Blog
Writing on GPUs, LLMs, MLOps, Kubernetes — and mindset · 3517 posts
#2026-03 765#english 592#culture 264#deep-dive 254#kubernetes 247#career 229#ai 216#llm 208#devops 193#2026-04 146#security 141#database 114#observability 113#communication 109#history 107#architecture 100#productivity 96#finance 88#economy 84#mindset 81#psychology 80#ai-papers 79#food 78#it 78#travel 78#deep-learning 77#japanese 77#networking 77#performance 72#business-travel 70#linux 70#gpu 69#ai-agent 66#cs-fundamentals 63#postgresql 60#rag 58#self-improvement 55#learning 53#mlops 53#ai-platform 51
NVIDIA GPU and CUDA Architecture Deep Dive: Why GPUs Dominate AI
From H100 hardware specs to Tensor Core WMMA API. A complete breakdown of the SIMT execution model, shared memory tiling, warp divergence, and why GPUs are purpose-built for AI workloads.
2026-03-18 · 14 min read #cuda#gpu#nvidia#matrix-multiplication#parallel-computingMulti-Agent Systems Compared: AutoGen vs CrewAI vs LangGraph — Which Should You Choose?
AutoGen, CrewAI, LangGraph — a practical comparison of three multi-agent frameworks with real code. Learn which one fits your use case, what each gets wrong, and the production pitfalls that apply regardless of which fra
2026-03-18 · 9 min read #multi-agent#autogen#crewai#langgraph#ai-agentMichael Jordan's Philosophy of Failure: What 9,000 Missed Shots Teach Developers
Michael Jordan's philosophy of failure and Carol Dweck's Growth Mindset research applied to developer resilience. From being cut from his high school team to 6 NBA titles, the story of how failure became a methodology.
2026-03-18 · 9 min read #michael-jordan#failure#growth-mindset#grit#motivationComplete LLM Serving Optimization Guide: KV Cache, PagedAttention, and Quantization
A complete engineering deep dive into LLM serving optimization. From the KV cache memory crisis to the virtual-memory revolution of PagedAttention, continuous batching, speculative decoding, quantization techniques, and
2026-03-18 · 30 min read #model-serving#kv-cache#paged-attention#vllm#quantizationLLM Agent Design Patterns: From ReAct to Multi-Agent Orchestration
ReAct, Chain of Thought, Plan-and-Execute, Reflection, Tree of Thoughts — a practical breakdown of which LLM agent design pattern to use and when, from an engineer who has built them in production.
2026-03-18 · 9 min read #ai-agent#react#llm#design-patterns#ai-developmentHybrid Search Guide: Combining BM25 and Vector Search for Better RAG
Pure vector search has a blind spot: exact keyword matching. This guide explains how to combine BM25 and vector search using RRF (Reciprocal Rank Fusion) to get the best of both worlds — with real code, benchmark numbers
2026-03-18 · 7 min read #hybrid-search#bm25#vector-search#rag#rrfCsikszentmihalyi's Flow: When Coding Becomes Meditation
This is the conceptual guide in the focus series. It explains why flow matters to developers through Csikszentmihalyi's model, deep work, interruption psychology, and the line between healthy immersion and burnout.
2026-03-18 · 17 min read #flow-theory#deep-work#csikszentmihalyi#focus#productivityDopamine Betrayal and Reconciliation: A Neuroscience Productivity Guide for Developers
Dopamine is not the pleasure hormone — it's the "wanting" hormone. From Kent Berridge's incentive salience theory to Anna Lembke's pain-pleasure seesaw: how to reclaim sustainable motivation from the technology that hija
2026-03-18 · 8 min read #dopamine#neuroscience#motivation#productivity#developercuDNN Internals: Why Deep Learning Operations Fly on GPU
From im2col transformation to Winograd algorithm, FlashAttention tiling, and TensorRT INT8 quantization. A complete breakdown of how cuDNN makes deep learning operations 10-100x faster than naive CUDA.
