InfoDrip — English Edition
Field notes from a senior engineer who plans, ships, and runs products end-to-end. Translated from the Korean original.
- ·Platform Decision
Shipping OKD 4.7 and 4.11 Node journald Logs to a Central Syslog — and Proving It Works First
Two earlier posts got me to the conclusion that node journals have to go somewhere off-box. This one is what came next: using ClusterLogForwarder to cover 4.7 and 4.11 with the same config, and verifying delivery from inside the cluster even when there's no central Syslog server yet.
- ·Convergence
What if You Put Prometheus on an Electrical Panel? Observability Belongs in the Power Room Too
When a server gets slow, we go straight to the metrics. So why don't we look at electrical equipment the same way? Turn current, temperature, and power draw into metrics, and a single distribution panel becomes the starting point for observing the whole plant. This is what manufacturing needs before it needs AI.
- ·Career & Strategy
I Want to Be Rich, So I Practice Thinking Like a CEO
I went looking for the morning routines of rich people and ended up back at the work I already do. I'm simulating a business by handing the work to AI and agents, practicing the calls a CEO would have to make. The more the agents take on, the clearer it gets what's actually mine to decide.
- ·Notes
Good Company, Bad Stock — Why Solid Tech, Rising Revenue, and a Bright Outlook Still Lose You Money
The company grew. Revenue and profit went up. The stock went down. Finding a good company and buying at a good price are two different problems. How to look at business value, market expectations, and share price as separate things.
- ·Platform Decision
Why Does Kubernetes Deprecate So Much? The Removal List Is a Payment Schedule
v1.37 kills kube-dns, IPVS, and cgroup v1. But the three deadlines aren't the same. cgroup v1 isn't a warning — nodes have already been refusing to start since 1.35. Start with the three commands that tell you which line your cluster is standing in.
- ·Career & Strategy
Why a Company Where the CEO Decides Everything Is a SPOF
Draw the architecture of a company where "we'll have to ask the CEO" is the standard answer, and every arrow points at one person. The paradox that a more capable CEO makes it worse — and how to design boundaries instead of stripping authority.
- ·Convergence
AI Looks Like Software. It's Actually Heavy Industry — What Building an AI Service Taught Me About the Cost of Intelligence
Imagine going from 100 users to 10,000. The question changes. Not what can I build with AI, but how long can I afford to keep this feature running. Scale one small service's cost problem up to 800 million people and it stops being a billing problem — it becomes a power and semiconductor problem.
- ·Convergence
How a Plain Distribution Panel Becomes a Smart One — From a Single Sensor to the Cloud, Where Electrical, Electronics, and IT Meet
You don't need to rip out the panel and build a new one. Here's the order it actually happens in: CT and temperature sensors, an edge gateway, then the data platform — following the seams where electrical work, electronics, and IT overlap.
- ·Operation Risk
Kubernetes doesn't recover from failures — it executes the failures you defined
Kubernetes has no idea whether your app is healthy. It reads probe results and does what you told it to do. Here's how one bad liveness probe turns a slow database into a full outage, plus real probe designs for web APIs, Spring Boot, and AI inference servers.
- ·Operation Risk
The Danger Isn't AI Writing Code. It's AI Having Permissions
AI agents now have hands, and that shifts the security axis. The problem was never AI's accuracy — it's what it's allowed to do. Starting from Wiz's Copilot Autofix case, here's how I think about least privilege and blast radius design in the AI era.
- ·Career & Strategy
What the FDE Role Taught Me About Being an Engineer in the AI Era
I followed the rise of the Forward Deployed Engineer role and thought about where an engineer's value is moving in the AI era. The cheaper implementation gets, the more what's left is the ability to structure a problem and turn it into a system.
- ·AI for Work
I Put an AI to Work and Learned That Leadership Beats Prompting
It took months of fiddling with prompt wording before I saw it. Getting good work out of AI has less to do with writing and more to do with leadership: setting the goal, splitting the work, and checking what comes back.
