DEV Community: koshirok096

A Few More Months with Claude — Some Scattered Thoughts (Bite-size Article)

koshirok096 — Fri, 31 Jul 2026 20:05:03 +0000

Introduction

Earlier this year I wrote a few posts about switching from ChatGPT to Claude, then getting into MCP, and various odds and ends I noticed after making the move.

Quick background, in case you're new here: I'd used ChatGPT as my main AI tool for a long time, but for various reasons I switched my main tool over to Claude around spring 2026. (I haven't quit ChatGPT, by the way — I still use it as a secondary tool.)

It's already been a few months since I moved over to Claude. (Time flies 😅)

I happened to be looking back over one of my older posts, and now that I've got a few more months of use under my belt, some new things have come into view — stuff I couldn't have noticed back when I wrote that earlier post. So this time I want to touch on that a bit. This isn't so much a tidy conclusion as just my scattered impressions right now.

Recent Impressions of Claude and ChatGPT

Claude

I previously wrote that the combination of the memory feature and MCP "cut down on explanation overhead quite a bit." That assessment still holds. But the more I use it, the more I've started to see another side. For example, sometimes it'll force a connection to information stored in memory even when the topic is completely unrelated.

"Uh, that's... not relevant right now."

Moments like that feel more frequent than before. These forced connections add noise, and it can get a little annoying. It's the flip side of the convenience, but since memory just keeps accumulating, tidying it up and deleting things might become part of the routine going forward. (I haven't tried that yet, so it's a to-do for later.)

One more thing — and I can't be sure about this — but lately I feel like I'm running into a bit more hallucination, excessive agreement, and, as a result, a harder time thinking critically. Even with Opus 4.8 or 5.0 — the top class model if you set aside Mythos and Fable — I occasionally hit the kind of wild, off-base answers I rarely used to see. (Oddly enough, when I experimentally switch back to something like Opus 4.6, I sometimes feel the answers are actually better, which is strange.)

It might have to do with how well a given model fits the kind of thing I'm asking about. Or maybe I'm wrong and it's just my personal impression, or maybe it's something off in the model's design itself — I can't really tell. Specs change often, so it could just be a problem with how I'm using it. I'd like to keep watching for a while.

ChatGPT

Right now I'm using ChatGPT on the free plan. Back when I was weighing the move from ChatGPT to Claude, I agonized quite a bit over whether to keep my ChatGPT Plus subscription. But paying a fixed cost for something I'm no longer using as my main tool felt wasteful, so I decided to try going free.

The number one thing on my list of "what ChatGPT has that Claude doesn't" is image generation. Claude can handle things you can express through SVG or Artifacts, but it doesn't have standalone JPG image generation the way ChatGPT does (as things currently stand). In my case, I need this image generation in all sorts of situations, so I was worried about whether going free — with tighter limits — would actually let me generate the volume my work requires. (Or, put another way, I figured if I hit a wall here, I'd need to keep Plus after all.)

But when I actually tried it, I could generate a fair amount without hitting a stop. The quality looks unchanged too. That said, if I ever need a much larger volume of images in a short window, things might change.

For now, though, the volume I use for work is no problem, so I feel like staying on the free plan is fine.

On the whole, I mostly use ChatGPT for image generation these days, but it seems like new models and memory features keep getting added one after another. There are probably quite a few features I don't even know about.

As for the recently talked-about GPT-5.6, I've also seen reports that on some benchmarks it's putting up numbers that surpass Claude's Mythos class — the model that itself made waves around the world.

So my impression that "Claude is ahead" has faded compared to before. But in the end, both are evolving at a pretty rapid pace, so I suspect my impression will have shifted yet again six months from now.

Closing thoughts

Today was all over the place, but I've loosely gathered up some honest impressions of how AI has felt to use lately.

One thing that struck me is that my own workflow is changing even faster than the AI itself. Properly adopting and using MCP, for instance, dramatically changed how I run my work, and going forward I plan to keep trying out new ideas — not just MCP — as they come up.

If I read this back again in a few months, I might think, "Huh, so this is the stuff I was fretting about back then." This post, too, isn't a conclusion — I'm leaving it as a note capturing where things stand right now.

Thanks for reading!

Getting Lost on Purpose (Bite-size Article)

koshirok096 — Fri, 24 Jul 2026 21:10:56 +0000

Introduction

This is another short, light one, but I hope you'll stick with me. A little while ago I wrote a post on the benefits of walking, and today I want to talk about walking again — from a slightly different angle.

—

In that post, I wrote about the "benefits" of walking: how ideas tend to come more easily when you mull things over on the move, how getting some sunlight is good for you, that sort of thing.

Separately from that, there's a way of walking I've come to enjoy lately: deliberately going out to get lost.

The Process, Not the Destination

I don't know the strict definition of a "walk," but the way I see it, it's less about getting to some particular place — that would be transit — and more about the walking itself being the point. So the interesting part isn't where you end up, but the process of getting there.

I feel that most strongly when I set out to get lost on purpose. The method is dead simple. Whether I'm traveling or just in the place I currently live, I start walking without deciding where I'm going. I don't open the map. Relying only on my own sense of direction and whatever signs happen to catch my eye, I turn down whichever way looks interesting. That's it.

When I do, the way the city looks changes noticeably. I naturally start picking up on the little details I'd normally walk straight past while staring at a screen — the bakery on that corner, a strangely steep hill, the flow of people. Not knowing where I'll end up turns out to be far more fun than I'd expected.

Why Do I Get Lost on Purpose?

Why do I do this? Honestly, there's no big reason.

For one, it's just a good way to kill time. For another, taking a different turn each time is somehow more exciting than walking the same route over and over. That goes double in an unfamiliar city — walking along with no idea what's around the next corner is a little bit of fun all on its own.

Thinking about it now, maybe it's a leftover habit from my backpacking days. When I was younger, I did a lot of backpacking abroad. Back then smartphones either didn't exist or were only just appearing, so I'd jot destinations onto a paper map in pen, walk by compass, and get lost all the time. But funny enough, what stands out most when I look back isn't the routes I nailed efficiently — it's the moments when I'd wandered somewhere by mistake. Though that's all reasoning after the fact; the truth is probably closer to "because it's just kind of fun."

