Posts
21 posts
RAG Failures After Retrieval
Specializing LLM Inference Stages
Perceived Effects and Measured Outcomes of AI Tools
How Collision Handling Changes Hash Costs
LLM Defenses at Action Boundaries
Compute Allocation in Multi-Agent Teams
Layers Where Transaction Costs Sit
Stopping Policy in Agentic Search
Measurement Boundaries in LLM Serving Optimization
Why Recovery Paths Need Separate Verification
From Token Alignment to Representation-Level Defense
Judging Multi-Agent Adoption on Reliability
Three Approaches to Reducing Consensus Latency
RAG Moves From Retriever to Compiler
The Gap Between Data Structure Theory and Libraries
The Tool Layer as the Agent Bottleneck
LLM Serving Seen Through the KV Cache
The Agent Memory Trilemma
From 2PC to the Commit Log
Storage Engine Optimization Moves Below the Kernel
Reranker Objective Shift