All Posts
Technical learnings, reflections, and daily discoveries
-
Why Your Retrospectives (KPT) Don't Stick — Fixing "Write It and Forget It"
You run retrospectives (KPT) every time, yet they never seem to pay off next time. The reason is structural: you write them, then they get buried. Here are three reasons retrospectives go hollow — and how to make past lessons resurface right when you need them.
-
"Why Study for Certifications When AI Exists?" — Returning to Basics
AI-driven development cuts learning costs to near zero and boosts productivity. The more I leaned on that convenience, the more I doubted whether studying still mattered. Here is why I returned to basics and chose to face the Information Processing Engineer Exam head-on.
-
Suggestions That Tell You "Why" — Tasukiba, the AI Operations Secretary
A relevance score of 0.82 alone isn't enough. The "Why?" feature in Tasukiba, your AI operations secretary, adds a plain-language reason to every past record surfaced by semantic search — so your project management decisions get the context behind each suggestion.
-
Just Talk to Tasukiba — Semantic Search Chat in Your AI Operations Secretary
"Wasn't there a retrospective about resource shortage last month?" Just ask, and Tasukiba, your AI operations secretary, returns past knowledge management records by semantic search. How it beats a normal project management tool where you hunt records yourself.
-
Six Months to Ship an AI Operations Secretary — A Solo-Dev Recap
From 2025-12 kickoff to 2026-06-01 release, a month-by-month retrospective of building Tasukiba, my AI operations secretary for project management and semantic search. 450+ improvements, shipped solo in 480 hours alongside a day job. The honest timeline.
-
Why I Drew the Tasuki Owl Three Times — General, Chat, and SNS
The Tasuki Owl, face of Tasukiba my AI operations secretary, has three master images, not one. General (logo / favicon / OG), semantic-search chat, and SNS. Why I refused to make one image do everything, and how five derivatives cover each context.
-
Why the Tasuki Owl Won — Scoring 4 Mascot Candidates on 3 Axes
The Tasuki Owl is the face of Tasukiba, my AI operations secretary. Owl, lighthouse, open book, abstract geometric — four candidates scored on knowledge management, project management, and security. Only the owl hit top marks on all three. Why navy blue won.
-
A "Cash Cow in Ten Years" — What My AI Operations Secretary Will Do for Me
What I want for Tasukiba, my AI operations secretary, in ten years isn't a hyper-growth IPO — it's a slowly rooted tree. A semantic-search tool for project management and knowledge management that grows on word of mouth, with profits fueling the next challenge.
-
I Won't Sell a Tool I Don't Use Daily — Dogfooding My AI Operations Secretary
Tasukiba, my AI operations secretary, runs on one premise: I'm the first user. From day one, all my day-job project management flows through it. It surfaces past knowledge by semantic search, and I sharpen it in the field every single day. Here's why.
-
One Week After Release — AI Operations Secretary Tasukiba's Phase 2/3 Roadmap
The June 1, 2026 release of AI operations secretary Tasukiba shipped all three signature features — suggestions, chat semantic search, and the 'Why?' feature. Phase 2 brings Stripe live billing and dogfooding feedback; Phase 3 brings multilingual support and a public API. The roadmap for strengthening your project management, shared one week in.
-
Two-Year Monetization Check — But the AI Operations Secretary Tasukiba Doesn't Stop
If AI operations secretary Tasukiba monetizes within two years, it takes the growth path. If not, sales stops — but the service doesn't, because I keep using it as my own project management tool. Separating business judgment from tool continuation, explained honestly.
-
One For All — How the AI Operations Secretary Tasukiba Designs Out Hidden Failure
Failure on AI operations secretary Tasukiba lands with me, not the member. Every report gets a heartfelt 'thank you for reporting.' We replace the urge to hide with a culture where reporting gets you help — the One For All team culture, blame-less postmortems, and day-job-first norms, in detail.
-
Equal-Founders Model — Why the AI Operations Secretary Tasukiba Refuses Top-Down
Final decisions on AI operations secretary Tasukiba rest with me, but the team runs as an 'equal-founders model,' not top-down. When opinions split, a three-layer discussion (team → outside → team) decides. It's designed to prove out a high-psychological-safety org that can sustain a project management service for the long run.
