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
Running GLM-5.2 on a slow computer — how colibrì streams a 744B model from disk
colibrì is a ~1,300-line pure-C inference engine that runs GLM-5.2, a 744B-parameter MoE model, on a consumer PC with 25GB of RAM. The trick is MoE sparsity plus disk streaming: only ~9.9GB of dense layers stay resident,
2026-07-11 · 5 min read #ai#llm#local-inference#moe#quantizationBTS After Military Service: The Scale and Staging of the Arirang Comeback
Arirang, the first group album BTS released after every member completed military service, came out on March 20, 2026 and debuted at No. 1 on the Billboard 200 with 641,000 album-equivalent units in its first week — the
2026-07-11 · 6 min read #kpop#bts#music#korea#touringEU Chat Control 1.0: what actually passed, and the client-side scanning question
On 9 July 2026 the European Parliament let a message-scanning regulation pass even though more MEPs in the room voted against it than for it. The label Chat Control hides two very different things: a voluntary derogation
2026-07-11 · 6 min read #privacy#encryption#eu-regulation#client-side-scanning#messagingWeb Agents Read Page Text as Commands — Cross-Site Prompt Injection and Prismata's Confinement
Prismata, posted to arXiv on July 9, 2026 by Villa, Ozdarendeli, Tan, and Popa, tackles the oldest weakness of autonomous web agents. Because agents interpret natural language as instructions, third-party and user-genera
2026-07-11 · 5 min read #prompt-injection#web-agents#ai-security#least-privilege#browser-agentsComputation as a Universal Concept — What Holds Up and What Is Overclaimed
A short course titled 'Computation as a universal and fundamental concept,' taught by Tim Roughgarden, has been trending. Read closely, it is careful computer-science theory: the halting problem, algorithmic efficiency,
2026-07-11 · 6 min read #computer-science#computation#complexity#theory#philosophyThe Elite Athlete Mindset, With the Poster Peeled Off — What the Research on Practice, Pressure, and Belief Actually Shows
The elite athlete mindset is a staple of motivational posters, but the actual sports psychology is more careful, and more interesting, than the slogans. This post walks through Carol Dweck's growth mindset (and the false
2026-07-11 · 5 min read #mindset#psychology#performance#sports#deliberate-practiceApple Has Sued OpenAI — Talent Mobility, Trade-Secret Law, and How Developers Should Read It
On July 10, 2026, Apple sued OpenAI, two former employees, and OpenAI's hardware arm io Products in the Northern District of California, over trade-secret misappropriation and breach of contract. This piece separates wha
2026-07-11 · 6 min read #openai#apple#trade-secrets#legal#industrySpinning Up and Killing Postgres on Kubernetes with CloudNativePG — Failover Measured at 23 Seconds
On a real 8-node Kubernetes cluster, I installed CloudNativePG (CNPG) v1.30.0, brought up a 3-instance Postgres cluster, and then actually killed the primary. From bootstrap through replication checks, to failover after
2026-07-11 · 5 min read #cloudnativepg#postgresql#kubernetes#operator#databaseChoosing How Much to Think Per Step: Ares and Adaptive Reasoning-Effort Routing
A practical look at Ares, a March 2026 preprint that treats reasoning effort as a per-step cost lever for LLM agents. A lightweight router reads the interaction history and predicts the lowest sufficient reasoning level
2026-07-11 · 4 min read #ai#agents#llm#efficiencyWhy Developers Are Leaving GitHub for Codeberg — and What You Actually Give Up
Posts titled 'I left GitHub for Codeberg' are multiplying. Behind them are real departures — Ghostty, Zig, Gentoo — plus frequent outages, GitHub's absorption into Microsoft's CoreAI division, and an April 2026 flip to o
2026-07-11 · 6 min read #github#codeberg#forgejo#self-hosting#open-sourceTwenty Years of ACSM Fitness Trends: From Aesthetics to Mental Health, Longevity, and Data
The American College of Sports Medicine (ACSM) ran its Worldwide Fitness Trends survey for the 20th time this year, and the 2026 top spot is again Wearable Technology. This post lists the top 10, but reads the 20-year ar
2026-07-11 · 6 min read #fitness#health#trends#wellness#wearablesThe All-Stadium Bet: What Blackpink's 'Deadline' Says About the Economics of Female K-pop
Blackpink's 'Deadline' World Tour was marketed as the group's first all-stadium tour, running from Goyang on July 5, 2025 to Hong Kong on January 26, 2026. The lead single 'Jump' arrived on July 11, 2025 and debuted at N
2026-07-11 · 5 min read #kpop#blackpink#music-industry#korea#touringWhen an AI Maintains Your Code, Write for Humans Anyway
On July 10, 2026, Scott Robinson revived an old maxim with a twist: an LLM reads your codebase as its style guide, so every shortcut you merge becomes training data it repeats back at machine scale. We walk through his d
2026-07-11 · 6 min read #ai#llm#code-quality#maintainability#craftPractical Stoicism: Applying an Ancient Philosophy to Modern Life
Stoicism began in Athens around 300 BCE, yet its core tools fit today's anxieties surprisingly well. Grounded in the actual texts of Epictetus, Seneca, and Marcus Aurelius, this post lays out practices you can use: the d
2026-07-11 · 8 min read #philosophy#stoicism#self-improvement#life-adviceHabits Aren't Built in 21 Days: Re-reading Micro-Habits Through the 66-Day Study
'It takes 21 days to build a habit' is folklore that traces to a 1960 surgeon's observation, not a habit experiment. Lally and colleagues' 2010 study, which tracked real behaviour, found a median of 66 days to automatici
2026-07-11 · 6 min read #self-improvement#habits#psychology#productivity#behavior-changeJapanese Walking vs the Real Science: What Interval Walking Training Actually Shows
The 'Japanese walking' that spread across social media in 2026 is really Interval Walking Training (IWT), a protocol that researchers at Shinshu University have refined for nearly two decades, repackaged under a new name
2026-07-11 · 4 min read #fitness#health#walking#exercise#cardioManage Your Energy, Not Your Time — a grounded read on the 2026 wellness trend
2026 self-improvement coverage is shifting its center of gravity from time management to energy management, nervous-system regulation, and emotional fitness. But the framing is actually an old idea from a 2007 Harvard Bu
2026-07-11 · 6 min read #self-improvement#wellness#productivity#energy#recoveryBuilding Effective AI Agents: A Reference on the Five Workflow Patterns and Agents
A practical reference distilled from Anthropic's engineering guide "Building Effective Agents." It covers the precise distinction between workflows and agents, the building block underneath everything — the augmented LLM
2026-07-11 · 8 min read #ai#agents#llm#engineering#anthropicBuilding a Kubernetes GPU Operator in Rust — Diagnosing a Real Cluster with kube-rs
Against a production 8-node homelab cluster (k8s v1.32.5), I used kube-rs to build and run a GPU operator in Rust myself. I defined a GpuInventory custom resource and launched two controllers (node scan → record CR statu
2026-07-11 · 6 min read #rust#kubernetes#operator#gpu#kube-rsSeparating Signal from Noise When You Evaluate AI Coding Models — Why SWE-bench Got Shaky
OpenAI's evals team argues that SWE-bench Verified, the most widely used coding benchmark, no longer gives meaningful signal because of contamination and design flaws. Look at benchmarks through two axes — signal (the po
2026-07-11 · 6 min read #llm-evaluation#coding-benchmarks#swe-bench#benchmarks#ai-coding