AI Agents Evaluation
Agent evaluation in 2026: beyond the benchmark illusion
Why benchmark scores mislead for agent systems — and what 2026's evaluation research says instead: three layers, log analysis and simulation caveats.
← Home
05.
Writing
Notes on RAG, agent systems, context engineering and LLM operations.
RAG AI Agents
Agentic RAG: from one-shot retrieval to research agents
RAG research in 2025-2026 shifted from one-shot pipelines to multi-step, decision-making systems. What actually changes, and when is it worth it?
RAG Long Context
Did long context kill RAG? What the 2026 evidence says
Did million-token context windows make retrieval obsolete? What comparative studies from 2025-2026 actually show — and how to decide in practice.
Inference MLOps
Inference engineering in 2026: speculative decoding, batching and quantization
Beyond API costs — if you serve your own models, the wins are at the systems layer. 2026's production measurements: speculative decoding, continuous batching, quantization and KV cache.
Research Video Generation
The 10 most upvoted papers in image and video generation (HF Daily Papers)
From Hugging Face Daily Papers' top-rated AI and vision papers, the generation side: playable game worlds, text-rich image synthesis, streaming video, diffusion acceleration and 3D.
Research AI Agents
The 10 most upvoted papers in visual agents and document AI (HF Daily Papers)
From Hugging Face Daily Papers' top-rated AI and vision papers, the agent side: visual computer use, GUI agents, robotic world models, OCR and document parsing.
RAG Retrieval
What RAG is (and when it fails)
A practical guide to Retrieval-Augmented Generation — what it is, when it works in production, and which problems make it the wrong tool.
AI Agents Tool Use
AI agent architecture: tool use, planning and error handling
Three things that make autonomous agent flows reliable in production: least-privilege tool design, disciplined planning and error handling.
Context Engineering Prompt Engineering
Context engineering: from prompts to context design
A model's answer is not shaped by the prompt alone — it is shaped by which context it gets, in which order and within which budget. A practical frame for context engineering.
LLM Cost Optimization
LLM cost optimization: caching, model choice and batching
The three levers that cut LLM costs without hurting quality — pick the right model, cache your context and batch your requests.