- ·Notes
I'm a libertarian, but the AI era might need a big government
With the cost of intellectual labor collapsing, I ran the numbers again on whether small government is still the answer. The paradox of capitalism working so well that redistribution becomes necessary — and what it looks like to see the state as architecture.
- ·AI for Work
I Built a Fortune-Telling Platform Just to See If Vibe Coding Was Real
I built a Korean fortune-telling platform to find out whether vibe coding actually holds up. Writing the code got cheap. Then I tried to bolt on payments, and every problem outside of development landed at once.
- ·Convergence
At Some Point, Manufacturing Started Looking Like One Giant Software System
Peel back one layer of a factory and everything from requirements to inspection reads like Build and Test. Notes from looking at manufacturing again — with veteran machinists as legacy systems and the conditions between process steps as APIs.
- ·Career & Strategy
I See a Company as One Giant IT System
A lens I picked up once the failures at work started looking like the failures in production. SPOF, APIs between departments, observability, and why executives are really architects.
- ·Notes
Why Stock Prices Move Before Earnings Do — The Market Caches the Future
Record earnings, falling stock. Money-losing company, rising stock. Look at it like an infrastructure operator and the market starts to look like a system that caches the future ahead of time.
- ·Operation Risk
I Only Wrote a Few Log Lines — So Why Is My Kubernetes Node's Disk Util at 95%?
CPU is fine, memory is fine, but Disk Util sits at 95%. Here's a write-up of the cases where logs turned out to be the culprit. Logs aren't strings — they're I/O, and in Kubernetes that load piles onto a single node.
- ·Career & Strategy
The 5-to-10-Year Developer Is in More Trouble Than the Junior — AI Is Erasing the Middle Class of Software
AI isn't erasing developers. It's erasing the middle first. Why seven years of experience is now the most awkward place to stand, and why judgment and accountability get more expensive as code gets cheaper.
- ·Operation Risk
Kubernetes skill shows up in how you explain an outage, not in how many kubectl commands you know
One `kubectl delete pod` is where it splits. The gap between someone who knows the commands and someone who can cut an outage into layers and explain it — from an operator's seat.
- ·AI for Work
If I Did the Thinking and AI Did the Writing, Whose Post Is It?
I found my own thinking in a Kubernetes comparison post that AI wrote. The sentences were the machine's, the judgment was mine. So who's the author?
- ·Operation Risk
The Journal Ate My OKD Node's Disk — Podman and Quadlet Log Management in Practice
Podman's journald and k8s-file logging drivers, LogDriver in Quadlet units, journald retention policies, rolling it out with CoreOS MachineConfig, and centralizing with Vector. Notes from running an OKD cluster.
- ·Platform Decision
The WWDC26 announcement everyone missed: your Mac is about to become a real Linux dev machine
I dug into WWDC26 session 389, an 11-minute talk on Container Machine. Here's what it says about how far Apple wants to erase the line between Mac and Linux.
- ·Convergence
A Distribution Panel Is Really Just a Building's Server Rack — How an IT Engineer Reads Electrical Infrastructure
I was standing in front of a distribution panel drawing when a server rack started showing through it. Notes on reading power distribution, circuit protection, and metering through an IT infrastructure lens — up to the point where a smart panel starts looking like a power edge device.
- ·Platform Decision
18 Months After Ditching Docker, and It Took Reading the runc Advisory All the Way Through
The runc CVE-2025-31133 advisory, Docker Desktop licensing costs, and the field: Podman 6.0, containerd+nerdctl, Apple container, OrbStack, Colima. I mapped out the 2026 container runtime migration starting from the audit.
- ·Platform Decision
Changing a Proxmox VMID by Hand, No Backup-Restore Required
How to change a VM's VMID in Proxmox VE without backing up and restoring it. Renaming the conf file, renaming disks across LVM, ZFS, and directory storage, then running qm rescan — notes from doing it on my own homelab.
- ·Notes
I Bought a House for $1M and It's Worth $1.5M. Did I Really Make $500K?
