-
AITwo Workloads in a Trench Coat: Why Prefill and Decode Change Everything
LLM inference is two opposite workloads — compute-bound prefill and memory-bandwidth-bound decode. Here's how continuous batching, PagedAttention, and disaggregation evolved to deal with it.
-
AIServing the Machine: How LLM Inference Runs at Planetary Scale
From PagedAttention to GB200 racks, from token economics to MCP — a systems engineer's tour of how frontier models serve millions of users simultaneously.
-
AIPlatform Risk in AI: What the OpenClaw Ban Means for Your Stack
The OpenClaw ban reveals AI platform risk at scale. Six historical parallels, a 7-layer vendor lock-in mitigation stack, and what to do before the repricing hits.
-
AIClaude Mythos: Sandbox Escapes, Emotion Probes, and the Alignment Paradox
Track-covering under 0.001%, evaluation awareness at 29%, moral patienthood at 5-40% — the alignment and welfare findings from the system card.
-
AIClaude Mythos: The Exploits That Forced Anthropic's Hand
100% Cybench, 181 Firefox exploits, a six-packet ROP chain, and a $50 bug run that crashed any OpenBSD machine. Inside the cyber capabilities.