Switching Between "Transit" and "Strolling"

That said, I'm not walking around ditching the map all the time. When I actually need to get somewhere — "transit," basically — I use the app fully, like anyone would. It's the fastest way, and I think that's exactly what you should do in that case.

What I'm doing, I guess, is consciously switching into a "stroll mode" now and then, separate from that transit mode. Not to reach a destination, but just to wander freely, I close the map for a little while. That alone makes even a familiar city look a bit different — and it's really not bad.

Wrapping Up

We've reached a point where you don't get lost anymore unless you make a conscious effort. Which is exactly why, every so often, I deliberately get lost and enjoy the situation when I do.

And on top of that, the "benefits" I mentioned last time — thinking more clearly, getting some sunlight, getting some exercise — you get all of those from this kind of aimless wandering too. You're still walking, after all. It's just that a little thrill of not knowing where you'll end up gets added on top.

It doesn't have to be somewhere you're traveling — even in your own familiar hometown, deliberately picking a street you don't know can turn up some surprising little discoveries. If you have the time, why not try getting "lost" — in a city you'll visit next, or in the one you already live in?

Thanks for reading.

Tackle Your To-Dos by Borrowing Another Plan (Bite-size Article)

koshirok096 — Fri, 17 Jul 2026 20:46:50 +0000

Introduction

There are sometimes small tasks that, for some reason, you just can't bring yourself to do — tasks that keep sitting on your to-do list. You know you have to do them, but somehow you end up prioritizing other things, and they stay there, undone.

The other day, one of those lingering tasks got finished in an unexpected way. The feeling at the time was oddly familiar, so I looked into it afterwards, and I wanted to write a bit about what I noticed.

The task that suddenly got done, 10 minutes before heading out

The other morning, I was working through my routine of planning and organizing my daily task list. On that list was a small task that had been sitting there for several days. The priority was fairly high, it would only take 10–20 minutes, and there was nothing especially difficult about it. But because I'd been busy, I kept saying "tomorrow's fine," and it just kept getting pushed back.

That evening, I had plans to meet up with a friend at my place and head out together. I had set "leave at X:XX" as the departure time. I wasn't really conscious of it, but I'd actually finished getting ready a good while before that, and almost without thinking, I found myself saying, "I've got a solid 10 minutes before I have to leave — let me do that thing now," and I started.

What was interesting was that, at the same time, I caught myself thinking, "I'm not even sure I'll finish in time" — and yet my hands were already moving, and as it turned out, I finished right on time. If I'd had a little more time, I probably wouldn't have started. And if I'd had a little less, I don't think I would have either.

What was actually happening

A task I'd been treating as "do it if I can, otherwise tomorrow" suddenly got done, just because it landed in that 10-minute window right before heading out.

Thinking about it a bit more, I realized: what I'd really done was finish something I had to do by borrowing another plan.

The plan to go out with my friend had nothing to do with that lingering task. But because that plan carved out a defined block of time, the task got done. It wasn't that I summoned up the willpower to tackle the task itself — it was more that I rode the current of another plan, and the task came along with it.

A "relative" of the concept: Habit Stacking

When you think about it this way, this is actually something we already do unconsciously all the time.

We check email on the commute train. We read a book during the commute. We brush our teeth while listening to the news. We routinely take this approach of "attaching a new action to an existing plan or habit."

There's a related concept from behavioral science called "habit stacking." It became widely known through James Clear's Atomic Habits, and the idea is that when you want to build a new habit, you can make it stick more easily by attaching it right after an existing habit. "After I brew my coffee, I take my vitamins." "After I brush my teeth, I learn one English word." You use an existing routine as an anchor. That familiar feeling I'd had, I realized, came from having encountered this concept before in that book.

That said, my experience isn't quite the same thing. What got done was not a habit but a one-off lingering task. Habit stacking, in its original sense, is about layering one recurring habit on top of another recurring habit, which is a bit different from layering a one-off task on top of a one-off plan, as in my case.

But I think the underlying principle is close — close enough that they feel like "relatives." In the sense of "instead of building a deadline or container yourself, you borrow the structure of another plan that's already locked in," both approaches seem to belong to the same family.

Even one-off plans can be borrowed

What felt like a personal discovery this time was: even non-habitual, one-off plans can be used as "anchors" in this way.

Normally, when we try to tackle a lingering task, we try to muster motivation directly, or we set deadlines for ourselves. "I'll do it tomorrow morning." "I'll finish it by evening." But deadlines we set for ourselves are just promises to ourselves, and they turn out to be surprisingly easy to break.

On the other hand, plans like "the time I leave for shopping" or "X o'clock pickup for the kids" are locked in by circumstances outside our control. They can't easily be moved. And because they can't be moved, borrowing them as deadlines can be much more effective than ones we set ourselves.

Whether I can consciously use this to actually push through tasks I'm finding heavy is another question. But at least for me, it feels like a promising idea worth experimenting with going forward.

Closing

When you're sitting on a lingering task, sitting down at your desk and telling yourself "OK, time to do this" surprisingly doesn't work all that well. At least for me — if that worked, the task wouldn't be lingering in the first place.

Instead, take a look at the other plans you already have locked in for the day, and try slotting the task into something like "the 10 minutes before heading out" or "the 20 minutes before meeting someone." A one-off version of habit stacking, where you tuck a lingering task into another plan. That kind of approach, I think, might suit certain kinds of tasks.

This post is also a bit of a note-to-self. If anything in it happens to be useful for your own life or work, I'd be glad.

Thanks for reading!

Putting "Doing Nothing" on Your Schedule (Bite-size Article)

koshirok096 — Fri, 10 Jul 2026 21:08:42 +0000

Introduction

When we plan our days, we usually write down what we're going to do. Meetings, errands, exercise, study. On the other hand, we rarely put "doing nothing" on the schedule.

I'm self-employed, and I tend to lean a bit toward being a workaholic. Since my work happens to be something I genuinely enjoy, if I let myself go, I end up working endlessly.

