Independent AI engineer building unconventional developer tools, agent systems, research software

github

Daniele Cangi • Peerlist

<h3>A surreal 16-bit political satire RPG built with Godot.</h3>
Civic Nightmare is a satirical, top-down RPG built in Godot 4.6 that explores the absurdity of modern global governance, corporate dominance, and the digital age. You play as a scrawny, debt-burdened citizen on a desperate quest to renew a passport, crossing high-detail districts where pixel-scale citizens, monumental political caricature, arcade confrontation, and CRT bureaucracy occupy the same collapsing reality.
<h4>🕹️ Game Features</h4>
<h3>🚗 The Administrative Approach</h3>
The game begins with a news bulletin and a mandatory drive through a surreal, bureaucratic landscape. Players must navigate obstacles and challenges that highlight the absurdities of modern systems. The free-roam gameplay begins with a persistent overworld theme.
<h3>🏛️ The Great Signature Quest</h3>
Players confront six different authorities, each with unique challenges and playable procedures:
<ul>
<li>Donald Trump: Focuses on patriotism and access.</li>
<li>Elon Musk: Features unfinished infrastructure and progress.</li>
<li>Ursula von der Leyen: A maze of instructions.</li>
<li>Vladimir Putin: A ceremonial palace with a war context.</li>
<li>Christine Lagarde: A stability vault protecting confidence.</li>
<li>Emmanuel Macron: Managed decline presented as elegance.</li>
</ul>
These encounters involve mini-games like blackjack, bureaucracy simulators, and arcade-style combat.
<h3>🛸 Optional Investigations, Deviations &amp; Anomalies</h3>
The game includes optional content such as:
<ul>
<li>Historical Contamination: A parody of historical figures.</li>
<li>The Southern Sanctuary: Featuring Sam Altman and AI.</li>
<li>Intercepted Channels: Interactions with global leaders.</li>
<li>The Quantum UFO: A surreal encounter with temporal echoes.</li>
<li>The Hidden Bunker: A high-stakes encounter with Zelensky and Death.</li>
<li>The Drone Escalation: A cybernetic battle with Bezos.</li>
</ul>
<h3>🤖 C.L.A.U.D.I.A. Assistant</h3>
A satirical 90s digital mascot provides cynical guidance throughout the game.
<h3>💾 Persistent Dossier</h3>
Player progress and encounter consequences are saved in a versioned local dossier. The game offers 'Continue' and 'New Game' options, with an 'Administrative Hold' feature for pausing gameplay.
<h3>🗂️ Office of Final Determinations</h3>
After all signatures are collected, a final processing office presents an interpretation of the player's choices. Web persistence relies on IndexedDB, with potential limitations in private browsing.
<h4>🎞️ Civic Nightmare 2.2</h4>
The game features runtime art loaded directly, including:
<ul>
<li>Actual gameplay: Visuals for key encounters and authorities.</li>
<li>Runtime art gallery: Showcasing various scenes and character art.</li>
</ul>
<h4>🎨 Technical &amp; Aesthetic Identity</h4>
<ul>
<li>Hybrid Civic-Surreal Style: Combines pixel art with detailed environments.</li>
<li>Portrait Pipeline: Automated processing for character portraits.</li>
<li>Authority Architecture: Consistent visual style for authority encounters.</li>
<li>Narrative Interiors: Unique room designs for different authorities.</li>
<li>VHS/CRT Effects: Custom shaders for a retro aesthetic.</li>
</ul>
<h4>🤝 Contributing</h4>
Contributions are welcome, including bug fixes, art, dialogue, and technical improvements. The project follows contribution guidelines and uses a MIT license for code and a separate notice for media and narrative content.
<h4>🚀 Play &amp; Run Locally</h4>
Requires Godot Engine 4.6. Players can clone the repository, open the project, and run the main scene. A web version is also available.
<h4>🧩 Project Structure</h4>
The project is organized into scripts for managers, data, assets, and scenes, with a clear separation of concerns for reusable components and game logic.
<h4>✅ Verification</h4>
Automated tests, including parsing, smoke tests, and web export, are run before deployment.
<h4>⚖️ License</h4>
Source code is under the MIT license. Media and narrative content have separate rights and contribution policies.