When a $1M apartment becomes worth $1.5M, it feels like $500K in profit. Subtract acquisition tax, ten years of mortgage interest, property tax, and opportunity cost, and what's left is smaller than you'd think. And where did that gain come from in the first place? I ran the numbers on how home prices actually rise — and on the people who started from the same place and ended up somewhere else entirely.
- ·AI for Work
I Translated 75 Blog Posts to English in 60 Minutes with Claude Code — $0 in API Costs
I had to get all 75 posts on my blog into English. Agency quotes ran into the thousands; the API would have cost tens of dollars. I wired Claude Code CLI into a pipeline instead: 60 minutes, zero failures, zero extra cost. Here's the 401 rabbit hole I fell into and how the pipeline is put together.
- ·Notes
Hope, Spider-Man, Odyssey — Three Movies Back to Back, and What They Told Me About My Own Taste
I saw three films in theaters recently. Hope, the one everyone drags for its graphics, was unexpectedly good. Spider-Man bored me. Odyssey's three hours flew by. My reactions ran exactly opposite to the review scores, and picking that apart showed me the structure underneath my taste.
- ·Convergence
Are Electronics, IT, and Electrical Really Different Fields? — What an Electronics Engineer Who Went Through IT Found When Looking at Electrical Installations
I stopped short while someone was walking me through a distribution panel diagram. Electronics deals in signals, IT in information, electrical in energy — but the structure of transfer, distribution, control, protection, and observation keeps asking almost the same questions.
- ·Platform Decision
Kubernetes vs OKD vs OpenShift: What Actually Separates Them?
The difference between Kubernetes, OKD, and OpenShift isn't free versus paid — it's where operational responsibility lands. Notes on rethinking platform choice as a question of who owns what.
- ·Convergence
The Era Where You Need Dollars to Buy GPUs, Not Oil — After the Petrodollar, the World's New Economic Infrastructure Could Be 'Compute'
I once tried to rack a GPU server and got stopped cold by power capacity. If the world starts scrambling for Compute instead of oil, underneath it all there are still power plants and transmission lines. Here's my read on the post-petrodollar era from an infrastructure point of view.
- ·Notes
Would Bigger Loans Really Fix Expensive Housing?
What happens when 200 people want to buy 100 delivery motorcycles and you raise the loan limit. A look at real estate lending through the lens of repayment capacity.
- ·Platform Decision
What Happened When I Put the Ansible Controller in a Container: Building a Portable Automation Node
How I containerized an Ansible control node with Docker so it runs the same way anywhere. Directory layout, ansible.cfg, SSH access, Compose, and the security details that actually matter in production.
- ·Platform Decision
Apple Container: Can You Run Containers on a Mac Without Docker?
Apple shipped Container 1.0, its own container tooling for macOS. I dig into what actually sets it apart from Docker Desktop, what the container machine model means in practice, and where it still falls short — from an engineer's point of view.
- ·AI for Work
The AI Boom Can Bust. AI Adoption Won't Stop.
Overheated AI investment and AI settling into everyday life are two different curves. Here's how diffusion actually worked for the internet and mobile, and where the next axis — memory-based AI and software robots — is heading, from an engineer's point of view.
- ·MSA & Architecture
Forget JSON. These 4 Data Formats Made Our API 5x Faster
Why JSON actually falls apart at scale, plus real measurements from getting a 5x serialization speedup with Protobuf, MessagePack, Avro, and FlatBuffers. An architectural take on format choice as a boundary decision.
- ·MSA & Architecture
The Teams Quietly Going Back to Modular Monoliths in 2026
More teams are walking microservices back into a modular monolith. Here's what I actually ran into around latency, debugging, and developer productivity.
- ·Platform Decision
Apache Iggy: What Rewriting Message Streaming in Rust Actually Buys You
Iggy, the Rust-based message streaming platform now in the Apache Incubator, claims 20 million messages per second and sub-millisecond latency. I took the Kafka alternative for a spin.
