M3 Ultra Mac Studio
32-core CPU, 80-core GPU, and 512 GB unified memory.
Private AI lab / Owner attested
Four local compute nodes. No public doorway into the network.
A four-system local lab intended for private model experimentation, reproducible evaluation, and customer-controlled deployment patterns. Specifications are owner-attested. A signed, sanitized hardware attestation and reproducible benchmarks have not yet been published.
Owner-attested inventory · signed receipt pending01
Owner-attested systems
RTX PRO 6000 workstation
AMD Ryzen 9 9950X3D, 256 GB host memory, and 96 GB dedicated GPU memory.
NVIDIA DGX Spark
NVIDIA GB10, 20 CPU cores, and 128 GB unified memory.
NVIDIA DGX Spark
NVIDIA GB10, 20 CPU cores, and 128 GB unified memory.
02
Truthful capacity
- Unified memory is not described as conventional VRAM
- Memory across machines is not one automatic shared pool
- No unverified dollar valuation
- No unverified uptime, PFLOPS, energy, or maximum-model claim
- Benchmarks publish their protocol, model, settings, and limitations
03
Attestation boundary
The published attestation protocol allows only model class, coarse capacity, verification method, date, and artifact hash. A future local attestor must sign that sanitized record for CI verification. Hostnames, addresses, ports, serials, topology, utilization, and model paths remain private.
Next mission
See the local-first promise
The owner-attested inventory documents available local capacity without claiming that hardware proves service quality or exposing private infrastructure.