civic-nightmare

<h3>Discover Your Perfect Living Situation in Denmark</h3>Roomee is an innovative mobile application designed to simplify the process of finding a room or a roommate in Denmark. Leveraging a user-friendly, swipe-and-match interface similar to popular dating apps, Roomee connects individuals seeking accommodation with those who have rooms to rent, and vice versa.<ul><li>Two-Sided Matching: Whether you're looking for a place to live or a compatible housemate, Roomee ensures meaningful connections by matching both parties' needs.</li><li>Verified Profiles: To ensure a safe and trustworthy experience, Roomee implements a verification process for its users, helping to prevent scams and create a secure environment.</li><li>Danish Market Focus: Tailored specifically for the Danish housing market, Roomee features Danish cities, currency, and is available in multiple languages, with a current focus on Denmark.</li><li>Intuitive Interface: Navigate through available rooms and potential roommates with a simple swipe. Like what you see? Swipe right to express interest.</li><li>Instant Chat: Once a mutual match is made, users can immediately start chatting within the app to discuss details and arrange viewings.</li><li>Map View &amp; Filters: Explore available rooms on an interactive map and utilize smart filters for price, housing type, and location to narrow down your search.</li><li>Save Favorites: Easily save rooms or profiles you're interested in for later review.</li></ul>Roomee aims to transform the shared living experience into a positive and seamless one. The app is currently in beta testing, with plans for a full launch on the App Store and Google Play soon. Join the beta program to experience Roomee before its official release.

Roomee

<h3>About Alchimista</h3>Alchimista is a self-hostable AI governance, document intelligence, RAG, audit, and privacy infrastructure layer. It ingests evidence, builds a tenant-scoped retrieval index, returns mandatory citations, and records AI decision evidence with factual privacy metadata. The baseline needs no GCP project, Auth0 tenant, external vector database, object store, or paid service. GCP, Vertex, Pub/Sub, GCS, Cloud Run, and OIDC remain optional adapters. Alchimista implements technical controls. It does not certify GDPR, AI Act, or other legal compliance, and its PII detectors are not guaranteed to find every sensitive value.<h3>Key Features</h3><ul><li>Self-hosted AI Governance: Manage AI deployments with a focus on governance, auditability, and privacy.</li><li>Document Intelligence: Extract insights and information from documents.</li><li>RAG (Retrieval-Augmented Generation): Integrate external knowledge bases for more informed AI responses.</li><li>Privacy and Auditable Evidence: Ensure data privacy and maintain auditable records of AI decisions.</li><li>Local-First Deployment: Designed for deployment within your own infrastructure, minimizing reliance on external services.</li><li>Optional Full Rizzo PII Detection: Advanced detection capabilities for Personally Identifiable Information.</li><li>Flexible Integration: Supports optional adapters for GCP services like Vertex AI, Pub/Sub, GCS, and Cloud Run, as well as OIDC for authentication.</li><li>Secure Data Handling: Implements technical controls for privacy and security, including encrypted vault for PII.</li></ul><h3>Quickstart Guide</h3>Prerequisites: Docker Engine with Compose v2 and Python 3.11 or newer.Follow these steps for a self-hosted deployment:<ol><li>Clone the repository: git clone https://github.com/Daniele-Cangi/Alchimista.git</li><li>Navigate to the directory: cd Alchimista</li><li>Initialize local environment: python scripts/init_local_env.py</li><li>Start the services: docker compose up --detach --build --wait</li></ol>Access Alchimista at http://127.0.0.1:8000. The application provides integrated workspaces for Home, Documents, Ask, Privacy, Audit, Governance, and System management.<h3>Technical Details</h3><ul><li>Database: PostgreSQL</li><li>Containerization: Docker Compose</li><li>Programming Language: Python</li><li>APIs: RESTful APIs for ingestion, RAG queries, and privacy management.</li><li>Authentication: Supports local token-based authentication and OIDC for enterprise integration.</li></ul>Alchimista offers a robust solution for managing AI systems with a strong emphasis on privacy, auditability, and self-hosting capabilities.

