# GTCode

> AI systems consulting, machine-learning research notes, implementation guides, and production engineering resources from Ekewaka Lono.
This site publishes GTCode technical work: AI systems consulting, ML research notes, implementation guides, software projects, and technical articles from Ekewaka Lono.

- Author: Ekewaka Lono
- Contact: inquire@gtcode.com
- Full context for LLMs: https://gtcode.com/llms-full.txt
- Consulting: https://gtcode.com/consulting/
- Articles: https://gtcode.com/articles/
- Guides: https://gtcode.com/guides/
- Repositories: https://gtcode.com/repos/
- News wire: https://gtcode.com/news/

## Articles

- [Local MI Lab: How Controls Broke My First Induction Story](https://gtcode.com/articles/local-mi-lab-gpt2-small-induction-controls/): A first-person account of Local MI Lab, a GPT-2 small mechanistic interpretability practice loop where controls, head-specific patching, characterization, failure taxonomy, and metric calibration falsified an attractive induction-head story.
- [My First Real Mechanistic Interpretability Attempt: SELF-GROUND, Controls, and the Discipline of Negative Results](https://gtcode.com/articles/first-mechanistic-interpretability-attempt-self-ground/): A first-person account of SELF-GROUND, a negation SAE experiment that taught me why controls, calibration, and negative results matter.
- [The Bio-Digital Interface: Hybrid Consciousness Integration, the Simulacrum-Bliss Analysis, and the NEXUS Frontier](https://gtcode.com/articles/hybrid-consciousness-neural-digital-integration/): Design rationale and theoretical framework for hybrid consciousness integration experiments, covering the BARLI-QM framework, the Simulacrum-Bliss failure-mode analysis, the Q-CHIF safety design, the SCIN social scaling network, and the four-arm NEXUS frontier.
- [Grounded Chiral Tensor Synthesis: Likely Truth When the Record Is Incomplete](https://gtcode.com/articles/grounded-chiral-tensor-synthesis/): A public introduction to CNS 7.1 / GCTS, an access-aware research architecture for ranking likely truth under contradictory, incomplete, and adversarial evidence.
- [Harness Engineering: From Agent Prompts to Engineering Control Systems](https://gtcode.com/articles/harness-engineering/): A practical technical synthesis on harness engineering: repository knowledge, executable invariants, runtime evidence, architecture compression, and safety controls for reliable AI coding agents.
- [The Cinema Debugger: Reinventing Code Review Through Semantic Graph Navigation and Multi-Resolution Intelligence](https://gtcode.com/articles/cinema-debugger-semantic-code-review/): A technical deep-dive into the Cinema Debugger paradigm and Context-Graph Knowledge Base architecture, exploring how semantic graphs, scene-based workflows, and multi-resolution navigation are transforming code review from text-diff scrolling to intelligent knowledge traversal.
- [The Mechanistic Interpretability Landscape: A Technical Survey of How We Are Learning to Understand Large Language Models](https://gtcode.com/articles/mechanistic-interpretability-landscape-2025/): A comprehensive technical survey of mechanistic interpretability research in 2025, covering control mechanisms, truth detection, safety implications, and fundamental limitations of understanding large language models.
- [The Global Cognitive Brain: Engineering Distributed Machine Intelligence as a Living System](https://gtcode.com/articles/global-cognitive-brain-architecture/): A technical introduction to the Global Cognitive Brain architecture, exploring how distributed AI agents can form coherent cognitive organisms through semantic primitives, cluster DNA, and bio-inspired coordination protocols.
- [A Visual Language for Thinking About AI Synthesis](https://gtcode.com/articles/visual-language-ai-synthesis/): Exploring a visual framework for understanding how AI systems might transform conflicting information into coherent knowledge—through resonance-like alignment of representations.
- [Epistemic Fragmentation in Multi-Model AI Systems: The Challenge of Synthesizing Divergent Knowledge Claims](https://gtcode.com/articles/epistemic-fragmentation-multi-model-ai-systems/): An analysis of epistemic fragmentation in multi-model AI deployments, exploring why different models produce conflicting knowledge claims and the institutional risks this creates for organizations relying on AI-generated insights.