2026-03-18 · 13 min read #cudnn#gpu#deep-learning-optimization#convolution#model-servingThe Bamboo's Five Years: A Developer's Meditation on Invisible Growth
Moso bamboo spends 5 years underground before shooting up 27 meters in 6 weeks. The secret of this extraordinary growth holds a profound insight for developers — you are growing right now, in places no one can see.
2026-03-18 · 9 min read #bamboo#growth#patience#invisible-effort#natureWhat Ant Colonies Teach Developer Teams About Collaboration and Resilience
Ant colonies have operated without a central command for 30 million years, solving some of nature's hardest optimization problems through distributed intelligence. Exploring E.O. Wilson's myrmecology, Marco Dorigo's Ant
2026-03-18 · 8 min read #ants#collaboration#complex-systems#teamwork#developerAgent Memory System Design: Everything About How AI Agents Remember
LLMs have no memory by default. How do we solve this fundamental limitation? From four memory types to production architecture and the Mem0 framework — a complete guide to agent memory systems.
2026-03-18 · 8 min read #agent-memory#ai-agent#long-term-memory#langchain#mem0Zhuangzi's Free Wandering: How Wu Wei Unlocks a Developer's Creativity
Zhuangzi's 4th-century BCE philosophy of effortless action (Wu Wei) offers a surprisingly precise diagnosis of modern developer burnout, over-engineering, and creative blocks. A deep dive into the Peng bird, Cook Ding, a
2026-03-18 · 10 min read #zhuangzi#taoism#wu-wei#creativity#philosophyVibe Coding: The New Paradigm of AI-Assisted Development — Potential and Pitfalls
A deep dive into Andrej Karpathy's vibe coding concept from 2025. Is it a genuine revolution, or a technical debt machine? An honest look based on real experience.
2026-03-18 · 7 min read #vibe-coding#ai-coding#cursor#github-copilot#ai-developmentThe Champion's Ritual: Nadal, Ichiro, and Your Developer Kata
Rafael Nadal's water bottle ritual, Ichiro Suzuki's 3-hour pre-game preparation, and the martial arts concept of kata — why champions use pre-performance routines, and how developers can build their own to unlock focus,
2026-03-18 · 10 min read #routine#habits#nadal#ichiro#sports-psychologyRAGAS Complete Guide: How to Quantitatively Evaluate Your RAG System
Learn to use RAGAS — the gold standard for RAG evaluation — to measure Faithfulness, Answer Relevancy, Context Precision, and Context Recall. Includes complete Python code, automated CI/CD evaluation pipelines, synthetic
2026-03-18 · 9 min read #ragas#rag-evaluation#llm-evaluation#ai-development#qualityPrompt Engineering 2025: Getting Maximum Performance from Modern LLMs
Even the best model gives poor results with a poor prompt. Six prompt engineering techniques proven in production, with working code for GPT-4o, Claude, and open-source models.
2026-03-18 · 7 min read #prompt-engineering#llm#ai-development#gpt-4o#claudeNPU Deep Dive: How Transformer Architecture Runs Directly on Silicon
A complete technical breakdown of how NPUs differ from CPUs and GPUs, how every transformer operation maps to hardware, and why LLM inference is memory-bound — not compute-bound. From Apple ANE to Qualcomm Hexagon, Groq
2026-03-18 · 22 min read #npu#transformer#ai-hardware#quantization#kv-cacheThree Thousand Kilometers: What the Monarch Butterfly Teaches Developers About Legacy
No single monarch butterfly completes the round trip from Canada to Mexico. The migration takes four generations — yet each generation carries knowledge it has never personally experienced. Developers carry the same kind
2026-03-18 · 15 min read #monarch-butterfly#migration#collective-knowledge#purpose#natureHow Matrices Fly on GPU: Complete Deep Dive from GEMM to FlashAttention
Matrix multiplication accounts for over 80% of deep learning compute. This post dissects every layer of optimization from naive O(n³) code to cache blocking, cuBLAS, Tensor Cores, and the IO-aware revolution of FlashAtte
2026-03-18 · 15 min read #gemm#matrix-multiplication#flash-attention#gpu#parallel-computing