-
Empathy as the Hiring Filter — Building the AI Operations Secretary Tasukiba With Me
The hiring filter for AI operations secretary Tasukiba isn't technical skill, it's empathy. Values alignment on three axes (purpose, target, method) comes first: can you share the belief in reshaping project management with semantic search? Why I weight empathy differently for known people vs outside applicants, written honestly.
-
Return the Controls — AI Operations Secretary Tasukiba's UI Without a Gatekeeper
Plan changes, cancellations, downgrades — all done with zero vendor round-trips in AI operations secretary Tasukiba. 'Your data is yours' export, the one self-role-change exception, and matching what the screen offers to what you're authorized to do. Project management put back in your own hands.
-
Starting at $0 — Pricing the AI Operations Secretary Tasukiba, All 3 Plans Open
AI operations secretary Tasukiba's three plans (Beginner/Expert/Pro) and usage billing are designed to last, not to maximize profit. Fixed $0/month, the case for free ops, capacity-based usage billing, and self-service cancellation — every pricing decision behind your project management, explained.
-
Tasukiba, the AI Operations Secretary — Semantic Search Surfaces "The Past You Forgot"
The core of AI operations secretary Tasukiba is its suggestion engine. Unlike ordinary project management tools where you search yourself, semantic search auto-surfaces your knowledge management history at three moments. Why push beats pull, the 3-axis score, and zero-cost-at-suggestion-time design.
-
Tasukiba, an AI Operations Secretary, Ships Today — Six Months in Review
Tasukiba Knowledge Relay launches June 1, 2026: an AI operations secretary for project management and knowledge management that uses semantic search to surface past project knowledge, risks, issues, retrospectives, and notes at project start. Three plans from $0/month, the tech stack, and core features — delivered with launch-day emotion intact.
-
Graduation, Not Leaving — Why Tasukiba, an AI Operations Secretary, Chose Discord
Why I picked Discord over Slack for the community around Tasukiba, an AI operations secretary, and why I call leaving 'graduating.' A lower exit barrier makes a place more comfortable to stay — a design call that runs continuous with the psychological safety project management teams need.
-
Meet the Tasuki Owl — Why Tasukiba's Mascot Is an Owl
The mascot of Tasukiba, an AI operations secretary, is an owl. Three symbols converge — wisdom (Athena's bird), memory (night vision), protection (guardian) — onto the product's three axes of knowledge management, project management, and security. Here's how the owl became the companion that surfaces related assets by semantic search.
-
How Is Tasukiba Different From an Ordinary Project Management Tool? Six Axes
Where an ordinary project management tool makes you full-text search and hunt yourself, Tasukiba — an AI operations secretary — surfaces related assets by semantic search. The six differentiation axes came from stubborn preference, not market research: semantic suggestions, pricing that keeps you going, retrieval of past assets, cross-tenant community, user autonomy, and a place for knowledge management.
-
Tasukiba's Tidy UI — Roots in a High-School Notebook
Tasukiba, an AI operations secretary, has a UI built on one belief: things belong where they belong. The root traces back to a high-school notebook redesigned for the reader every term — a habit that, twenty years later, shapes how project management and knowledge management feel on screen. Told as five definitions.
-
Changing Sunday Night and Monday Morning — Tasukiba's Worldview as an AI Operations Secretary
Sunday evening, you sigh about Monday's escalation. Tasukiba, an AI operations secretary, surfaces the project management context you need by semantic search before you go looking. Every product decision is judged by one question: does this brighten Monday morning? Here's the worldview, written with its emotional temperature intact.
-
An Hour a Day Was Evaporating — Why I Built Tasukiba, an AI Operations Secretary
Six people times ten minutes equals an hour a day, evaporating from meetings just to find a file. Data piles up unused while judgment leans on a few heads. I built Tasukiba, an AI operations secretary, to dissolve that with semantic search across project management and knowledge management.
-
Think Again and the Binary Bias I Couldn't See in Myself
Adam Grant's 'Think Again' named a habit I'd never noticed — collapsing nuance into either/or. Four thinking modes and three daily practices for the gray zone.
-
Five hours that felt like two — Adler's 'now, here' as I actually felt it
A night studying for an InfoSec certification when five hours felt like two: body exhausted, time forgotten. That, I realized, is what 'live now, here' actually feels like. The closing post of the three-part Adler series, woven with my habit of overthinking my career and the moment a teammate said 'thank you' for an AI rollout.
