and MeTTa tooling contributions. Our research on neural-symbolic code intelligence directly informs Meridian.

Company Name (if applicable)

Archway Research

Project details

Meridian: Inference-Native Developer Intelligence for Rholang and MeTTa

 

## The Problem

 

Blockchain developers writing Rholang and MeTTa deserve tools whose intelligence matches the mathematical precision of the languages themselves. Current developer tooling either uses hand-written heuristics or wraps LLMs that statistically predict what code looks like without understanding process calculus semantics, channel linearity, or behavioral types. Meridian takes a different approach: every code intelligence feature is a Hyperon inference over Atomspace knowledge representations. Every suggestion is logically guaranteed, not statistically likely.

 

## Why TypeScript for the LSP Server

 

The Hyperon/MeTTa runtime has mature JavaScript bindings, enabling direct Atomspace queries from the LSP process without FFI overhead. Second, TypeScript LSP servers have proven production patterns (typescript-language-server, vscode-eslint) with battle-tested frameworks like vscode-languageserver-node. Third, TypeScript dramatically expands the contributor pool: any SingularityNET developer who writes MeTTa can also contribute to the tooling. The inference intelligence lives in Atomspace and Hyperon, not in the implementation language.

 

## Architecture: Three Components, Inference-First

 

### 1. Dual-Language LSP Server with Atomspace Code Graph

 

The TypeScript LSP server maintains a live Atomspace graph of the developer's codebase. Every function definition, type annotation, channel binding, and contract interaction is represented as atoms with typed relationships. When the developer requests a completion, the LSP formulates this as a Hyperon backward-chaining query: given the current type context and available bindings, what expressions satisfy the type constraints? The result is type-directed synthesis producing completions that are logically guaranteed well-typed.

 

For Rholang, the LSP tracks channel names through par compositions, reasons about name equivalence under restriction, and diagnoses linearity violations. For MeTTa, it leverages Hyperon's native type system for pattern-aware completions including grounded atoms, custom types, and higher-order patterns.

 

The Atomspace code graph is the foundation everything else builds on. It captures definitions, types, imports, and dependencies as typed atoms. This enables queries like "what functions accept this channel type" or "what contracts depend on this name" as inference problems rather than text search.

 

### 2. AI Coding Assistant with DevNet Deployment CLI

 

The AI assistant uses Hyperon inference, not LLM generation. When a developer describes a contract, the assistant decomposes the specification into sub-goals using backward chaining over Atomspace, then constructs code where every step has an explicit proof trace. It implements concept blending for novel patterns: combining known contract structures through Atomspace compositional reasoning with proofs that composition preserves properties.

 

The DevNet CLI handles compilation, deployment to BlockDAG infrastructure, CBC Casper finality monitoring, and Singularity Compute resource estimation. It uses Atomspace analysis for dependency ordering and gas optimization. The CLI captures deployment metadata as atoms, building knowledge that improves future deployments.

 

The self-extending feedback loop begins here: accepted and rejected suggestions become evidence atoms. Hyperon evolutionary optimization retains inference strategies that produce accepted code and prunes those that do not.

 

### 3. Web Playground

 

A thin wrapper: Monaco editor connecting to the LSP server via WebSocket, managed DevNet node for live deployment, and guided tutorials. The playground demonstrates inference vs. statistical prediction with side-by-side comparison showing proof traces alongside confidence scores. It serves as onboarding for new Rholang and MeTTa developers.

 

## Why This Matters for SingularityNET

 

This is multi-paradigm AI in practice: neural perception understands developer intent, symbolic reasoning performs type-directed synthesis via Hyperon, evolutionary optimization improves suggestions over time, concept blending creates novel patterns. The intelligence lives in Atomspace inference rules, open-source and inspectable. No centralized API, no opaque model. Decentralization as the lever against dystopian AI outcomes, applied to the developer experience.

 

## Technical Foundation

 

Meridian builds on Hyperon alpha runtime, Atomspace storage, and MeTTa interpreter. The LSP is TypeScript using vscode-languageserver-node with Hyperon JS bindings. The CLI is a standalone Node.js binary with embedded MeTTa runtime. The playground uses the same LSP over WebSocket. All components share an Atomspace instance that accumulates developer knowledge. CBC Casper consensus monitoring tracks deployment finality. BlockDAG navigation resolves dependencies. Singularity Compute integration estimates resources. DevNet connectivity ensures tools work against live infrastructure from day one.