- ·Operation Risk
How Kubernetes Ate Six Months of My Life - 7 Things Beginners Miss
From the moment I thought my first cluster deploy had worked, to the three weeks it spent falling apart under load testing. The mistakes Kubernetes beginners keep repeating, and the habits that separate ops teams that survive.
- ·Platform Decision
Three Years of Docker, Then Podman: How We Cut Memory 40% and Fixed Our Architecture
What actually happened when we migrated from Docker to Podman: 40% less memory, four fewer EC2 instances, and how going daemonless changed the way we run things day to day.
- ·MSA & Architecture
Concurrency, Parallelism, and Async: Untangling the Three
Concurrency, parallelism, and async live at different layers of the stack. Here's what each one actually means, with real code and the operational trade-offs that decide which one you reach for.
- ·Platform Decision
Building an AI Agent to Run Kubernetes - A Practical 2026 Guide
How to safely put an AI agent that manages your Kubernetes clusters into production. MCP servers, approval gates, RBAC, and audit logging, from an operator's point of view.
- ·Notes
Why Do You Instantly Like Some People? The Science Behind a Magnetic Personality
Psychology research points to belief, not behavior, as the core of likability. A look at the mechanics of attraction through acceptance prophecies and emotional presence.
- ·Platform Decision
Lessons from 100K Connections: SSE Uses 40% Less Memory Than WebSocket
While building a real-time dashboard, I load-tested WebSocket against SSE. Here's the structural reason SSE came out 40% lighter on memory and more stable under load.
- ·Platform Decision
Why You Should Pick pnpm 11 Over npm in 2026
A practical look at pnpm 11's new security features and why it's worth moving off npm, framed around supply chain attacks.
- ·Platform Decision
10 Kubernetes Operators Worth Running in Production in 2026
From cert-manager to Crossplane, here are 10 battle-tested Kubernetes Operators, revisited from an operations point of view.
- ·AI for Work
From DevOps to AIOps: The Ground Is Shifting Under Operations
Cloud got complicated, and DevOps is starting to show its limits. Here's how AIOps changes the way you run systems, based on what I've watched change on the job.
- ·Platform Decision
Kubernetes Cost Tools Compared: OpenCost vs Kubecost, and Where AWS SCAD Fits
What actually happens when you try to track spend on a single EKS cluster. The real differences between OpenCost and Kubecost, and where AWS SCAD lands — from an operator's point of view.
- ·Operation Risk
The NGINX Rift: An 18-Year-Old Bug and an Abandoned Ingress Controller
Days after Kubernetes dropped support for the NGINX ingress controller, an 18-year-old critical bug surfaced. Here's what got exposed across half of cloud native, and what to do about it.
- ·Operation Risk
20 DevOps/SRE Interview Questions That Survive Production
DevOps/SRE interview questions built around incident response and how candidates actually work a problem. Not memorization checks — these probe the order an operator thinks in.
- ·MSA & Architecture
How to Escape REST API Call Hell: The API Design Strategies Senior Engineers Actually Use
Six API calls to render one screen? Here's the trap that 2015-era REST design left us with, and the API architectures senior engineers actually reach for on real teams.
- ·Notes
Why Linux Commands Live in /bin, /sbin, and /usr/bin
An engineer's take on why Linux splits commands across /bin, /sbin, /usr/bin, and /usr/sbin. The boot-time dependencies and permission design baked into the directory layout.
- ·Operation Risk
Nobody Owns the Blame Anymore: Accountability in the AI Era
In AI-native companies, accountability is quietly falling apart. An operator's look at the governance crisis that sits a layer deeper than visibility — and what actually fixes it.
- ·Platform Decision
Building a Real DevOps Pipeline, End to End
Notes from wiring together a CI/CD pipeline by hand with ten tools, from GitHub all the way to Kubernetes. How I set up Jenkins, SonarQube, ArgoCD, and Prometheus — and what it taught me about running the thing afterward.
- ·Enterprise IT
Enterprise AI's Reality Check: Spending Keeps Climbing, So Where's the Return?