Alchimista

<h3>About Butterfly</h3>Butterfly is an interactive 3D story world where a deterministic engine verifies ways to keep a chosen moment after conditions change. It allows users to explore a city as an interface, where a deterministic TypeScript engine dictates events. The diorama, timeline, MomentFrame, and causal inspector all present the simulation's results.<h3>Key Features:</h3><ul><li>Interactive 3D Story World: Explore a dynamic 3D environment where changes in conditions lead to different outcomes.</li><li>Deterministic Engine: A core engine ensures predictable and verifiable simulation results.</li><li>Causal Inspector: Understand the cause-and-effect relationships within the simulation.</li><li>MomentFrame: Preserve specific moments and explore alternative scenarios to achieve them.</li><li>Time-Aware Routes: Features explicit endpoints, transport modes, durations, and availability windows for routes.</li><li>Local Development: Run the application locally with Node.js, using a local sample without requiring API keys or network connections for the core journey.</li><li>Sanity Integration: Integrates with Sanity for content management, allowing for live content updates and authoring.</li><li>Testing Frameworks: Includes Playwright for end-to-end browser testing and Vitest for unit testing.</li></ul>The project is built with modern web technologies including React, Next.js, and Three.js, providing a rich visual and interactive experience. It emphasizes a clear separation of concerns with distinct modules for the app, world, engine, scene, UI, and Sanity integration.

Butterfly

<h3>WorldLoop - Procedural Roblox World Generator</h3>WorldLoop is a powerful procedural generation tool designed specifically for Roblox creators. It streamlines the process of building complex and playable worlds within Roblox Studio by offering a browser-based workflow that generates architecture, terrain, caves, and water features. Powered by the OMNI Engine, WorldLoop composes unique architectural elements rather than relying on static templates, ensuring limitless creative possibilities.<ul><li>Procedural Architecture: Generates real CAD parts for buildings, houses, towers, and temples, ensuring accurate measurements and playable interiors with rooms, stairs, doors, and openings.</li><li>Voxel Worlds: Creates dynamic terrain, water bodies, and caves that can be composed around the generated architecture, bringing your worlds to life.</li><li>Studio Sync: Seamlessly integrates with Roblox Studio via the Creator Store connector, allowing you to send generated assets directly into your projects.</li><li>Live 3D Preview: Inspect and refine your generated worlds in real-time before delivery, ensuring the final result meets your vision.</li><li>AI-Assisted Workflows (MCP Beta): Connect a local AI agent to inspect builds, use the CAD catalog, and receive validation reports for enhanced creative control.</li></ul>The intuitive workflow allows creators to choose a preset, generate the world, preview the results, and send them directly to Roblox Studio, significantly reducing development time and effort.

WorldLoop - Procedural Roblox World Generator

<h3>About Derivative and Forge</h3>Derivative is a computational reasoning substrate, and Forge is its software-building pipeline. This system transforms natural-language requirements into executable Python software. It runs the generated code in isolation, validates it independently, and only packages builds that pass all validation gates. The process involves turning a requirement into a typed contract, then a grounded plan, followed by code generation, isolated execution, and independent validation, ultimately leading to a verified package or explicit failure evidence.<h3>Key Features:</h3><ul><li>Automated Software Generation: Builds software from natural-language requirements.</li><li>Isolated Execution: Runs generated code in a secure, isolated environment (Docker sandbox) to ensure safety and reproducibility.</li><li>Independent Validation: Implements multiple layers of validation to ensure the generated code meets specified contracts and quality standards.</li><li>Fail-Closed Architecture: Designed to fail securely, preventing the packaging of unverified or invalid builds.</li><li>Traceability: Maintains traceability of requirements throughout the build process, from source text to final artifact.</li><li>Quality Contracts: Enforces executable obligations for security, persistence, rate limiting, auditability, and observability.</li><li>Bounded Repair: Implements a mechanism for targeted repair of failures, requiring material artifact changes before revalidation.</li><li>Extensibility: Supports various reasoning capabilities through optional installation profiles, including symbolic, probabilistic, causal, quantum, and physical reasoning.</li><li>Evidence Generation: Produces typed evidence for each stage of the build process, including build specifications, plans, code artifacts, and validation results.</li></ul>The system is built with Python and leverages Docker for its sandboxing capabilities. It is designed for developers seeking a robust and trustworthy way to automate software creation and validation.