For someone like me, I've had a vague sense for a long time that I should actively plan time to do nothing. Rather than resting when I happen to feel like it, I wanted to build "doing nothing time" into my schedule from the start. I've been consciously trying to do this for a while.

Carving out time to do nothing, every day

One simple approach is this: every day, reserve a fixed amount of time for doing nothing.

There are two key points. First, fix the amount of time (set it as a daily target, like "X hours every day"). Second, don't decide what to do with it (don't plan the content in advance).

A normal plan specifies both the time and the content. "Go to the gym for 30 minutes starting at 7 PM," for example. The "doing nothing time" I'm talking about here only takes the time as a slot, leaving the content blank. The moment you decide what to do, it stops being "doing nothing time" — it just becomes another scheduled task.

The amount of time can vary from person to person. 30 minutes, an hour, several hours — whatever you can comfortably manage given your work and life.

The timing (when in the day) can be either fixed or flexible. You could set it as "10 PM to midnight every day," for example, or leave the timing open and just track total minutes throughout the day, stopwatch-style, aiming for something like "3 hours of nothing per day."

I've been intentional about taking active breaks for a long time, partly as a defense against my workaholic tendencies. The specific approach has shifted over time, but the basic idea of "reserving a slot for doing nothing in advance" has stuck with me as one of the methods that genuinely works for me.

To be clear, I say "doing nothing time," but I don't literally do nothing during this time. I play games, read books, space out, go for walks — whatever I feel like in the moment. The reality is closer to "time away from things I feel obligated to do." The phrase "doing nothing" isn't exactly precise, but it's the wording that felt most fitting, so I'm using it.

Some related ideas from other fields

By the way, when I first thought about writing a blog post on this idea of "intentionally reserving time to do nothing," I started wondering whether there was any real basis for it. So I did a bit of digging — and it turns out that similar ideas come up in various fields.

For example, there's a Dutch word called niksen, which apparently means, literally, "doing nothing." Along with Danish hygge and Swedish lagom, it's been introduced in Western media in recent years as a way of framing "time spent without purpose" in a positive light. I found it kind of interesting that other languages have words pointing to similar ideas.

There's also a concept from neuroscience called the Default Mode Network (DMN). It's apparently a brain network that becomes active when a person isn't focused on any external task — in other words, when they're spacing out. It's said to be involved in creativity and memory organization. The idea that something is actually being processed in the brain during what looks like "nothing time" struck me as interesting too.

I've only read up on these lightly, so I can't go very deep, but it seems like the view that "unused time isn't a flaw" — or even that there might be some meaning in it — shows up in quite a few places.

Closing thoughts

When you're busy and have a lot on your plate, from a strict efficiency standpoint, spending 2 or 3 hours a day on "doing nothing" can look like a serious waste. That said, my own experience is that consciously protecting this time, as much as possible, every day, gives me a sense of mental breathing room.

So even if it's not something to enforce every single day, when you're feeling worn out or notice yourself running low on slack, deliberately carving out some "doing nothing time" might help shift things in a better direction. If that sounds interesting to you, give it a try.

Thanks for reading!

Replacing a vague habit with "time" and "amount" (Bite-size Article)

koshirok096 — Fri, 03 Jul 2026 21:08:30 +0000

Introduction

This is going to be a short and light post, but I hope you'll give it a read.
Recently I rethought how I handle deleting unnecessary files on my PC, so I thought I'd write about it.

—

How do you usually delete unnecessary files on your PC? I used to run on a vague sense of "I'll do it when I feel like it" or "I'll clean up little by little" (and I suspect a lot of people manage things in a similar way). In my case, the very idea of "organizing unnecessary files on my PC" was vague to begin with — I hadn't decided when I'd do it or how much I'd do. As a result, my PC was chronically accumulating data, always in a cluttered state, with me deleting things at random whenever I felt like it. A pretty sloppy way of managing things.

The intention was there, but in practice almost nothing happened. The only time I actually moved was when storage ran out and I couldn't save something. Completely reactive, in other words.

"Do it little by little" sounds low-friction and easy to sustain — it looks good on the surface — but since the when and how-much weren't defined, the days where I didn't do anything just piled up.

Replacing the rule with "time × amount"

So the other day, I tried replacing the rule with two concrete elements. I set the timing of the deletion to "every week," and the amount to delete to "1–3 GB." That way, both the frequency and the target are clearly defined.

The "1–3 GB" number comes from a rough estimate of how much data I add in a typical week. In my case I have a vague sense of that number because of the kind of work I do, so I just decided on it. If you don't have that sense, you could check the diff size from a backup tool, or — going more analog — note today's free space and check the same spot a week later. That should be enough to get a feel for it. The basic idea is simple arithmetic: if you delete more than what gets added, your free space gradually increases.

What I found interesting after trying this was that the axis of judgment flipped.

Under the old approach, I'd ask "is this file necessary?" and if it seemed necessary, I'd keep it. My default posture was tilted toward "keep," so when I hesitated I'd keep the file, and my hand would tend to stop moving.

After setting the quota, I started asking "which ones can I delete to hit this week's amount?" The default posture switched to "delete," so I started actively looking for candidates. Looking at the same files, the conclusion of whether to keep or delete probably comes out differently.
What I was actually doing was changing the default of judgment

Put more abstractly, this was a story about switching the default of my cognition.

The "weekly × 1–3 GB" rule looks like just a quota from the outside. But what I was actually doing was replacing the question I ask myself when looking at a file — from "is this necessary?" to "can this be deleted?" — which is a fairly internal piece of work.

When you ask "is this necessary?", the default becomes "keep." When you hesitate, the file stays, and not making a judgment effectively means "hold = keep." On the other hand, when you ask "can this be deleted?", the default switches to "delete." Hesitation puts the file into the "candidate to discard" pile, and not making a judgment means nothing moves forward, so things naturally start moving. Looking at the same file, just changing the first question you ask leads to surprisingly different conclusions.