Open Source Licensing

MIT - Massachusetts Institute of Technology License

Background & Experience

Archway Research builds developer infrastructure combining formal methods, decentralized systems, and AI-native tooling. Our team has compiler engineering expertise (LSP implementations, type systems) and deep Hyperon/Atomspace experience. We ship production TypeScript developer tools for blockchain platforms with active Rholang and MeTTa tooling contributions. Our research on neural-symbolic code intelligence directly informs Meridian.

Links and references

MeTTa Language: https://metta-lang.dev/

Hyperon Implementation: https://github.com/trueagi-io/hyperon-experimental

MeTTa Examples: https://github.com/trueagi-io/metta-examples

Language Server Protocol: https://microsoft.github.io/language-server-protocol/

SingularityNET Developer Portal: https://dev.singularitynet.io/

Proposal Video

Not Avaliable Yet

Check back later during the Feedback & Selection period for the RFP that is proposal is applied to.

  • Total Milestones

    3

  • Total Budget

    $50,000 USD

  • Last Updated

    13 Feb 2026

Milestone 1 - Dual-Language LSP with Atomspace Code Graph

Description

Build the TypeScript LSP server for Rholang and MeTTa powered by Hyperon inference over Atomspace code representations. This milestone delivers ONE thing: a production-quality LSP that maintains a live Atomspace graph and formulates all code intelligence as inference queries. The server provides completions, diagnostics, hover info, go-to-definition, and refactoring for both languages. Completions use Hyperon backward-chaining: given type context and bindings, what expressions satisfy constraints? Results are logically guaranteed well-typed. For Rholang: channel tracking through par compositions, name equivalence reasoning, linearity violation diagnostics. For MeTTa: native type system integration, pattern-aware completions, grounded atom support. The Atomspace code graph captures every definition, type, channel binding, import, and dependency as typed atoms. This is the foundation for M2 and M3. Implementation uses vscode-languageserver-node framework with Hyperon JavaScript bindings for direct Atomspace queries. Ships as VS Code extension.

Deliverables

1. TypeScript LSP server supporting Rholang and MeTTa with Hyperon JS bindings for inference queries against Atomspace code graphs. 2. Atomspace code graph builder parsing both languages into typed atom representations with definitions, types, channel bindings, imports, dependencies. 3. Type-directed completion engine formulating requests as Hyperon backward-chaining queries, returning well-typed suggestions with proof traces. 4. Diagnostic engine: Rholang channel linearity violations, unbound names, type mismatches; MeTTa undefined atoms, type errors, pattern match failures. 5. Hover information and go-to-definition via Atomspace graph traversal. 6. VS Code extension with install-and-go activation. 7. 200+ test cases covering both languages across all LSP features. 8. Architecture docs, Atomspace schema reference, setup guide.

Budget

$10,000 USD

Success Criterion

1. LSP starts in <3s, completions within 200ms for projects up to 50 files in both languages. 2. Type-directed completions achieve 95%+ type-correctness (every suggestion type-checks). 3. Diagnostics detect 90%+ of common errors in both languages (tested against corpus of 50+ files with injected errors). 4. Atomspace graph correctly represents 100% of top-level definitions for test corpus of 20+ contracts. 5. VS Code extension installs without errors on Windows, macOS, Linux. 6. All 200+ tests pass in CI. 7. Three independent Rholang/MeTTa developers validate suggestions are superior to existing tooling.

Milestone 2 - AI Coding Assistant & DevNet Deploy CLI

Description

Deliver Hyperon inference AI coding assistant and BlockDAG DevNet deployment CLI with CBC Casper finality. The AI assistant uses Atomspace reasoning for type-directed synthesis: developer describes contract, assistant decomposes into Hyperon backward-chaining sub-goals, constructs code with proof traces. Concept blending synthesizes novel patterns from Atomspace compositional reasoning. The DevNet CLI handles compile, deploy, verify, monitor for both languages on BlockDAG infrastructure. CBC Casper finality tracker reports deployment status in real-time. Singularity Compute integration estimates resources pre-deployment. Atomspace-informed dependency ordering and gas optimization. Self-extending feedback: accepted/rejected suggestions become evidence atoms for evolutionary optimization of inference strategies. Both components integrate with M1's shared Atomspace instance.