Derivative

<h3>About XCP</h3>XCP is an open-source development stack designed for programmable, verifiable execution on Xbox Series X Developer Mode. It enables developers to explore, reproduce, and contribute to a Windows-to-Xbox development workflow, turning measured results into independently verified execution.<ul><li>AI builds: Leverage AI for the creation and development process.</li><li>Xbox executes: Facilitates execution on Xbox Series X Developer Mode.</li><li>XCP verifies: Ensures the integrity and verifiability of the execution.</li></ul>The project is open source under the Apache 2.0 license. You can explore the source code on GitHub and contribute to its development.<h3>Getting Started with XCP</h3>XCP offers multiple paths to get started:<ol><li>Any Development Machine: Run the portable core, including XVM, project contracts, conformance vectors, and the Python test suite without needing an Xbox. Start by cloning the repository.</li><li>Windows: Rebuild the entire development stack, including Build Studio, the worker, and supporting components from source. Compare the produced distribution with its verification manifests by following the build guide.</li><li>Xbox Series X: Reproduce the development on hardware by testing newly built development packages on your own console in Developer Mode and reporting new, independently identified results. Follow the device protocol.</li></ol>The first run does not require an Xbox. You can try XCP on your computer with a supported Python development environment to run the portable tests and conformance suite, then verify and execute a bundled XVM vector.<h3>XCP Studio</h3>XCP Studio provides a unified workspace for creation, source adaptation, project evolution, Xbox playtesting, and structured evidence collection. It offers three interfaces:<ul><li>Guided: Utilizes visual tools over the declared project contract.</li><li>Expert: Provides detailed access to the authoritative project.</li><li>Agent: Integrates external AI through the same portable toolchain.</li></ul>The project maintains one authoritative source, integrity-checked on save, with three interfaces to access it.<h3>The XCP Lifecycle</h3>XCP extends beyond initial generation, encompassing validation, execution, observation, correction, versioning, and rollback on real hardware. The lifecycle ensures that what the console reports becomes input for the next correction, enabling agents to act without human translation.<ul><li>Idea or repository: Start with an idea or an existing codebase.</li><li>AI agent + XCP Studio: Utilize AI and the Studio for development.</li><li>Editable XCP project: Work with an editable project structure.</li><li>Deterministic build: Ensure reproducible builds.</li><li>Real Xbox execution: Execute on actual Xbox hardware.</li><li>Observation + evidence: Collect and analyze execution data.</li><li>Correct · adapt · evolve: Iterate and improve the project.</li></ul>Whether created from an idea or adapted from existing source (like Godot projects), XCP projects follow a repeatable, observable, and verifiable lifecycle. The platform emphasizes measured, not implied, proof, with evidence generated during execution.Key Features:<ul><li>Complete blind-agent lifecycle operations.</li><li>Real-source adaptation operations.</li><li>Exact update and rollback capabilities.</li><li>Structured correction recorded as it happens (Ledgers).</li><li>Deterministic project and bundle digests (Identity).</li><li>Measured execution on retail Xbox Series hardware (Real hardware).</li><li>Zero arbitrary native-code upload.</li></ul>XCP facilitates reasoning on the PC and deterministic execution on Xbox, with the AI planning, creating, and correcting from the PC side, while the Xbox executes bounded, validated artifacts.