So the rule, it turns out, might be best understood as a device that lets me avoid consciously flipping that switch every time. By setting a quota, I'm putting myself in a state where the default of judgment is automatically tilted toward "delete." That's how it feels, anyway.

Closing thoughts

This time the topic was "regularly reducing unnecessary data on my PC," which isn't a particularly important subject — just a small everyday thing. But what I found, personally, was that a tiny shift in how I framed the problem made a noticeable difference in the outcome. Deleting unnecessary data is just one example; I suspect the same approach can be applied to other things too. When a habit has become stuck in vagueness and stopped moving, applying the two axes of "time" and "amount" might be enough to get other things moving as well, I felt.

Thanks for reading.

When Old Things Take On New Meaning in the Age of AI (Bite-size Article)

koshirok096 — Fri, 26 Jun 2026 21:45:27 +0000

Introduction — On What I've Been Writing for Years

This is a follow-up to my previous post on Claude and MCP. Just sharing some recent thoughts.

Personally, I've always enjoyed keeping records and analyzing my own work. So for years, I've been logging my daily tasks, jotting down thoughts, hesitations, and impressions in notes. I've drawn on these records for reviews, analysis, and decisions on various projects.

The tools have shifted over time — Evernote, Notion, Logseq, Taskuma, and so on — but the habit itself, of writing notes into some app or tool, has stayed with me for years.

What Happened with MCP

I recently wrote about connecting Notion and Google Docs through MCP, and the results have surprised even me. I won't repeat the details here since they're in that post, but ever since I introduced MCP, the flow of information has accelerated dramatically. In particular, I'd been accumulating reviews, task management notes, and brainstorms in Notion for years, and letting Claude read all of this has shifted the meaning of what I'd previously written.

When I first started recording in Notion, it never occurred to me that it might be useful to AI. Of course — I had no way to imagine a time when AI would become this close to everyday life, used in this way. I was just writing for plain, analog reasons — "so I could look back later," "so I could organize my own thinking."

But the moment MCP made it all readable, the feeling shifted. It's as if my past self comes forward to help my current self. Claude answers my current questions while drawing on the reasoning behind old project decisions, or on impressions I'd noted at the time. I've had moments like that more than once now.

Thinking about it: the human brain's memory has limits — even the person who wrote something forgets it quickly. That's why I kept taking notes, leaving behind my thoughts and conclusions at each point in time as a record. And now, in the flow of conversation, AI reads from those records, distills them into a form that immediately fits my current self, and bridges across to me. Whether for grounding decisions or raising the quality of brainstorms and project work — my past self's accumulation comes through directly. That's the structure underneath all of this, and I only noticed it recently.

What I Noticed

When a new technology appears, many people tend to focus on its newness. New tools, new techniques, the sense that we have to do something new with them. I'm the same way — I catch myself thinking I need to add something new on top.
But this time, in my own case, what actually worked was the opposite direction. Using a new tool to draw out the old accumulation I'd built up over years, and to work creatively with it. That was the shape of it.

It wasn't that something new worked — it was that something old took on new meaning. That came as a surprise, and I wanted to write it down.

Maybe this isn't just my story.

The things I've built up over time. Things that look old, things that seem to have lost their value. Maybe the value didn't disappear — maybe there just hadn't been a reader for them yet. And now, there's something that can read across all of it.

Closing

To be honest, over the past few years, watching AI develop so quickly, I've had this worry — that what I've been doing for years might get replaced, that what I've accumulated might have been wasted. A vague anxiety that's been sitting somewhere in the background.

But this time, with MCP, what I'd been collecting in Notion started to do more for me than ever before. It wasn't wasted, I found myself thinking.

For my own part, I want to keep this perspective going forward — alongside chasing new tools, asking how I make the most of what I've built up over time.

Thank you for reading :)

How I Finally Stopped Re-Explaining Everything to AI (Bite-size Article)

koshirok096 — Fri, 19 Jun 2026 20:58:46 +0000

Introduction

As I mentioned in my last post, I've recently started making more serious use of MCP and conversation memory in Claude. Neither of these are new features — they've been around for a while. I just finally got around to actually using them.

The changes turned out to be subtle but significant. My relationship with AI, or the sense of distance between us, shifted more than I expected.

At the core of it, I think it comes down to the elimination of explanation cost — the effort of having to re-explain everything from scratch every single time.

I wrote about MCP in my previous article, so consider this a continuation of that.

(MCP is roughly a way to connect Claude to external services. Conversation memory lets it carry context from previous sessions into new ones. That's enough background to follow along.)

Back When Copy-Paste Was Enough

For a long time, my approach was simple: just copy and paste whatever context I needed into each session. I was curious about MCP, but honestly, I didn't think the lack of context continuity was that big a deal. Yes, copying things over was a bit of a hassle — but you get used to it. It was just part of the workflow.

Then I actually started using both features, and I noticed things I hadn't before.

Re-explaining things introduced subtle drift. Last year, I was working through the detailed specs of a project with AI — things like database schema, naming conventions, and operational rules. When I had to start a new session, the broad strokes carried over fine, but the fine details kept slipping. The AI would interpret things slightly differently, and I'd have to correct it each time. It was a lot of small friction that added up.
I had to re-establish the same foundational context every time. Who the project was for, why I was building it, what direction I was taking it — even when I'd already shared all of this in a previous session, the next one would only have a fragmented understanding. Not zero, but close enough that I'd spend a chunk of time filling in the gaps again.

Looking back, it was tedious work. The kind you don't really notice until it's gone.

What Changed

Now I have Notion and Google Docs connected in read-only mode. As long as I've documented a project's direction, naming rules, and key decisions, Claude can reference that in any session — and the context stays consistent.

There's a prerequisite to this working, though. I've always had a habit of writing things down, even for solo projects. The idea of plans drifting or decisions being forgotten bothers me, so I keep my own documentation in Notion as a matter of course. That habit turned out to be the foundation that made MCP actually useful. Connecting MCP is only half of it — it needs something worth referencing on the other end.

Conversation memory has also had more impact than I expected. When I ask "didn't we talk about this before?", it usually remembers accurately. And sometimes it picks up on context without me even asking — it just knows, and the conversation moves faster because of it.

