Rich Dunbar

Vibe coding the Nexus suite.
Build. Measure. Learn.

I’ve been using AI-assisted development to build a connected suite of applications for creating, training, compressing, recovering, and running AI locally.

The work spans:

• Nexus Foundry — creating language models from random initialization.
• Nexus Compression & Recovery Studio — exploring smaller models and recovering capability after compression.
• Nexus Adapter Optimization Studio — developing specialized model adapters.
• Nexus Benchmark — measuring performance and identifying settings that can be applied to the actual applications.
• Nexus.cpp — developing my own inference engine, with a focus on Radeon hardware.

Alongside these, I’m working on Nexus Office, Imaging Studio, virtual machine tools, and the longer-term Nexus OS project.

My approach to vibe coding is hands-on: define the objective, work with AI to implement it, run the software, inspect the results, and bring the failures back into the next iteration.

A setting needs a working backend. A benchmark needs to represent the real workload. An exported model needs to load and produce useful results.…

Some forge with fire. I tuned the AMD Radeon RX 7900 XT to train AI with PyTorch on Windows 11.
My method: build, measure, learn.
That’s the approach behind Nexus Foundry, my AI creation tool optimized for the AMD Radeon RX 7900 XT on Windows 11.
Build.Measure.Learn
During my latest run, training a 20-million-parameter model from random initialization, Nexus Foundry reported:

• 24,893 training targets per second. • 66% GPU utilization. • 3,663 MB of VRAM in use. • Approximately 16 GB of remaining VRAM capacity. • 10% CPU utilization.

Nexus Foundry brings model creation, PyTorch training, configuration, checkpoints, and performance monitoring into one local application.

The remaining VRAM gives me room to experiment with larger batches and larger models. My next milestone is a 120-million-parameter model, testing sustained throughput alongside model quality.

Every run informs the next improvement: build the capability, measure what actually happens, and learn where to optimize.

The throughput shown is the application’s recent 3-second moving average. Hardware readings are snapshots; unused memory does not guarantee proportional speed gains.…