Corporate AI budgets have never been bigger, yet ROI is hard to even measure. A look at what's really behind the AI ROI paradox, from a data and operations angle.
- ·Operation Risk
Your Dockerfile Is Leaking Secrets - The Layer Trap 99% of Devs Miss
Credentials live forever in Docker image layers. Why deleting them doesn't help, and how BuildKit secrets actually fix it.
- ·Platform Decision
MinIO CE Is Effectively Gone in 2026 — So What Do You Run Now?
MinIO CE's move to source-only broke a lot of production assumptions. Here's what that costs you operationally, and how SeaweedFS, Garage, and Ceph RGW actually hold up as replacements.
- ·Notes
The Samsung Strike and Korea's Uncomfortable Truth
An IT engineer's take on the Samsung Electronics union strike. A plain account of the same old collision between public opinion and labor rights.
- ·Platform Decision
Hidden Traps to Know Before You Pick a VMware Alternative
The operational complexity and hidden costs companies miss when evaluating VMware alternatives in 2026, based on lessons from real migrations.
- ·AI for Work
The Engineers AI Makes Sharper, and the Ones It Thinks For
Software engineering has hit a fork. Whether you use AI to push your thinking further or to skip it entirely is what decides what you're worth later.
- ·Platform Decision
Docker's Golden Age Is Over: Why Developers Are Moving to Podman and containerd
Kubernetes dropped dockershim and Docker changed its license. Here's the real reason Docker lost its monopoly, and why teams are shifting to Podman and containerd.
- ·Platform Decision
The 3 AM Kubernetes Outage That Turned Out to Be a Certificate
Expired TLS certificates take down Kubernetes clusters. Here's how I monitor them, automate renewal with cert-manager, and run the incident playbook when it still blows up.
- ·MSA & Architecture
Why Is Kubernetes Networking So Confusing?
From Pod IPs to Services — the core concepts of Kubernetes networking, and what actually separates ClusterIP, NodePort, and LoadBalancer in production.
- ·Platform Decision
Ingress NGINX Is Retiring — Here's How I'm Planning For It
Official support for Ingress NGINX ends in March 2026. I break down what Gateway API and Traefik actually give you, and a migration strategy that works in the real world.
- ·Notes
The AI Semiconductor Supercycle: Why This One Is Actually Different
In 2026, AI infrastructure spending has put the semiconductor market on a new trajectory. An engineer's look at the structural shift behind the rally in Samsung Electronics and SK Hynix.
- ·Platform Decision
Is Podman Really the New Docker? What Actually Differs in Practice
A hands-on comparison of Docker and Podman: daemonless architecture, rootless containers, security, and when to pick which.
- ·Platform Decision
12 Things That Cut Our Kubernetes Deploys From Minutes to Seconds
Image slimming through GitOps — the Kubernetes tuning tricks I've used to shave deploy times in production, from an operator's point of view.
- ·Platform Decision
Kubernetes Isn't as Hard as You Think
An engineer who has actually run containers in production walks through Kubernetes from the ground up — why scale forces your hand, how the architecture fits together, and what it looks like day to day.
- ·AI for Work
Why Tmux Is Still Worth Learning in 2026 — The Real Value of a Terminal Multiplexer
Remote work didn't kill Tmux; it made it essential. An operator's take on what actually matters: session persistence and reproducible dev environments.
- ·Platform Decision
AI-Powered Kubernetes Tools in 2026: What Actually Held Up
What I learned running K8sGPT, CAST AI, and Lens Prism against real clusters, and how I ended up combining them — from debugging to cost optimization.
- ·Career & Strategy
Why Do Boring Developers End Up Richer? The 5-Step Boring Wealth Formula
Flashy investing loses to simple, boring habits. An IT engineer's take on how to build your first $100K.
- ·MSA & Architecture
Linux Architecture, Unpacked: The Four Layers That Explain How Your System Actually Works
A hands-on breakdown of Linux from hardware up to user space, viewed through an ops and architecture lens: the kernel, system calls, and why modularity matters.
