-
BF16 vs NVFP4 for a Local Coding Model: I Built a Benchmark to Decide
— Oct 14, 2026
-
One Flag Doubled My Local LLM's Speed: MTP Speculative Decoding on vLLM
— Oct 13, 2026
-
Testing My Sandboxed AI on the 1918 Flu Found a Hole in the Sandbox
— Sep 26, 2026
-
Sandy: An AI Assistant That Can't Leave the Sandbox
— Sep 1, 2026
-
Xid 79, Six Days Later: The Clock Lock Held, and the Motherboard Stays in the Box
— Aug 27, 2026
-
The GPU Fell Off the Bus, Again (and Again): Chasing Xid 79 to a Silicon RMA
— Aug 21, 2026
-
Xid 79: A Stress Test That Passed for the Wrong Reason, and a Reddit Tip That Might Fix It
— Aug 21, 2026
-
The GPU Fell Off the Bus: A Postmortem on the Blackwell Box's First Hardware Failure
— Aug 17, 2026
-
How to Create Atomic PRs Using AI Agents: A Spec-Driven Approach
— Aug 9, 2026
-
Building a Local LLM Server on Blackwell: Everything That Went Wrong
— Jun 4, 2026
-
Anthropic Changed the Rules: Revising My OpenClaw Split Deployment for Zero Cloud Spend
— Apr 9, 2026
-
Building a Split-Brain OpenClaw Deployment: Gateway + Inference Server Over Tailscale
— Mar 31, 2026
-
OpenClaw Hardware Showdown: Mini PC vs. Gaming Rig — When a Discrete GPU Changes the Equation
— Mar 31, 2026