-
The day I stopped trying to move other people — Adler's separation of tasks, finally landing
From the days when I was wearing myself out trying to make other people move, to the moment 'how others evaluate me is their task' actually hit me at work. A reading note on how The Courage to Be Disliked's separation of tasks finally landed, at home and at the office — and the 1-on-1 phrase I use to keep the principle alive in daily conversation.
-
'Only I protect me' — how The Courage to Be Disliked handed me back the steering wheel
From caring deeply about how others saw me, to noticing 'how others evaluate me is their task' and starting to show up as myself. The Courage to Be Disliked's teleology rejects causal blame of the past and returns the steering wheel of life to your own hands. A reading note on the path through a job change and side-project that the book quietly backed me through.
-
Knowing psychological safety isn't enough — the four stages: understand, design, implement, reproduce
Many people can say psychological safety matters. Few can actually design it into a team, and fewer still can reproduce the same level in a different team. A four-stage model — built on Amy Edmondson's research — for moving from understanding to reproduction, the four leader behaviors that make it concrete, and why psychological safety only works when paired with accountability.
-
The urge to decide fast is your strength's shadow — Need for Cognitive Closure and the 2-week rule
People praised as fast decision-makers are often the most vulnerable to the Need for Cognitive Closure — the psychological pull to escape ambiguity by reaching for an answer too early. A look at Arie Kruglanski's concept, why structural thinkers are especially susceptible, and three small daily habits — the 3-column journal, the 2-week rule, and stopping at three Whys — that I'm using to retune my own judgment.
-
When I realized my biggest gap wasn't technical — what an outside review showed me
Approaching a career change at 27, I asked someone outside my circle to do a gap analysis on me. The most important line in the report wasn't about new skills or certifications. It was about turning the experience I already had into transferable knowledge, and seeing my own thinking habits more clearly. A reflection on the gap mid-career engineers most often miss.
-
Reading Notes on Reinventing Organizations — Why Pyramids Fail and How to Build a Self-Driven Team
A reading note on the illustrated introduction to Reinventing Organizations (Teal). Tracing organizational evolution from Red to Teal, this article unpacks why pyramid structures break at scale, what genuine delegation looks like, and how a rookie team lead can stop accidentally killing intrinsic motivation.
-
The Art of Spending Money — Invisible Value and the Real Wealth Called Independence
Reading notes on The Art of Spending Money. The book frames the real value of money as the gap between what you have and what you want, and argues that invisible value — not visible status — is what generates happiness and independence. I tie its core ideas to my own habit of paying with gratitude toward the people behind a product or service.
-
Thinking 101 — Fluency, Confirmation Bias, and the Opposite Question
Reading notes on the Yale intensive lecture-based book Thinking 101. Five concepts that sharpened how I form and test hypotheses at work: fluency-driven overconfidence, why plans are always too optimistic, confirmation bias as a survival strategy, and the surprisingly simple technique of asking the opposite question.
-
5 Internal SEO Tricks — Tag Pages, noindex, Redirects
Code-level walkthrough of 5 internal SEO tricks: dynamic tag pages, 120-160 char descriptions, noindex for thin tags, URL canonicals, and an internal-link hub.
-
Personal Blog: 1 Month, +733% Clicks, +496% Impressions
One month of Search Console data: Clicks 6→50, Impressions 77→459, CTR 7.8%→10.89%, Position 38→18.97 — quantitative review of what moved the numbers.
-
Personal Blog SEO — Lift Position to 18 at Zero Cost
From position 38 to 18.97 and CTR 7.8% to 10.89% in a month — the SEO playbook on a zero-cost personal blog: sitemap, hreflang, descriptions, GSC, redirects.
-
RPA 40h/Month Savings — A Requirements Pattern That Works
A real RPA engagement that delivered 40h/month savings: target selection, As-Is/To-Be decomposition, Power Automate design, and operations built for handover.
-
Cutting Regression Test Effort 90% — A 3-Layer Pattern
A real engagement: regression test effort cut by 90% via selection, automation, and operations. Four lenses for what to keep, plus CI patterns and migration.
-
An Objective Self-Portrait — Five Strengths, Five Weaknesses
Subjective intro is on the profile page. Here is the third-party fact-based view: five strengths and five weaknesses with evidence, plus what I am improving.
-
From Funeral to Adventure: A Manager's Real Art of Asking
A manager's words decide if meetings feel like funerals or adventures. From The New Art of Asking: expedition mindset, four rules, the see-build-ask cycle.
