Discover insights, tutorials, and thoughts on technology, homelab, and development.
A deep technical investigation into why MAX Engine allocates 117GB regardless of limits on NVIDIA DGX Spark's unified memory architecture - and why the solution doesn't exist in userspace
How I built an NVML library replacement to unlock GPU monitoring on Grace Blackwell unified memory architecture
A follow-up to my retracted post on self-improving models. The fidelity metric improved 18% but actual performance dropped 11%. Here's what went wrong, what I learned about Goodhart's Law, and why reproducibility filtering might be the answer.
We fine-tuned a security agent to 100% skill differentiation in probing tests, but it collapsed to a single behavior in deployment. This gap led us to develop a trust diagnostic framework.