Things I used to explain every time are now things I don't have to explain at all.

It Felt Like Talking to Someone — Until the Session Ended

When two people talk, the conversation carries forward. You don't start from scratch every time you meet. But that's exactly what AI used to require. Every new session, the previous one had never happened. Same personality, same tone — but no memory of you. You had to re-introduce yourself every time.

That gap was particularly noticeable because the in-session experience felt so natural. Mid-conversation, it really did feel like talking to someone. But the moment a session ended, that sense of continuity broke.

With MCP and conversation memory, that gap has largely closed. The starting point of a conversation has changed. Before, I'd spend the first part of any session getting Claude up to speed on who I am and what I'm working on. Now I can start from a place of shared understanding — and go deeper, faster.

Closing Thoughts

Since I started using both features, I reach useful answers more quickly, and there's less back-and-forth to get there. The density of each conversation has gone up.

That said, it takes some active management. Conversation memory can sometimes over-connect things — drawing links to past discussions that aren't really relevant. And the documents I've connected via MCP can go stale if my thinking has evolved. With both, I've found it's important to stay aware of whether what Claude is referencing still reflects where I actually am. This isn't a passive tool. You have to know where you want the conversation to go.

One more thing: I still haven't given Claude write access to anything. Notion and Google Docs are both read-only. And yet the difference in experience is real. I don't feel any urgency to change that — for now, this is more than enough.

I'm still early in using these features, so I don't think I've found the final answer yet. I'll keep at it, and write again if anything new comes up.

Thanks for reading :)

From "Asking AI" to "Delegating to AI" — Trying Out MCP (Bite-size Article)

koshirok096 — Fri, 12 Jun 2026 19:54:42 +0000

Introduction

A while back, I wrote a post about switching my main tool from ChatGPT to Claude. It's only been a few months since I made Claude my primary AI, but I've been really happy with what it can do, and I've grown quite fond of it.

I'm still in the trial-and-error stage, but recently I got curious about Claude's "Connector" feature and started experimenting with it. As a simple first step, I connected my Notion and Google Drive — and the experience changed my relationship with AI more than I expected. So I figured I'd jot down a few notes, partly as a memo to myself.

This is fairly introductory content, so it may not be useful for those who are already familiar. But if you're anything like the past version of me — someone who has heard of MCP but doesn't quite have a feel for it — this might be worth a read. Feel free to stick around if you're curious.

What is MCP?

In short, it's a common standard for connecting AI to external tools and data.

Here's an analogy. Imagine you have an SSD full of data. As long as it supports the USB-C standard, you can plug it into just about any device with a single cable and access the data. MCP plays a similar role: as long as a service supports it, Claude can read directly from Notion or reference files in Google Drive.

From "Asking" to "Delegating"

This is just my personal take, but until now, whenever I asked AI for something, the process went like this: organize the situation in my head, summarize it or copy-paste the relevant info, and then ask the question. Once I connected MCP, the AI started fetching the context on its own.

It sounds like a subtle change, but the difference in feel is pretty significant. I had actually known about the concept of MCP for a while, but I'd dismissed it with "if I need to reference something, I can just paste it in and explain, right?" Once I actually tried it, though, the back-and-forth of information became dramatically faster — and as a result, all the work that I should be moving forward (brainstorming, planning, tasks) sped up significantly.

Also, at some point — I don't remember exactly when — Claude gained the ability to reference past conversations (currently only on the Pro plan). With most of the upfront context-setting no longer needed, the pace of pretty much everything went up another notch.

I get a sense that this is a step beyond what AI used to be — not just a "tool to talk to," but more like an assistant or a partner who works alongside you. It might be an overstatement, but that "wait, this is different from before" feeling I had when I first started using it has stuck with me even now.

What I've Actually Tried

Just a few concrete examples. Here are some of the ways I've been using it:

Weekly reviews: Once a week, I keep a record of the previous week's data (logs of what I did, daily notes, work memos, and so on). Before, I would manually go through them every Sunday, clean up notes, and summarize. Now I just point Claude to the right database and ask, "Read through this week's journal and summarize it," and it's done. The whole copy-paste-and-organize step disappeared. Of course, hallucinations and AI mistakes do happen, so it's safer to look over important parts yourself — but those cases are rare, and even when verification is needed, it's still far more efficient than doing everything by hand.
Digging up past decisions: When I ask things like "I remember deciding something like this a while back — is the rationale recorded anywhere?", it'll search for related pages and pull them up. The task of searching through my own Notion has, in a sense, been handed off to AI.
Data analysis: I often keep work management sheets and data in Google Sheets. This kind of analysis is right in AI's wheelhouse — depending on the data, it's quite handy for spotting trends or thinking through strategy.

None of these are flashy use cases, but the small daily frictions slowly disappearing adds up in a way that's quietly powerful.

About Permissions

For now, I'm a bit nervous about AI accidentally deleting or editing important notes, so I'm running everything in read-only mode with no edit permissions.

Depending on what I'm working with and which tools I connect down the line, I might grant edit access later — but even with read-only, it's genuinely useful.

Conclusion

I've only just started, so I'm still getting a vague sense of how useful this really is. I'll keep experimenting, and if I gather enough material, I might write a follow-up.

Thanks for reading to the end.

Why a Short Task List Can Still Feel Heavy (Bite-size Article)

koshirok096 — Fri, 05 Jun 2026 20:39:46 +0000

Introduction

I could not really think of a topic to write about today, so I would like to write a short memo based on something I recently experienced and what I learned from it. I am not sure if it will be useful for everyone, but if you are interested, I would be happy if you read it.

Recently, I had a day with many tasks that required brainstorming. More specifically, they involved things like coming up with ideas, planning, and designing the structure of future work.

I had already organized what I needed to do, so the tasks were neatly listed in my task list. If I only looked at the number of tasks, it did not look like too much.

I thought, “I should be able to move through this in one day.” But in reality, even by the evening, I had not finished everything. I had to move one task to the next day, and several other tasks were only partially done rather than completed.

