← Back

Technical Logs

CUDA Pipelines
Jun 2026
How CUDA pipelines and asynchronous copies overlap data movement with computation, when multi-stage buffering pays off, and why deeper pipelines can hurt occupancy.
CUDA GPU Performance Memory
Moving a Distributed GPU Service into TEEs
Mar 2026
A practical walkthrough of running GPU workloads inside Trusted Execution Environments using NVIDIA's Hopper GPU Confidential Computing, covering hardware isolation, attestation, and performance overhead.
GPU Security CUDA TEE
Code Is Cheap Now. Engineering Isn't.
Feb 2026
A short reflection on how AI has made code generation trivial, shifting the real work in software development to architecture, judgment, and turning ideas into reliable, evolving systems.
Development Tools AI