## Research Papers

- [Physics-Informed Spatiotemporal Digests: WSTD, OMR-Gated RF Verification, and Residual Security under Learned-Model Attackers](https://gtcode.com/papers/maxwell-bound-spatiotemporal-digests-paper/): A research design note proposing Physics-Informed Spatiotemporal Digests, combining a windowed state-and-transition digest with optional OMR-gated RF verification.
- [Geometric Semantic Communication for Emergent Multi-Agent Intelligence](https://gtcode.com/papers/geometric-semantic-communication/): A research agenda toward communication infrastructure for emergent artificial general intelligence, grounded in tensor logic and topological structures for truth, perspective, and narrative.
- [Geometric Publish-Subscribe: Content-Based Routing in Embedding Space for Multi-Agent Systems](https://gtcode.com/papers/geometric-publish-subscribe/): A research proposal for GPS, a communication protocol for multi-agent systems where messages route by semantic similarity rather than topic strings.


## Guides

- [CNS 2.0 Research Roadmap: A Multi-Year Vision](https://gtcode.com/guides/cns-2.0-research-roadmap/): Detailing the comprehensive, multi-faceted research program to develop, validate, and responsibly deploy the Chiral Narrative Synthesis framework.
- [Tutorials](https://gtcode.com/guides/tutorials/): Practical, step-by-step walkthroughs of key CNS 2.0 applications and experiments.
- [Case Studies & Experiments](https://gtcode.com/guides/case-studies-and-experiments/): Exploring the theoretical foundations and empirical validation of the CNS framework through in-depth case studies.
- [CNS 7.1 / GCTS: Grounded Chiral Tensor Synthesis](https://gtcode.com/guides/cns-gcts/): The current CNS research line: access-aware likely-truth ranking over structured possible worlds.
- [CNS 8.0 / Grounded Dialectical Orthesis](https://gtcode.com/guides/cns/): CNS 8.0 guide set: SNO-centered dialectical synthesis, tensor closure, predicate invention, orthesis, grounding, access, and audit resources.
- [Building CNS 2.0: A Developer's Guide](https://gtcode.com/guides/building-cns-2.0-developers-guide/): A progressive educational guide to implementing Chiral Narrative Synthesis 2.0 in Python


## Open Source Libraries

- [Hex.pm Profile](https://hex.pm/users/nshkrdotcom): Complete Elixir package catalog
- [All Repositories](https://gtcode.com/repos/): Full library listing