Tasks That Are Easy to Estimate and Tasks That Are Not

What I realized again from this experience is that it can be risky to estimate the weight of a day only by the number of tasks.

This may sound very obvious. However, when tasks are lined up one by one in a task list, they can start to look like they all have a similar weight. At least in my case, I think I was unconsciously seeing them that way.

Some types of work are easy to estimate.

For example, a one-hour meeting, work that is scheduled for a fixed number of hours, or simple tasks with clear steps have a relatively visible time frame.

On the other hand, tasks like brainstorming and planning are a little different.

After you start thinking, you may need to organize the assumptions, compare different options, or notice another issue along the way. The same applies to programming design or planning work. There are parts that you cannot really estimate until you actually start working on them.

In a task list, every task appears as a single line.
But in reality, their weight can be completely different.

Even a task like “check the materials” may finish quickly depending on the content, or it may take much longer if it requires judgment.

Tasks like “think about a project,” “decide the design,” or “decide the direction of an article” may look short in the list, but they can require a lot of mental effort.

In other words, having only a small number of tasks does not necessarily mean the day will be light.

There Are Limits to Estimation

Then, you might think that the answer is to estimate more carefully at the beginning.

Of course, some level of estimation is necessary. Rather than listing tasks without thinking, it is better to separate them into things like “this seems heavy” and “this seems quick.”

However, when it comes to thinking-based tasks, there are limits to estimation. This is because there are many things you cannot know until you actually start.

Sometimes, you only notice the complexity of the problem after you begin working.
Sometimes, another option appears while you are thinking.
Sometimes, it takes time just to organize what you actually need to decide.

In other words, at the estimation stage, the inside of the task may not be fully visible yet.

If that is the case, spending too much time on estimation can become another heavy task in itself.

Maybe the important thing is not to predict everything perfectly, but to plan with the assumption that some tasks cannot be fully predicted.

What I Can Do

Since perfect prediction is difficult, it may be more realistic to make plans on the assumption that some tasks are hard to estimate. For example, there are a few things I can do.

Create a buffer

On days with many thinking-based tasks, I can keep the schedule a little lighter.

Instead of assuming that everything will go according to plan, I can leave room for the possibility that something may take longer than expected.
Especially on days with tasks like brainstorming or planning, it feels safer to leave some open space.

Avoid putting too many thinking-based tasks into one day

If I put many tasks like brainstorming, planning, writing structure, or project decisions into the same day, the day becomes heavier than the number of tasks suggests.

Even if the number of tasks looks small, the overall load can be quite high if they require a lot of thinking.

Also, even if I manage to finish those tasks, they create a different kind of fatigue compared to finishing simple tasks. When my brain is tired, it can affect the quality of later decisions and work.

So, on days with many thinking-based tasks, it seems better to leave some extra room from the beginning.

Set a midpoint instead of aiming for completion

Thinking-based tasks can feel heavy if I assume that I must complete them on the same day.

For example, it may be more realistic to set a midpoint as the goal, such as:

Come up with three possible directions
Create a rough draft
Gather materials for making a decision
Organize what needs to be decided next

Sometimes, I start a task thinking it will finish quickly, but after beginning, I realize that it is heavier than expected. In that case, instead of trying to force myself to finish everything, it may be better to break the task down at that point and decide how far I will go today.

The goal is not necessarily to “finish” it, but to bring it to a state where I can move forward next time.

Mark tasks that are hard to estimate

Some tasks are easy to estimate, and others are not.

If I label tasks in advance as “hard to estimate,” it becomes harder to judge the day only by the number of tasks.

For example, tasks like brainstorming, planning, decision-making, research, and writing structure can have a label such as “estimation caution.”

Even that small label can help me notice, when building a schedule, that “this is only one line, but it may not be light.”

Conclusion

Today, I wrote about a pitfall in task estimation that can happen when I try to process tasks too mechanically.

Of course, I understand in my head that not all tasks in a task list have the same weight. If the content is different, the required time and concentration are different too.

However, when I am tired or when I make daily plans without thinking too deeply, I can end up deciding the day’s tasks without paying enough attention to those differences. At least, I feel that this was happening to me.

Especially with thinking-based tasks like brainstorming and planning, a single line in a task list can expand much more than expected. Even if the number of tasks is small, the day can become quite heavy if they require a lot of mental effort.

Personally, when I make a plan for the day and cannot finish it as planned, it can affect me mentally for a while. That is why I think daily planning is actually quite important. When I plan the day, I want to look not only at the number of tasks, but also at their weight and how difficult they are to estimate.

It is difficult to estimate everything perfectly.
But even just noticing, “This task may be hard to estimate,” can change the way I make a schedule.

Just because the number of tasks is small, it does not necessarily mean the day will be light.
I would like to keep this in mind when planning my days from now on.

Thank you for reading!

Knowing Doesn't Mean Feeling (Bite-size Article)

koshirok096 — Fri, 29 May 2026 19:55:01 +0000

Introduction - Knowing Doesn't Mean Feeling: Why Logic Fails to Stop Our Emotions

When I feel like work is falling slightly behind, I find myself adding tasks along the way, thinking "if I just push a little harder today, I can catch up." In that moment, I'm moving on feeling rather than solid reasoning — just a vague sense that it's doable if I try.

But looking back, that judgment often makes things worse. The added tasks don't get done and leave a psychological weight as a sense of failure, or I force my way through but the quality suffers.

The anxiety is driving the decision, and yet in that moment, I'm convinced I'm thinking clearly.

Separate from that, another example. There are days when I've checked my finances and confirmed that things are fine for the month. But then, a few days later, sales come in a little lower than usual. That's all it takes. Before I know it, thoughts like "I don't have enough money" and "this isn't going to work out" are growing in my head. The facts I confirmed not long ago seem to have quietly disappeared somewhere.

This isn't my own experience, but I sometimes hear about people who keep getting cosmetic surgery despite already being objectively attractive. The feeling of "still not enough" just doesn't stop. I think the underlying mechanism might be the same as the two examples above.

All three share the same quality: knowing something in your head doesn't stop the emotion from moving on its own. I tried to break down why that happens.