Deliverables

1. AI assistant with natural language interface decomposing specs into Hyperon sub-goals, producing code via type-directed synthesis with proof traces. 2. Concept blending engine synthesizing novel contract patterns from Atomspace composition with preservation proofs. 3. DevNet deployment CLI: compile, verify, deploy, monitor for Rholang and MeTTa on BlockDAG. 4. CBC Casper finality monitor tracking deployment transactions through consensus with real-time status. 5. Singularity Compute integration for resource estimation and gas optimization via Atomspace execution path analysis. 6. MeTTa-based verification pipeline blocking deployment of contracts failing spec checks. 7. Self-extending feedback capturing interactions as Atomspace evidence atoms with evolutionary optimization. 8. CLI docs with DevNet deployment tutorials.

Budget

$35,000 USD

Success Criterion

Deliver Hyperon inference AI coding assistant and BlockDAG DevNet deployment CLI with CBC Casper finality. The AI assistant uses Atomspace reasoning for type-directed synthesis: developer describes contract, assistant decomposes into Hyperon backward-chaining sub-goals, constructs code with proof traces. Concept blending synthesizes novel patterns from Atomspace compositional reasoning. The DevNet CLI handles compile, deploy, verify, monitor for both languages on BlockDAG infrastructure. CBC Casper finality tracker reports deployment status in real-time. Singularity Compute integration estimates resources pre-deployment. Atomspace-informed dependency ordering and gas optimization. Self-extending feedback: accepted/rejected suggestions become evidence atoms for evolutionary optimization of inference strategies. Both components integrate with M1's shared Atomspace instance.

Milestone 3 - Web Playground & Community Integration

Description

Deliver browser-based playground as thin wrapper around M1 LSP and complete ecosystem integration. Monaco editor connects to LSP server via WebSocket (no WASM compilation needed since LSP is TypeScript). Managed DevNet node for live contract deployment. 10+ guided tutorials using AI assistant from M2. Side-by-side comparison showing Hyperon inference suggestions with proof traces vs LLM confidence scores. Community Atomspace aggregation collecting anonymized patterns across users to improve inference ecosystem-wide. Cross-component integration ensuring LSP, AI assistant, CLI, and playground share Atomspace state. Performance optimization, documentation, video walkthroughs, onboarding materials.

Deliverables

1. Web playground with Monaco editor connected to LSP via WebSocket, full code intelligence for Rholang and MeTTa. 2. Managed DevNet connection for deploying contracts to live BlockDAG without local setup. 3. 10+ interactive tutorials: Rholang basics, MeTTa fundamentals, contract patterns, deployment workflows. 4. Side-by-side comparison mode: inference suggestions with proof traces vs LLM predictions with confidence scores. 5. Community Atomspace aggregation collecting anonymized developer patterns to improve inference. 6. Cross-component integration tests: LSP + AI + CLI + playground with shared Atomspace. 7. Performance: playground loads <5s, completions <500ms. 8. Launch docs, video walkthroughs, community onboarding.

Budget

$5,000 USD

Success Criterion

1. Playground loads and interactive within 5s on standard broadband, completions within 500ms. 2. New developers complete intro Rholang tutorial in 30min and deploy first contract, validated with 10+ testers. 3. Side-by-side comparison: 80%+ of test users correctly identify inference vs statistical suggestions. 4. Community Atomspace processes 50+ sessions and demonstrates measurable suggestion improvement. 5. All cross-component integration tests pass. 6. Playground supports 50+ concurrent users. 7. Docs receive positive feedback from 5+ community reviewers.

Join the Discussion (0)

Expert Ratings

Reviews & Ratings

    No Reviews Avaliable

    Check back later by refreshing the page.

Welcome to our website!

Nice to meet you! If you have any question about our services, feel free to contact us.