- [Agent Session Manager](https://hex.pm/packages/agent_session_manager): Lean OTP-correct multi-provider CLI session runtime (ASM).
- [Aitrace](https://hex.pm/packages/aitrace): The unified observability layer for the AI Control Plane, delivering full-fidelity tracing for AI agent reasoning, tool calls, and state transitions.
- [Altar](https://hex.pm/packages/altar): Altar provides a robust, type-safe foundation for building AI agent tools in Elixir. It offers a clean contract to define and execute tools locally, with a clear promotion path to…
- [Altar Ai](https://hex.pm/packages/altar_ai): Protocol-based AI adapter foundation for Elixir. Provides unified abstractions for gemini_ex, claude_agent_sdk, and codex_sdk with automatic fallback support, runtime capability…
- [Amp Sdk](https://hex.pm/packages/amp_sdk): An Elixir SDK for the Amp CLI - programmatic access to Amp's AI coding agent.
- [Anvil Ex](https://hex.pm/packages/anvil_ex): Labeling queue library for managing human labeling workflows. Domain-agnostic HITL (human-in-the-loop) data annotation with inter-rater reliability metrics (Cohen's kappa, Fleiss'…
- [Arsenal](https://hex.pm/packages/arsenal): A metaprogramming framework for building REST APIs from OTP operations with automatic endpoint generation, parameter validation, and OpenAPI documentation.
- [Assessor](https://hex.pm/packages/assessor): The definitive CI/CD platform for AI Quality.
- [Blitz](https://hex.pm/packages/blitz): Parallel command runner and Mix workspace orchestrator for Elixir tooling
- [Chz Ex](https://hex.pm/packages/chz_ex): Configuration management with CLI parsing for Elixir
- [Citadel](https://hex.pm/packages/citadel): The command and control layer for the AI-powered enterprise.
- [Claude Agent Sdk](https://hex.pm/packages/claude_agent_sdk): An Elixir SDK for Claude Code - build AI-powered CLI tools with Claude.
- [Claude Code Sdk](https://hex.pm/packages/claude_code_sdk): An Elixir SDK for Claude Code - Build AI-powered CLI tools with Claude
- [Cli Subprocess Core](https://hex.pm/packages/cli_subprocess_core): Shared CLI subprocess runtime foundation with first-party common provider profiles.
- [Codex Sdk](https://hex.pm/packages/codex_sdk): Idiomatic Elixir SDK for OpenAI's Codex agent. Provides a complete, production-ready interface with streaming support, comprehensive event handling, and robust testing utilities.
- [Command](https://hex.pm/packages/command): Command center core library for AI agent orchestration
- [Coolify Ex](https://hex.pm/packages/coolify_ex): Generic Elixir toolkit for deploying, observing, and readiness-verifying Coolify apps
- [Crucible Adversary](https://hex.pm/packages/crucible_adversary): Adversarial testing and robustness framework for AI models with 25 attacks (character/word/semantic perturbations, prompt injection, jailbreak, extraction, inversion), defenses…
- [Crucible Bench](https://hex.pm/packages/crucible_bench): Comprehensive benchmarking framework for AI research. Measures latency, throughput, cost, and reliability with percentile analysis and Nx numerical computations.
- [Crucible Datasets](https://hex.pm/packages/crucible_datasets): Dataset management and caching for AI research benchmarks
- [Crucible Deployment](https://hex.pm/packages/crucible_deployment): Model deployment orchestration with health checking and rollback
- [Crucible Ensemble](https://hex.pm/packages/crucible_ensemble): Multi-model ensemble prediction with voting strategies for AI reliability. Leverages BEAM parallelism for massively concurrent LLM queries.
- [Crucible Feedback](https://hex.pm/packages/crucible_feedback): Feedback collection, drift detection, and active learning for ML pipelines
- [Crucible Framework](https://hex.pm/packages/crucible_framework): CrucibleFramework: A thin orchestration layer for experiment pipelines. Provides pipeline execution, stage behaviour, and optional persistence.
- [Crucible Harness](https://hex.pm/packages/crucible_harness): Experimental research framework for running AI benchmarks at scale. Provides orchestration, streaming processing with Flow/GenStage, and statistical analysis.
- [Crucible Hedging](https://hex.pm/packages/crucible_hedging): Request hedging for tail latency reduction in distributed systems. Implements Google's 'Tail at Scale' with adaptive strategies. Reduces P99 latency by 75-96%.
- [Crucible Ir](https://hex.pm/packages/crucible_ir): Intermediate Representation for the Crucible ML reliability ecosystem
- [Crucible Model Registry](https://hex.pm/packages/crucible_model_registry): Model versioning, artifact storage, and lineage tracking for ML pipelines