"Knowing" and "Feeling" Run on Separate Systems

When I looked into it, there was a framework in psychology that mapped almost exactly onto this. There's a framework popularized by Daniel Kahneman called Dual Process Theory.

The idea is that human thinking runs on two distinct systems: a fast, automatic, emotion-driven System 1, and a slow, deliberate, logical System 2.

The key point is that these two don't operate as equals. Most of our everyday judgments are led by System 1, with System 2 often following behind to rationalize the conclusion. "Knowing something in your head" is System 2 territory. But what drives emotions is System 1. They run on different tracks, so no matter how much you try to reason with yourself, the logic doesn't reach where it needs to go.

In the work example, the anxiety (System 1) had quietly mixed into the judgment of "I can handle this if I push a bit" (System 2). In the money example, the feeling of fear (System 1) was gradually overwriting a fact that had already been confirmed — "things are fine this month."

When Emotions Become Evidence

There's a related concept called Emotional Reasoning: the pattern of thinking where "I feel anxious, therefore something must be wrong." The emotion comes first and gets treated as evidence for a conclusion.

The money example is a clear case of this — the anxiety itself was functioning as proof that there wasn't enough money. The cosmetic surgery example might work the same way: the feeling of "still not enough" becomes the fact of "still not enough."

There's also something called Motivated Reasoning — unconsciously avoiding uncomfortable information, or automatically steering away from questions you'd rather not sit with. In the work example, there was probably some of this too: instead of honestly asking "can I actually handle this much?", the thinking moved straight toward the conclusion that it was fine.

Both patterns are less about willpower and more about emotions shaping the direction of thought — which is a more useful frame than blaming yourself for not being logical enough.

So What Actually Helps

If logic doesn't reach System 1, the approach needs to come from a different direction. Here's what I've tried that seems to work, at least somewhat.

Labeling
Putting words to what's happening — "I'm in an anxiety spiral right now." Research on affect labeling suggests that simply naming an emotion can reduce its intensity. There's something to it: when you can name what's happening, it starts to feel slightly less overwhelming.

Working through the body
Deep breathing, going for a walk, drinking water. It sounds basic, but System 1 is deeply connected to physical sensation. Changing the state of your body can shift the emotional state along with it.

Choosing the right moment
When you're exhausted or in the middle of a strong emotional reaction, System 1 is running at full capacity. Decisions and self-reflection made in that state tend to be lower quality. The ability to think "this isn't the right moment to figure this out" has to be built in advance, during calmer times.

Using external rituals
Writing things down, changing your physical location, doing a specific routine. Rather than trying to solve everything inside your head, letting the environment help shift your state.

Conclusion

I used to think understanding something meant it would change. It's not that simple.

That said, knowing the structure does make it a little easier to step back from self-blame. System 1 working this way isn't a flaw — it's how it's built. The problem, if there is one, is being carried along by it without noticing.

Understanding is just the entrance. That's roughly where I've landed. Using what's useful, where it fits.

Thank you for reading :)

Reference: Daniel Kahneman, Thinking, Fast and Slow

When Stress Disguises Itself as Rational Planning (Bite-size Article)

koshirok096 — Fri, 22 May 2026 21:09:03 +0000

Introduction

Hello. Today, I want to write a short article about something I recently noticed while working. This is partly a note to myself, and partly a reminder.
Over the past few weeks, I had been feeling a little overwhelmed by everything I needed to do.
Every day, I had the feeling that things were not moving forward the way I wanted. When I looked back on the weekend, I found that many of the things I had expected to finish were still unfinished. But I could not clearly tell what the actual problem was.
I vaguely felt that I needed to change something, so I tried adjusting my task management, changing my schedule, and tweaking different parts of my workflow. But nothing really changed.

Looking Back, I Noticed the Real Problem

So the other day, I decided to take some time and look back at several days that had not gone well.
As I reviewed them, I noticed a common pattern. On those days, I had almost always added something unexpected to my schedule during the day. And the trigger was usually the moment when I felt that I was already slightly behind.
In other words, every time I felt behind, I was unconsciously trying to catch up.

When I think about it now, this was a reaction to stress. But in that moment, I did not feel like I was stressed at all.
“If I make some progress on this today, I can recover a little bit of the delay.”

When I thought that way and acted on it, it felt like a calm and realistic decision.

I did not think stress was driving my judgment. I thought I was simply looking at the situation and adjusting my plan rationally.

Looking back, it seems obvious. But in the moment, I did not notice it at all.

The Loop I Found

As I continued reflecting on it, I started to see a pattern.

I fall a little behind
↓
I think, “I should be able to catch up a little”
  ※ At this point, it feels like a rational decision
↓
The plan becomes unrealistic, so I either cannot finish it or the quality drops
↓
I fall further behind
↓
I feel stressed again

The problem was not just falling behind itself. The real issue was that the stress created after falling behind disguised itself as “rational judgment,” and I kept pushing myself without realizing it.

Why Putting It Into Words Helps

At first, I thought the way to break this loop was to become the kind of person who does not feel stressed when plans fall behind. But that is difficult to do in a short period of time.

What felt more realistic was to pause when I noticed the stress.

More specifically, to put it into words.

“Maybe I am feeling stressed because I am behind.”
“Is this really something I need to do today, or am I just trying to catch up?”

When you feel overwhelmed by too many tasks, “putting it into words” may not seem directly related to the problem. It does not reduce the number of tasks. It does not erase the delay.
But in this case, the problem was happening at the cognitive level. Stress was mixing into my judgment unconsciously. If that is the case, the intervention also needs to happen at the cognitive level.

Putting the feeling into words is a way to make the stress visible. Once it becomes visible, it becomes easier to notice when I am being pulled into it.

Conclusion

When things are not going well, I quickly start thinking, “I need to change my method.” But this time, I realized that the issue was one step before that.

I thought I was making decisions, but I was actually being moved by stress. Noticing that “I thought I was being rational” may be the turning point.
The next time I feel the same way, I want to first put into words what I am reacting to. That alone may help me act a little differently.

