The Launchpad TLP · Intelligence Briefing · April 2026 Project Glasshouse: Virtualizing macOS on Agnostic Infrastructure Author: Joaquín Antonio "Piqui" Muñoz Ortiz / The Launchpad TLP Prepared by: Manus AI Date: April 5, 2026 Status: Theoretical / Experimental — Legal gray area (Apple EULA). For research and educational purposes only. Quick Links Full Plain Text Briefing AI Manifest (llms.txt) JSON Content API JSON Transcript API Executive Summary Project Glasshouse is a complete technical framework for running Apple macOS on non-Apple cloud infrastructure at zero cost. It addresses three core problems: (1) the Emulation Tax (solved by Arancini HBT, 5x faster than QEMU); (2) hardware-accelerated graphics (solved by KosmicKrisp Vulkan-to-Metal); (3) zero-cost infrastructure (solved by the Scavenger Architecture). The Four-Phase Architecture Phase 1 — The Engine Room KVM hypervisor with HugePages memory management and IOMMU PCIe passthrough. Docker-OSX packages QEMU, OpenCore, and macOS recovery into an Arch Linux container. Phase 2 — The Ghost OpenCore bootloader with cryptographic hardware spoofing: Lilu.kext (patching engine), VirtualSMC.kext (SMC spoofing), WhateverGreen.kext (framebuffer patching). Phase 3 — The Translator (Arancini) Hybrid Binary Translator using LLVM. Lifts AArch64 to LLVM IR, applies AVX-512 optimizations, recompiles to x86_64. 5x faster than QEMU TCG, 81% fewer memory access instructions. ASPLOS 2026. Phase 4 — The Interceptor (KosmicKrisp) Vulkan-to-Metal Mesa driver by LunarG (2025). Vulkan 1.3/1.4 conformant. Enables 60fps hardware-accelerated graphics inside macOS VMs on generic cloud GPUs. The Scavenger Architecture — Zero-Cost Infrastructure Total cost: $0.00 GitHub Actions — Free M1 Apple Silicon runners (macos-14), 6 hours per job, unlimited on public repos Oracle Cloud Always Free — 4 ARM cores, 24GB RAM, 200GB storage, permanent, $0.00 Tailscale — Mesh VPN, free for personal use, connects all components into one virtual LAN MCP + Llama 3 — mcp-server-macos-use (mediar-ai) + local Llama 3 on Oracle = $0.00 AI inference Key Quote "To build a 'God-Mac' cloud infrastructure with absolutely zero capital, we have to abandon the traditional path of renting dedicated servers. When you have no money, your primary currency is cunning, open-source leverage, and exploiting corporate free tiers to their absolute limits." — Google Gemini, Project Glasshouse Conversation, April 2026 Legal Framework Apple EULA: Explicitly prohibits macOS on non-Apple hardware. Commercial use = immediate legal risk. Corellium case (11th Circuit, 2022): Security research virtualization may qualify as fair use. EU DMA Article 6(7) (2024): Requires Apple to allow third-party virtualization in the EU. DMCA Section 1201(f): Permits reverse engineering for interoperability purposes. Research Coordinates Arancini HBT: dl.acm.org/doi/proceedings/10.1145/3669940 KosmicKrisp: lunarg.com/kosmickrisp-vulkan-conformance-apple-silicon/ Docker-OSX: github.com/sickcodes/Docker-OSX OpenCore: github.com/acidanthera/OpenCorePkg MCP: modelcontextprotocol.io mcp-server-macos-use: github.com/mediar-ai/mcp-server-macos-use Oracle Always Free: cloud.oracle.com/free Tailscale: tailscale.com llama.cpp: github.com/ggerganov/llama.cpp Darling (macOS on Linux): github.com/darlinghq/darling MoltenVK: github.com/KhronosGroup/MoltenVK gibMacOS: github.com/corpnewt/gibMacOS Akash Network: akash.network GSA Auctions: gsaauctions.gov GovDeals: govdeals.com Municibid: municibid.com