-
Anything But Absolute Yes Is a No — Essentialism in Practice
Essentialism: anything but absolute yes is a no. The pickle jar, sleep as the foundation of mastery, and the courage to stop. Applied to engineer life.
-
Launching Tasukiba — an AI Operations Secretary That Passes Knowledge Like a Baton
My fourth personal product, Tasukiba Knowledge Relay, enters trial operation: an AI operations secretary for project management built around knowledge management. The insights one team accumulates get handed to the next like a relay baton, surfaced again later by semantic search. Here's the backstory and the world I'm aiming for.
-
Claude Prompt Mastery — complete guide index across 9 posts
Wrapping up a nine-day series on Claude prompting. Built on the Anthropic official docs, the series covers three core principles, a workflow, and 42 ready-to-paste patterns. This index shows which post to read depending on what you're trying to accomplish.
-
Claude for life, not just work — 18 patterns for writing, career, emotions, and daily tasks
Claude is powerful in non-technical domains too. Here are 18 ready-to-paste prompt patterns for writing, career decisions, emotional support, and everyday tasks — with a developer's real-life examples of each.
-
Claude isn't 'just a coding AI' — 13 prompt patterns for research, learning, and decisions
Claude is a powerful partner for research, learning, and decision-making, not only coding. Here are 13 ready-to-paste patterns I use to accelerate my thinking, with personal examples from development and everyday work.
-
Claude Code-only features — Skills, Subagents, Hooks, Plan Mode explained
Claude Code has features the Web and API versions don't. This post organizes the most important ones — Skills, Subagents, Hooks, Plan Mode, Auto Mode — and shows how I combine them in my personal projects.
-
Writing instructions Claude gets right on the first try — CLAUDE.md, motivation, XML, few-shot
Are you repeating the same instructions to Claude every session? Anthropic's prompt engineering guide highlights five techniques that raise prompt quality across the board: CLAUDE.md, stating motivation, XML structuring, few-shot examples, and document ordering.
-
Delegate development to Claude — 11 copy-paste prompt patterns for coding, review, and debugging
When you hand off development tasks to Claude, what kinds of prompts reliably produce high-quality code? Here are 11 battle-tested patterns I use daily for implementation, review, debugging, refactoring, test generation, and more.
-
Never let Claude jump straight into code — the Explore → Plan → Code → Commit workflow
The Claude Code official best practices recommend a four-stage workflow: Explore, Plan, Code, Commit. Here's why skipping Plan Mode produces 'correct solutions to the wrong problem', and how the discipline has saved me hours in practice.
-
Treat Claude like a brilliant new intern — the three principles Anthropic prioritizes
Anthropic's official prompt engineering guide distills everything down to three principles: the Golden Rule, context-window management, and always providing a way for Claude to verify its own work. Here is what those principles look like in daily practice.
-
The Claude Code prompt playbook was hiding in the official docs — kicking off a new series
Starting a series built on Anthropic's Claude Code best practices and the Claude 4 prompt engineering guide. Here's the overview and a quick look at how reading the official docs concretely changed my day-to-day development.
-
From Built-and-Never-Used to Three Apps at Zero Cost
Engineer who built local-only apps met Claude Code, cleared the server hurdle, runs three products at zero cost — and the constraint-driven design behind it.
-
AI That Implements Without Hesitation — How to Write a 20-Section, 3,500-Line Design Doc
Publishing the structure of the design doc that reduced 16 person-days to 2 hours in AI-driven development. The roles and priorities of 20 sections, and the 'must-write' points for each.
-
16 Person-Days of Development Done in 2 Hours — Design Doc Completeness Determines Speed
In a personal project, AI-driven development completed 16 person-days of work in 2 hours. The biggest factor wasn't AI speed — it was the completeness of the design document that eliminated all decision-making.
-
The Power of Small Contributions — Reflections from Tokyo Blaze Symphonic Band's 13th Concert
Attending the Tokyo Blaze Symphonic Band's concert for the third consecutive year, I witnessed how individual efforts converge into a single piece of music — and what it taught me about team building and psychological safety.
-
Don't Bring Motivation to Work — Deliver Consistent Results Through Systems, Like Brushing Your Teeth
Motivation-driven action inevitably burns out. Understanding how self-preservation instincts block change, and how systematizing actions like brushing teeth enables consistent output regardless of mood.