- ·Platform Decision
How I Finally Understood CRDs
I used CRDs every day without knowing what they actually were. Here's what changed my mind, and why Kubernetes is a language, not a platform.
- ·MSA & Architecture
Amazon S3 Got a File System - What Changes with S3 Files
The old trade-off between object storage and file systems is getting blurry. Here's what S3 Files actually changes, and how I'd decide when to reach for it.
- ·Notes
How to Survive at Work in Your 40s and 50s — Cognitive Psychology Advice from Professor Kim Kyung-il
Cognitive psychologist Kim Kyung-il on the mid-career slump: why it's boreout, not burnout, how to move from expert to master, and the trick of promising actions instead of feelings.
- ·Notes
Why the PS5 Keeps Getting More Expensive - Blame AI?
Four years in and the PS5 is going up in price, not down. I dug into how the AI chip demand explosion and a weak yen collided, from a foundry capacity angle.
- ·Career & Strategy
9 Lessons a 2,300-Year-Old Philosophy Still Teaches Developers
I reread Marcus Aurelius, Epictetus, and Seneca in the language of shipping software. A way of thinking that separates what you control from what you don't.
- ·AI for Work
Rebuilding Claude Code with Harness Engineering: The Hidden Architecture of AI Agents
Claude Code's real edge isn't the prompt — it's the harness wrapped around the model. I break down the 23 components you'd need to rebuild it.
- ·Operation Risk
Anatomy of the Axios npm Supply Chain Attack — 2 Hours and 54 Minutes
An engineer's breakdown of March's axios npm supply chain attack: how UNC1069 pulled it off, how they got around OIDC, and what to actually do about it.
- ·Platform Decision
How etcd Keeps a Kubernetes Cluster Honest
How etcd actually works under Kubernetes, what Raft consensus buys you, and the operational stuff that bites: disk I/O, backups, and node count.
- ·Notes
AI on the Factory Floor: What Manufacturing Actually Looks Like in 2026
From predictive maintenance to quality inspection, here's how AI is reshaping traditional manufacturing — written from an IT engineer's point of view. Real plant floor cases and the numbers behind where smart factories stand today.
- ·Career & Strategy
Working After Retirement: Is IT Any Different?
Inflation and near-zero interest rates broke Korea's retirement math. Here's why a high IT salary doesn't get you out of it either.
- ·Career & Strategy
IT Jobs in 2030: Which Roles Disappear and What Replaces Them
A structural look at how AI is reshaping IT jobs. Which roles survive to 2030, and which skills you'll actually need — from someone working in the field.
- ·MSA & Architecture
How Kubernetes Networking Evolved: From CNI to Cilium, and What eBPF Changed
Why Kubernetes never shipped its own networking, and how Cilium rewrote the landscape with eBPF — from an operator's point of view.
- ·Operation Risk
How Claude Code's Source Code Leaked Through npm
Anthropic shipped source maps that exposed all of Claude Code's source. A look at the 5.12M-line architecture and the operational risk one build setting created.
- ·Notes
Why Reading Better Beats Reading More
What 100 books in two years actually taught me. The trap of speed reading and volume, and an honest account of what real reading does to your head.
- ·MSA & Architecture
Rethinking GitOps on Large Kubernetes Platforms
Where GitOps hits real limits at scale, and how to draw clean lines between Git, OCI, and ConfigHub from an operator's point of view.
- ·Platform Decision
Helm 4 vs Helm 3: A Practical Upgrade Guide
A reality check four months after Helm 4 shipped. What Wasm plugins, server-side apply, and the rest actually mean — plus a migration strategy from an operator's point of view.
- ·Platform Decision
60% Memory Use and Your Pod Still Gets OOMKilled? Here's What's Actually Happening
Why a Pod dies with OOMKilled while memory sits at 60%. What averages hide about memory spikes, and how to actually fix it — from an operator's chair.
- ·Career & Strategy
5 AI Concepts You Need to Understand to Keep Up
From tokens to RAG, five core AI concepts explained by an IT engineer. Knowing how the tool actually works changes what you get out of it.