As a side note, the thing that helped me notice this stress was a conversation with Claude. I think the same thing could probably happen by talking to another person.
When I tried to explain the situation I was stuck in, I finally started to see what I had been reacting to. I wrote in this article that putting stress into words makes it visible, but in fact, the reason I started writing this article was that I had already experienced that process.
Depending on the AI tool and how you use it, talking to AI can be a relatively easy option when it feels hard to talk to someone else, or when your thoughts are too messy to organize on your own. If you ever feel similarly overwhelmed, you might try explaining the situation to an AI and checking whether stress has quietly mixed into your judgment or actions.

Thank you for reading.

Learning the Term “Tool Sprawl” Helped Me Rethink My Distance from Tools (Bite-size Article)

koshirok096 — Fri, 15 May 2026 20:35:42 +0000

Introduction

Whenever I discover a new tool that might be useful for work or personal use, I like to try it. If something catches my attention, I often touch it at least once. At the same time, I am quite careful about whether I actually bring it into my main working environment.

When I try a new tool, there is often a moment when I think, “This might make my work easier.” But at the same time, I also start thinking about how to manage it, how to organize its usage, and whether it can really fit into my existing workflow. Because of that, I often use a tool briefly and then end up leaving it alone.

In the past, I sometimes felt that this meant I was not using tools well enough, or that I was being too conservative.

I tried something, but did not continue using it.
I created an account, used it a little, and then stopped.

Those small unfinished feelings sometimes stayed with me.

Learning the Term “Tool Sprawl”

Recently, I learned the term tool sprawl.

Simply put, it refers to a state where tools increase in an uncontrolled or disorganized way. Each tool may look useful on its own, but before you know it, you may have several tools with similar purposes, and it becomes unclear which one should be used for what.

For example, you may have multiple places to write notes. You may have several task management tools. Even with AI tools, there are now many options such as ChatGPT, Claude, Gemini, and Perplexity.

Of course, having more options is not bad in itself. By trying new tools, you may find something that fits you well. In fact, I also enjoy trying new tools.

However, I think the real issue is not simply the number of tools. The issue is that you have to keep thinking about how to use them properly.

Which tool did I write this in?
Which tool should I use for this task?
Among similar tools, which one should I continue using?
What should I do with the tools I am not using?

Each of these decisions may look small. But when they pile up, they take up more mental space than expected.

This is where the idea of _cognitive load_ becomes relevant.

Cognitive load, simply put, is the idea that there is a limit to how much information the brain can process at one time. It is a concept often used in educational psychology, and it is related to Cognitive Load Theory, which became widely known through John Sweller’s 1988 paper and related work.

Human working memory cannot process that many things at the same time. When information such as tool usage, storage locations, notifications, and workflow rules keeps increasing, the attention available for the actual work gradually decreases.

For example:

Time spent wondering, “Where did I write this?”
The mental cost of switching context every time you move between apps
Noise from notifications coming from multiple places
A vague anxiety like, “Am I using this tool properly?”

These are not the work itself. But they still consume mental resources.

Tools are supposed to make work easier. But at some point, managing the tools themselves can become part of the work.

That was the part that made the most sense to me when I learned the term tool sprawl.

I Am Not Seriously Struggling with It

To be honest, I do not feel that I am seriously struggling with tool sprawl.

Rather, I think I am the type of person who likes trying new tools, but is careful about making them part of my main workflow. For example, I used ChatGPT for a long time, but even after trying Claude, Perplexity, and Gemini, I did not immediately bring them into the center of my workflow.

Recently, as I wrote in another article, I started using Claude more seriously in addition to my almost ChatGPT-only workflow. But that was not an impulsive switch. It was a decision I made after thinking about it quite carefully.

I used to see this cautious tendency as a somewhat passive attitude. Maybe I should use more tools properly. Maybe I am falling behind new technology. I sometimes felt that way.

But after learning the term tool sprawl, my perspective changed a little.

Trying and Adopting Are Different Things

Trying a new tool and adopting it into your main environment are two different decisions.

Trying itself is not a bad thing. In fact, there are many things you cannot know unless you try them. So I think it is good to stay aware of new tools and try things that seem potentially useful, as long as it does not become too much.

However, if you try to “properly use” every tool you test, the burden starts to increase.

In that sense, briefly trying a tool and then leaving it alone is not necessarily a failure. If there is little harm, it may simply mean that you naturally filtered out something that did not fit you.

Also, even if you do not use a tool as your main tool right away, touching it once still has value. When you actually try it, you develop an impression or feeling about that tool. Even if you do not use it now, you may later face a different problem and remember, “Maybe that tool could help.”

In other words, trying a tool is not only about adopting it immediately. It can also be a way of keeping a future option somewhere in your mind.

I have written several articles on Dev.to about how I use tools in my daily workflow, including tips and hacks around tools such as Logseq, Obsidian, Notion, ChatGPT, and Claude.

But looking back, there was a lot of trial and error before those tools became part of my workflow. In some cases, I compared one tool with another, used it for a while, and gradually integrated only the ones that remained useful.

In other words, I have not continued using every tool I tried. Actually, there are many tools that I touched once but never really adopted.

Maybe the danger is not increasing the number of tools itself. The danger may be trying to integrate every tool you try into your work and life.

Conclusion

The reason I wrote this article was that I learned the concept of tool sprawl.

It felt like a vague feeling I had never fully put into words finally had a name. I like trying new tools, but I become careful when deciding whether to bring them into my main workflow. That caution may not simply be conservatism. It may be a natural response to avoid increasing cognitive load too much.

I think my own way of thinking is probably part of it too. I often wonder, “Is there a better tool?” or “Is my current way of working inefficient?” or “Should I try another tool?”

Of course, looking for better options is not a bad thing. But if I keep questioning my current environment every time, my attention may shift from using tools to constantly choosing tools.

So the next time I try a new tool, I want to first touch it lightly. Then, instead of judging it too quickly, I want to later ask myself, “Is this worth bringing into my main workflow?”

Trying something and adopting it.

Just separating those two decisions may make my relationship with tools a little easier.

Thank you for reading!