-
From "What Is Git?" to GitHub Actions — The Big Picture New Engineers Wish They Knew
A step-by-step guide covering Git, branches, GitHub, GitHub Actions, and GitHub Pages. Includes real CI/CD configurations (auto-deploy, scheduled publishing, stress testing) from two personal projects with actual code.
-
Qiita CLI × Claude Code — Automating Article Management with Trend Analysis, SEO, and Scheduled Publishing
I built a system combining Qiita CLI, Claude Code, and GitHub Actions to manage articles entirely in Git. This covers the full journey including a duplicate article ID trap and data-driven publishing time optimization.
-
"Essentialism" and "Effortless" — The Two Wheels of What to Do and How to Do It
Reading Essentialism (what to do) and Effortless (how to do it), I explored how to focus on essential tasks and build systems to make them easier. Applying the Pickle Jar Theory, 10,000-hour rule, and broaden-and-build theory to engineering work.
-
"Black Box Thinking" — Life Is Too Short to Experience Every Failure Yourself
Reading 'Black Box Thinking' revealed how ego blocks learning from failure. I explore the mechanisms of self-esteem-driven denial and how engineers can build systems to detect, report, and leverage failure.
-
Why I Built This Homepage — Walking with Astro v6
Why I built this homepage and what I poured into it. Plus the characteristics of Astro v6, a framework still under-documented in Japanese.
-
Get to Know Teppei Suyama in 5 Minutes — A Guide to This Site
A navigation map for first-time visitors to quickly understand who Teppei Suyama is — career, dreams, philosophy, side projects, and the shortest routes to each section of this homepage.
-
What Money Really Is — Connecting People
An essay on what money really is. Money has no inherent value; it pays for human labor. With Zimbabwe hyperinflation, an engineer view on why value attaches.
-
What Reading Gave Me Wasn't Knowledge — It Was the Habit of Thinking from Multiple Angles
The greatest value from reading isn't the volume of knowledge gained, but the ability to interpret things from multiple perspectives. Here's how the self-questioning habit cultivated through reading benefits code reviews and requirements analysis.
-
"No Effort Goes Unrewarded" — How the Concept of Accumulating Luck Changed My Work
I used to believe some efforts are wasted. A single book changed my definition of effort: time spent for others accumulates as luck. Here's how this shift transformed my approach as an engineer.
-
Janet's Law — How to Counter Time Feeling Shorter as You Age
Janet's Law: why each year feels shorter as we age. Countermeasures using new experiences and continuous learning — stretching subjective time as an engineer.
-
"Words Are Blades" — How a Single Phrase Can Stay with Someone Forever
Inspired by a line from Detective Conan — 'Words are blades' — I reflect on the dual nature of language: the power to uplift and the power to wound. As an engineer, I practice 'maturing words before speaking.'
-
INTJ-AS (The Lone Grand Designer) — My Working Style Through Personality Assessment
My extended MBTI (64-type) result is INTJ-AS. I've articulated my communication style, team behavior, and growth areas as self-disclosure for future collaborators.
-
"The Testing Phase Disappeared" — AI-Driven vs Traditional Development Compared
A developer with zero Flutter experience used Claude Code to ship a production app in 3 weeks. This article compares every phase — from planning to testing — with real data showing 6-9x faster delivery and 6x higher test density.
-
YumeHashi v2.1.0 — 4 Things I Did to Stretch Firestore's Free Tier
Running my Flutter Web app YumeHashi at zero monthly cost, I implemented 4 cost optimization measures in v2.1.0 — gzip compression, extended debounce, compact JSON, and size monitoring — to delay exceeding Firestore's free tier.
-
From "YumeLog" to "YumeHashi" — Building a Bridge Between Dreams and Reality
The story behind renaming my personal app from YumeLog to YumeHashi, the 3-step design philosophy (write, break down, keep going) for turning dreams into action, and why I keep building despite zero revenue.
-
YumeHashi's Tech Stack and 5 Implementation Challenges Solved by v2.1.0
A deep dive into the serverless architecture (Flutter Web + Drift + Riverpod + Firebase) behind YumeHashi, a side project running at zero monthly cost, and 5 real implementation challenges solved with code.
-
I Launched My Homepage — Consolidating Scattered Information into One Place
I built a personal homepage with Astro v6, GitHub Pages, and AI-driven development. Here's why I consolidated my career, skills, products, and blog into a single site, and what I hope to achieve.