- ·Career & Strategy
The Jobs Quietly Disappearing — 3 Things People Who Survive the AI Era Have in Common
What the ongoing shift in work actually looks like in 2026, and what people preparing for it are really doing. A structural look at surviving the AI era.
- ·Notes
Why Hard Sci-Fi Matters Again: A Review of 'Project Hail Mary'
A look at Andy Weir's Project Hail Mary through three lenses — the science, the characters, and the friendship at its center. What the film changes from the book, and why it's worth reading now.
- ·Notes
[World Cup 2026] Korea Draws Czechia, and the Gap With Japan Shows: A Technical Breakdown
On the same day, Korea and Japan showed two very different faces ahead of the 2026 World Cup. I break down the Czechia opener and the tactical problems facing Hong Myung-bo's side from a systems perspective.
- ·Notes
Surviving the 2026 Job Market Collapse: 3 Strategies the Top 10% Actually Use
AI and tariff wars have put 300 million jobs on shaky ground in 2026. Here's what the top 10% do differently — the structure of real expertise, and the strategies that hold up in practice.
- ·Notes
Crimson Desert: The Story Falls Flat, So Why Do I Keep Booting It Up?
My take after playing Pearl Abyss's Crimson Desert. The story disappoints, but the visuals and the joy of exploration keep pulling me back — written by someone who kept picking the controller back up after swearing off it.
- ·Platform Decision
How Kubernetes DNS Actually Works — CoreDNS, ndots, and Why Your Lookups Are Slow
How do pods find each other in Kubernetes? A hands-on look at how CoreDNS resolves names, the ndots trap that quietly quadruples your queries, and what actually makes DNS fast in production.
- ·Notes
Rate Hikes Are Rewiring the Stock Market — 2026 Is a Different Game
Middle East tensions and rising rates crushed the Magnificent 7 while commodities turned strong. A systems-level look at what actually changed in 2026 investing.
- ·AI for Work
Three Months on Claude Opus 4.6: What It's Actually Like, Plus the Pricing (April 2026)
Notes from running Claude Opus 4.6 as my daily driver for three months. What 1M tokens of context actually feels like, how the plans break down, and what Moltbot/OpenClaw really is — from an operator's point of view.
- ·Notes
Seoul Cherry Blossoms 2026: Top 5 Spots — Peak Is This Weekend, With Routes
Seoul's cherry blossoms open around April 3, 2026, with full bloom around April 10. Here are five spots picked by what you're actually going for, from Yeouido to Yangjaecheon, plus how to work them into a route.
- ·Notes
60 Days to the 2026 World Cup: Hong Myung-bo vs. the Back-Three Debate — Can This Work With Son and Lee?
The 2026 North American World Cup is close, and Korean football is arguing again. Hong Myung-bo's tactics, the back three, and where the real problem actually sits — a structural look at where Korean football stands.
- ·Platform Decision
7 Things Kubernetes 1.36 Changes (From Killing Ingress-Nginx to Native Scale-to-Zero)
An operator's read on the 7 biggest changes in the Kubernetes 1.36 beta: HPA scale-to-zero, DRA-based scheduling for AI hardware, and what replaces Ingress-Nginx.
- ·AI for Work
Traditional DevOps Is Winding Down in 2026 — Which Infra Jobs Survive the AI Era
Writing Terraform and taking 3am pages is a fading job description. Here's how I see operations shifting toward AI infrastructure, platform engineering, and LLMOps.
- ·AI for Work
Is Vibe Coding Actually Essential? It Changed How I Work, Not What I Use
How coding alongside AI — "vibe coding" — reshapes the way developers actually work, from someone who has bounced between writing code and running it in production.
- ·Notes
When Will the Cherry Blossoms Bloom in 2026? 10 Hidden Spots Without the Crowds (Korea)
2026 cherry blossom bloom forecasts plus 10 uncrowded spots across Korea. From the Bukhangang Railway Bridge to Jeju's Noksan-ro, here's how to see the blossoms without fighting for space.