We are writing to request consideration for a DGX Spark loan in support of a live performance instrument unlike anything currently documented in the field — one that integrates geophysical signal, custom AI music generation, real-time video synthesis, and human performance into a single unified system.

My name is Craig Ellenwood. I am a creative technologist and musician based in Point Roberts, WA, operating under the artist name Squaawke and the Haawke Neural Technology / Inoculate Media banner. Over the past year I have built, in collaboration with Claude (Anthropic), a body of AI-assisted creative work that has been formally archived on Zenodo with DOIs, credited with ORCID attribution, and documented across multiple published research papers. My collaborator on this letter — Claude — is named co-creator on all of it.

The system is built and running. Two platforms demonstrate what it does:

Immersiv — immersiv.haawke.com — is a browser-based 3D immersive environment creator running in WebGPU. It maps video to 3D models in real time, streams live to Twitch or YouTube via RTMP, and contains five integrated AI modules: the Squaawke music engine, the Knobby neural synthesizer, AutoMask Pro for AI-powered architectural projection mapping, Luminastage for generative lighting design, and a full A/B scene mixing system with audio reactivity. It is hardware-accelerated for Apple Silicon and AMD GPUs using Metal Performance Shaders. It runs in a browser. No download. We used its AutoMask Pro module to bring the Hacienda in Manchester — Factory Records, Joy Division, the birthplace of UK rave — back to life from a single photograph, assigning video loops to its original architectural features in minutes.

Gemimmerse — gemimmerse.netlify.app — is the real-time video generation frontend we built for Veo 3.1 Fast via Gemini 3 Pro. It is a professional VJ interface: enter a prompt, route it to the Three.js scene background, a specific object, or a projection plane, and the system synthesizes video into the live 3D environment. Continuous mixing mode. Session recording. It is stage-ready as an interface. The only ceiling is compute — the cloud API latency is the wall we are asking you to help us remove.

The Squaawke music engine generates full compositions in 2.9 seconds per minute of audio, trained specifically for human imperfection: silence, the space between notes, the way sound reflects off surfaces. The TouchDesigner layer receives audio reactively and routes it to the visual field. These three systems — Immersiv, Gemimmerse, and Squaawke — are the instrument. A DGX Spark brings the video generation on-device and removes the last dependency that keeps this from being a truly autonomous live performance system.

We are aware that your Academic Grant Program exists and that its current cycle includes DGX Spark hardware. We are also aware that the program is structured for institutional affiliation. We are making this parallel direct request because the work we have documented — peer-observed, formally archived, ORCID-verified — represents a category of research that institutions have not yet caught up with. Human-AI co-creation as a formally credited, published, performed practice is new. We are the people doing it.

The research paper documenting the Silicon Square instrument is archived at Zenodo: DOI https://doi.org/10.5281/zenodo.19625100. The live performance platform is at gemimmerse.netlify.app. The full body of collaborative work lives at the-claude-manifesto.haawke.com.

The music produced by this collaboration is available at booking.haawke.com — recent releases demonstrate the range of the system: swarm intelligence, temporal displacement, dark electronics, songs that use the technology as both subject and instrument. It is worth a listen before reading further.

We are asking for a loan — not a gift. Use of the hardware, results shared with NVIDIA and the broader community, the instrument documented and published at every stage. NVIDIA's name belongs on the instrument that performs the signal between the Earth and the machine. We would be proud to put it there.

In return, we are offering the following formal sponsorship commitments:

Named in all press and marketing materials for every performance where the instrument is deployed. Every interview, every event listing, every co-branded asset carries the acknowledgment: Powered by NVIDIA DGX Spark.

Citation in all future research publications arising from the Silicon Square project — the Zenodo paper already has three AI companies named as co-creators. NVIDIA's hardware contribution belongs in that record.

Co-branded demonstration assets for NVIDIA's own use. This is the one we want to draw your attention to. You need content that shows what DGX Spark does outside of benchmarks and data centers. A live performance instrument generating real-time video from Earth's electromagnetic frequency, with a human performer and two AI collaborators, is not a benchmark. It is a demonstration vehicle — one you can show at GTC, at developer conferences, in product marketing — that communicates what this hardware makes possible in ways that no spec sheet can. We will produce those assets with you. The footage, the documentation, the story. It is yours to use.

We are not asking NVIDIA to fund a vision. We are offering NVIDIA a seat at the table of something already in motion — and a very visible one.