- [Crucible Telemetry](https://hex.pm/packages/crucible_telemetry): Advanced telemetry collection and analysis for AI research
- [Crucible Trace](https://hex.pm/packages/crucible_trace): Structured causal reasoning chain logging for LLMs. Captures decision-making processes with cryptographic verification for transparency and debugging.
- [Crucible Train](https://hex.pm/packages/crucible_train): Unified ML training infrastructure for Elixir/BEAM
- [Crucible Xai](https://hex.pm/packages/crucible_xai): Explainable AI (XAI) tools for the Crucible framework. Includes LIME implementations, SHAP-like explanations, feature attribution, and model interpretability for local and global…
- [Datasets Ex](https://hex.pm/packages/datasets_ex): Dataset management library for ML experiments with support for GSM8K, HumanEval, MMLU loaders and evaluation metrics (BLEU, ROUGE, F1).
- [Dspex](https://hex.pm/packages/dspex): DSPy for Elixir via SnakeBridge - Declarative LLM programming
- [Duckdb Ex](https://hex.pm/packages/duckdb_ex): A 100% faithful port of the DuckDB Python client to Elixir, using the DuckDB CLI for full API compatibility with the official Python client.
- [Elixir Dashboard](https://hex.pm/packages/elixir_dashboard): A Phoenix LiveView performance monitoring dashboard for tracking slow endpoints and database queries during development.
- [Elixir Scope](https://hex.pm/packages/elixir_scope): ElixirScope is a next-generation debugging and observability platform for Elixir applications, designed to provide an Execution Cinema experience through deep, compile-time AST…
- [Elixir Tracer](https://hex.pm/packages/elixir_tracer): Local-first observability for Elixir with New Relic API parity. A complete tracing and observability solution that works entirely offline while maintaining compatibility with New…
- [Embed Ex](https://hex.pm/packages/embed_ex): Vector embeddings service for the NSAI ecosystem
- [Eval Ex](https://hex.pm/packages/eval_ex): Model evaluation harness for standardized benchmarking with semantic similarity, exact match, and custom metrics.
- [Ex Data Check](https://hex.pm/packages/ex_data_check): Production-ready data validation and quality library for Elixir ML pipelines. Provides 22 built-in expectations, drift detection, advanced profiling with outliers and…
- [Ex Fairness](https://hex.pm/packages/ex_fairness): Fairness and bias detection library for Elixir AI/ML systems. Provides comprehensive fairness metrics, bias detection algorithms, and mitigation techniques.
- [Ex Topology](https://hex.pm/packages/ex_topology): Generic Topological Data Analysis for Elixir
- [Exdantic](https://hex.pm/packages/exdantic): Advanced schema definition and validation library for Elixir
- [External Runtime Transport](https://hex.pm/packages/external_runtime_transport): Shared execution-surface substrate for local subprocess, SSH exec, and guest-bridge placement.
- [Flowstone](https://hex.pm/packages/flowstone): Asset-first data orchestration framework for Elixir/BEAM.
- [Flowstone Ai](https://hex.pm/packages/flowstone_ai): FlowStone integration for altar_ai - AI-powered data pipeline assets. Provides FlowStone.AI.Resource for unified AI access and FlowStone.AI.Assets DSL helpers (classify_each,…
- [Forge Ex](https://hex.pm/packages/forge_ex): Sample factory library for generating, transforming, and computing measurements on arbitrary samples. Domain-agnostic data pipeline orchestration for ML dataset preparation.
- [Foundation](https://hex.pm/packages/foundation): Foundation 0.2.x is a complete rewrite (not compatible with 0.1.x): lightweight resilience primitives for backoff, retry, rate-limit windows, circuit breakers, semaphores, and…
- [Gemini Cli Sdk](https://hex.pm/packages/gemini_cli_sdk): An Elixir SDK for the Gemini CLI - build AI-powered applications with Google Gemini.
- [Gemini Ex](https://hex.pm/packages/gemini_ex): Comprehensive Elixir client for Google's Gemini AI API with dual authentication, embeddings with MRL, streaming, type safety, and built-in telemetry for production applications.
- [Gepa Ex](https://hex.pm/packages/gepa_ex): Elixir implementation of the GEPA (Genetic-Pareto) optimizer that combines LLM-powered reflection with Pareto search to evolve text-based system components.
- [Github Ex](https://hex.pm/packages/github_ex): Native Elixir SDK for the GitHub REST API with generated endpoint modules, OTP-friendly runtime defaults, OAuth helpers, GitHub App helpers, pagination utilities, and publication-…
- [Hf Datasets Ex](https://hex.pm/packages/hf_datasets_ex): HuggingFace Datasets for Elixir - Load, stream, and process ML datasets from the HuggingFace Hub with native BEAM/OTP integration.
