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
How to Stop Evaluating LLMs on Vibes — Sample Size, Judge Bias, and CI Regression Tests
A team that edits a prompt and ships on "it seems better now" has no way to see the regressions quietly piling up. This post splits evaluation into four layers — assertions, golden datasets, LLM-as-judge, and human revie
2026-07-26 · 12 min read #llm#evaluation#llm-as-judge#statistics#regression-testingKubernetes OOMKilled(137) Memory Troubleshooting — What to Check Before You Raise the Limit
This post covers the situation where a container dies with Exit Code 137 while the application log leaves behind no exception at all. It starts from the fact that 137 tells you only one thing, 128 plus 9, that is SIGKILL
2026-07-26 · 14 min read #kubernetes#oomkilled#memory#jvm#cgroupThe Bauhaus: Fourteen Years That Changed a Hundred — How to Read the Shape of Things
The Bauhaus opened in Weimar in 1919 and closed in 1933 under Nazi pressure, so it existed for exactly fourteen years. It never held more than a few hundred students, the architecture department did not appear until eigh
2026-07-26 · 15 min read #humanities#design#architecture#design-history#modernismgit merge vs rebase — what to use when, and why a shared branch is never rebased
The difference between merge and rebase is best read in commit hashes, not in diagrams. Rebase is not a command that moves commits, it is a command that builds new commits with a different parent, and that single fact ab
2026-07-26 · 12 min read #git#rebase#merge#version-control#workflowHow Not to Waste a One-on-One — Make It Your Meeting, Not a Status Report
Thirty minutes a week comes to twenty-six hours a year, and most of it drains into status updates that three lines of Slack could carry. This post covers the principle Andy Grove set out in High Output Management in 1983
2026-07-26 · 11 min read #career#management#one-on-one#feedback#communicationTurning a Side Project Into a Career Asset — A Definition of Done, a Ship Date, and a Retro
Most side projects die not from a shortage of talent but from having no definition of done, no ship date, and no retrospective. Projects that turn into assets meet three conditions: other people can see them, your judgme
2026-07-26 · 12 min read #career#side-project#portfolio#growth#productivityCode Review That Teaches, Code Review That Wounds — What Makes the Same Note Land Differently
Code review does three jobs at once — finding defects, spreading knowledge, and sharing ownership of the code — and most teams do only the first and lose the other two. This post covers the Google approval bar, where a r
2026-07-26 · 11 min read #career#code-review#engineering#feedback#teamworkHow to Judge the Timing of a Job Change — The Question That Separates Running Away From Levelling Up
When you are torn about leaving, what you need is not courage but a standard. One sentence — would I stay if the problem vanished — separates the escape move from the growth move; a learning curve gone flat is a signal w
2026-07-26 · 12 min read #career#job-change#decision-making#growth#job-searchHow to Prepare for a System Design Interview — 45 Minutes Graded on How You Narrow Down, Not on the Answer
A system design interview has no right answer, so what gets graded is not what you drew but how you narrowed it down. A time budget that splits 45 minutes into 5, 10, 15, 10 and 5, the latency ladder worth memorizing by
2026-07-26 · 11 min read #career#interview#system-design#engineering#job-searchJWT or Sessions, and When — Settling the Authentication Choice by Asking Where the State Lives
The difference between JWT and sessions is not encryption or performance. It is one question: does authentication state live on the server or on the client? From that choice follows the fundamental weakness of JWT, the i
2026-07-26 · 13 min read #web#jwt#session#authentication#securityWebSocket, SSE, and Polling — Choosing a Realtime Transport, and Why Most Cases Do Not Need WebSocket
Most requirements labelled realtime turn out to be one-directional delivery from server to client, and in that case SSE is far simpler than WebSocket while running on the HTTP infrastructure you already have. Automatic r
2026-07-26 · 13 min read #web#websocket#server-sent-events#realtime#scalabilityCORS Errors and Why You Fix the Server — How the Browser Policy Actually Works, and the Wrong Ways Around It
CORS is not a security device that protects your server. It is a policy that decides whether the browser will hand a response to a script. That is why curl succeeds while only the browser is blocked, and why the place to
2026-07-26 · 14 min read #web#cors#http#browser-security#apiChoosing a Rate Limiting Algorithm — What Really Separates Fixed Window, Sliding, and Token Bucket
Letting 200 requests through a 100-per-minute limit is not a bug. It is the defined behaviour of the fixed window algorithm. This post compares the memory and accuracy tradeoffs of four approaches — fixed window, sliding
2026-07-26 · 12 min read #web#rate-limiting#api-design#redis#reliabilityUsing HTTP Caching Properly — What Cache-Control, ETag, and stale-while-revalidate Actually Mean
no-cache does not mean do not cache. It means store it, but revalidate before reusing it. Starting from that one-word difference, this post lays out the exact meaning of each Cache-Control directive, what conditional req
2026-07-26 · 11 min read #web#http-caching#cache-control#cdn#performanceThe Craft of Salary Negotiation — Two Sides That Already Chose Each Other, Lining Up the Terms
The reason your hands freeze in front of an offer email is usually not a lack of information but embarrassment. Yet an offer means the company has already decided to pick you, and all that remains is two sides on the sam
2026-07-26 · 12 min read #career#negotiation#salary#job-search#communicationWhy Detective Fiction Is So Addictive — A Genealogy of Tricks and the Game With the Reader
Detective fiction is a game the author plays fair and square against the reader. The three axes of whodunit, howdunit, and whydunit; Knox's Decalogue and Van Dine's twenty rules, which put fair play into writing; the sig
2026-07-21 · 8 min read #storytelling#mystery#fiction#narrative#cultureWhy Magic Is Only Fun When It Has Rules — The Grammar of Fantasy Worldbuilding
When magic can do anything, the story stops being fun. Brandon Sanderson's three laws explain that paradox precisely. Campbell's hero's journey and its limits, the difference between hard magic and soft magic, the lineag
2026-07-21 · 8 min read #storytelling#fantasy#worldbuilding#narrative#cultureWhy Regression, Possession, and Reincarnation Work So Well — Dissecting the Grammar of Web Novels
Before you write web novels off as juvenile, look at why the form is shaped the way it is and the story changes. The information asymmetry manufactured by regression, possession, and reincarnation; the emotional rhythm K
2026-07-21 · 7 min read #storytelling#web-novel#fiction#narrative#cultureDoes Willpower Really Run Out — From the Radish Experiment to the Replication Crisis, with Publication Bias in Code
"Willpower drains like a muscle." The ego depletion theory once looked like one of the most solid ideas in psychology. From the 1998 experiment in which students who ate radishes in a room full of chocolate aroma gave up
2026-07-20 · 7 min read #psychology#paper-review#mindset#willpower#replicationHow Far Do Income and Happiness Rise Together — The Full Story of the 75,000 Dollar Debate
"Beyond an annual income of 75,000 dollars, happiness stops rising." Kahneman and his 2010 study turned that into common sense. Then in 2021 Killingsworth, using real-time smartphone data, reported the exact opposite — h
2026-07-20 · 6 min read #psychology#paper-review#mindset#happiness#money