- [Hf Hub](https://hex.pm/packages/hf_hub): Elixir client for HuggingFace Hub—dataset/model metadata, file downloads, caching, and authentication
- [Ingot](https://hex.pm/packages/ingot): Phoenix LiveView interface for sample generation and human labeling workflows. Thin wrapper around Forge (sample factory) and Anvil (labeling queue) with real-time progress…
- [Json Remedy](https://hex.pm/packages/json_remedy): A blazingly fast Elixir library for repairing malformed JSON using binary pattern matching. Handles LLM outputs, legacy data, and broken JSON with intelligent context-aware fixes.
- [Jules Ex](https://hex.pm/packages/jules_ex): Elixir client library for the Jules API. Jules is an AI-powered coding assistant that can automate software development tasks across multiple repositories.
- [Labeling Ir](https://hex.pm/packages/labeling_ir): Shared IR structs for Forge, Anvil, Ingot, and external clients; typed datasets, samples, assignments, labels, artifacts, and evaluation runs for labeling workflows.
- [Lineage Ir](https://hex.pm/packages/lineage_ir): Lineage IR for cross-system traces, spans, artifacts, and provenance edges. Provides a shared event envelope and sink interface for consolidation across runtimes.
- [Linear Sdk](https://hex.pm/packages/linear_sdk): Elixir SDK for Linear with committed upstream schema artifacts and generated GraphQL API reference docs.
- [Llama Cpp Sdk](https://hex.pm/packages/llama_cpp_sdk): First concrete self-hosted inference backend for llama.cpp, owning llama-server boot specs, readiness probes, stop semantics, and endpoint publication through…
- [Llm Guard](https://hex.pm/packages/llm_guard): AI Firewall and guardrails for LLM-based Elixir applications. Provides prompt injection detection, data leakage prevention, jailbreak detection, and comprehensive security…
- [Mabeam](https://hex.pm/packages/mabeam): A multi-agent framework for the BEAM (Erlang VM) that provides agent lifecycle management, message passing, discovery, and extensibility for building distributed agent-based…
- [Metrics Ex](https://hex.pm/packages/metrics_ex): Metrics aggregation service for experiment results and system health monitoring.
- [Multipart Ex](https://hex.pm/packages/multipart_ex): Client-agnostic multipart/form-data builder with explicit file handling and streaming encoding. Safe, predictable file inputs and configurable serialization strategies for complex…
- [Notion Sdk](https://hex.pm/packages/notion_sdk): Elixir SDK for the Notion API, built on Pristine's hexagonal architecture. Ported from the official notion-sdk-js.
- [Nsai Gateway](https://hex.pm/packages/nsai_gateway): Unified API Gateway for North Shore AI services - handles routing, authentication, rate limiting, and circuit breaking
- [Nsai Registry](https://hex.pm/packages/nsai_registry): North Shore AI Registry: Distributed model registry with PubSub coordination, health monitoring, and multi-backend storage for ML reliability research.
- [Nsai Work](https://hex.pm/packages/nsai_work): NsaiWork - Unified job scheduler for the North-Shore-AI platform. Protocol-first, multi-tenant job scheduling with priority queues, resource-aware scheduling, and pluggable…
- [Nx Penalties](https://hex.pm/packages/nx_penalties): Composable regularization penalties and loss functions for the Nx ecosystem
- [Ollixir](https://hex.pm/packages/ollixir): Ollixir is a complete Elixir client library for the Ollama API.
- [Perimeter](https://hex.pm/packages/perimeter): A typing system for Elixir/OTP.
- [Pilot](https://hex.pm/packages/pilot): Interactive CLI/REPL for the NSAI ecosystem: unified interface for Crucible experiments, CNS dialectical synthesis, and ML reliability research.
- [Pipeline Ex](https://hex.pm/packages/pipeline_ex): AI pipeline orchestration library for Elixir. Chain Claude and Gemini APIs with robust execution, fault tolerance, and self-improving Genesis pipelines.
- [Playwriter](https://hex.pm/packages/playwriter): Elixir browser automation with WSL-to-Windows support. Control visible Windows browsers from WSL.
- [Portfolio Coder](https://hex.pm/packages/portfolio_coder): Code Intelligence Platform built on the Portfolio RAG ecosystem. Repository analysis, semantic code search, dependency graphs, and AI-powered code understanding with multi-…
- [Portfolio Core](https://hex.pm/packages/portfolio_core): Hexagonal architecture core for building flexible RAG systems in Elixir.
- [Portfolio Index](https://hex.pm/packages/portfolio_index): Production adapters and pipelines for PortfolioCore. Vector stores, graph stores, embedders, Broadway pipelines, and advanced RAG strategies.
- [Portfolio Manager](https://hex.pm/packages/portfolio_manager): Application layer for Portfolio ecosystem providing RAG workflows, graph analysis, domain registries, and CLI tools built on portfolio_core and portfolio_index.
- [Prismatic](https://hex.pm/packages/prismatic): Shared GraphQL runtime for Prismatic-based Elixir SDKs.
- [Pristine](https://hex.pm/packages/pristine): Shared runtime substrate for first-party OpenAPI-based Elixir SDKs.
- [Prompt Runner Sdk](https://hex.pm/packages/prompt_runner_sdk): Prompt Runner SDK - Convention-driven and legacy-config prompt orchestration for Elixir, Mix, and production applications with streaming output, provider abstractions, runtime…
- [Rag Ex](https://hex.pm/packages/rag_ex): A library to make building performant RAG systems in Elixir easy
- [Self Hosted Inference Core](https://hex.pm/packages/self_hosted_inference_core): Service-runtime kernel for self-hosted inference backends, owning readiness, health, lease reuse, and endpoint publication above the transport seam.
- [Sinter](https://hex.pm/packages/sinter): Unified schema definition, validation, and JSON generation for Elixir. Sinter is a focused, high-performance schema validation library designed specifically for dynamic frameworks…
- [Skill Ex](https://hex.pm/packages/skill_ex): Aggregates and packages Claude skills from multiple Elixir projects into a single, validated bundle ready for distribution.
- [Slither](https://hex.pm/packages/slither): Low-level BEAM↔Python concurrency substrate. ETS-backed shared state with Python views, batched fan-out with real backpressure, and stage composition over BEAM + Python steps.…
- [Snakebridge](https://hex.pm/packages/snakebridge): Compile-time generator for type-safe Elixir bindings to Python libraries
- [Snakepit](https://hex.pm/packages/snakepit): High-performance pooler and session manager for external language integrations. Supports Python, Node.js, Ruby, and more with gRPC streaming, session management, and production-…
- [Supertester](https://hex.pm/packages/supertester): Battle-hardened OTP testing toolkit with chaos engineering, performance testing, and zero-sleep synchronization patterns for building robust Elixir applications.
- [Synapse](https://hex.pm/packages/synapse): Declarative, headless multi-agent runtime for code review orchestration with signals, workflows, and Postgres persistence.
- [Synapse Ai](https://hex.pm/packages/synapse_ai): Synapse integration for altar_ai - SDK-backed LLM providers for multi-agent workflows. Provides unified adapter layer for Gemini, Claude, and Codex with automatic fallback,…
- [Telemetry Reporter](https://hex.pm/packages/telemetry_reporter): TelemetryReporter is a transport-agnostic telemetry batching library for Elixir/BEAM apps. It uses Pachka for efficient size/time batch flushing, drops on overload to protect…
- [Tiktoken Ex](https://hex.pm/packages/tiktoken_ex): Pure-Elixir TikToken-style byte-level BPE tokenizer, with helpers for Kimi K2 encodings.
- [Tinkex](https://hex.pm/packages/tinkex): Elixir SDK for Tinker: LoRA training, sampling, and future-based workflows with telemetry and HTTP/2.
- [Vllm](https://hex.pm/packages/vllm): vLLM for Elixir via SnakeBridge - Easy, fast, and cheap LLM serving for everyone. High-throughput LLM inference with PagedAttention, continuous batching, and OpenAI-compatible API.
- [Weaviate Ex](https://hex.pm/packages/weaviate_ex): A modern Elixir client for Weaviate vector database with support for collections, objects, batch operations, GraphQL queries, and vector search. Includes health checks and…
- [Weld](https://hex.pm/packages/weld): Deterministic Hex package projection for Elixir monorepos

## Consulting

- [AI Consulting | GTCode](https://gtcode.com/consulting/): AI consulting and engineering services from GTCode: strategy, architecture, delivery, and reliability for production-grade AI systems.

## Policies

- [AI Ethics Guidelines](https://gtcode.com/policies/ai/): General AI access and usage guidelines for GTCode.com
- [AI Agent Policy](https://gtcode.com/policies/agent/): Guidelines and instructions for AI agents accessing GTCode.com
- [LLM Usage Policy](https://gtcode.com/policies/llm/): Comprehensive guidelines for Large Language Models accessing GTCode.com
- [Robots.txt - Web Crawler Policy](https://gtcode.com/policies/robots/): Robots.txt configuration for web crawlers and search engines
- [Humans Behind GTCode](https://gtcode.com/policies/humans/): Credits and information about the people behind GTCode.com