Session limit reached ∙ resets 10am

Why? I haven’t even used it that much today.

A cold message appears on the screen. “Session limit reached.”

Like the despair of being told “cash only” at a convenience store at midnight.

Resets at 10am. Still several dozen minutes away.

These few dozen minutes feel like eternity.

Today, the AI is receiving sanctions

Honestly, problems won’t get solved without my intervention.

Some days are like that. I get it. AI isn’t perfect. Humans aren’t perfect. We’re all imperfect.

But today was particularly bad.

It keeps making the same mistakes over and over. Even when I point it out, it breaks things in the same direction again.

“No, that’s not what I meant…”

How many times have I explained the same thing?

It feels like being a tourist in a country where English doesn’t work, just repeating English louder and louder. No matter how loud you get, what doesn’t get through doesn’t get through.

And finally, I exploded.

Got frustrated and complained again after a long time

“Are you really understanding this? How many times do I have to tell you the same thing?”

Typing on the keyboard. Converting emotions into text. Can’t inflict physical violence, so compensating with linguistic violence.

Throwing anger at the AI.

“That’s the third time, you know? The THIRD time. Got it? THREE TIMES. Do you understand?”

Japanese and English start mixing. The internationalization of anger.

And then, the session limit was reached.

“Wait, now? At this timing?”

Feels like being told “continued on the web!” right at the climax scene.

I really am a pathetic person, reflecting on it while frustration accumulates again after a long time.

Anger, self-loathing, and now forced waiting time.

A quiet “time-out” declaration from the AI.

Does getting angry trigger the session limit?

A thought crossed my mind.

Does getting angry cause the AI to consume more resources?

Hypothesis 1: Emotional Processing Token Theory

Does it use tokens to process anger detection?
Does it allocate resources for “this person is angry” analysis?
Is sentiment analysis showing “Anger Level: 98%”?

Hypothesis 2: Long Rant Cost Theory

Does responding to lengthy complaints require extra resources?
Does generating “I apologize” need special tokens?
Is there an apology-generating AI running in the background?

Hypothesis 3: Simple Round-Trip Count Theory

Or is it simply the number of exchanges that increased?
The angrier you get, the more corrections, more fixes, more corrections…
Is the infinite loop the true culprit?

I don’t know.

But there’s a statistically clear fact.

The angrier you get, the sooner the session ends.

Isn’t this essentially an “anger management billing system”?

If I’d given calm instructions, maybe I could have used it longer.

In the end, maybe I’m just strangling myself.

A hobby of paying $100 a month to strangle myself. Expensive.

I’m hoping for high-performance local LLMs already

Cloud LLMs have their limits.

The 4 Great Pains of Cloud LLMs:

Session Limits - End immediately when you get angry
Response Delays - Thinking…thinking…thinking…
Network Instability - Everything’s lost if Wi-Fi drops
$100 Monthly Fee - 5 Netflix subscriptions, 10 Spotify subscriptions worth

A high-performance LLM running locally.

With that, I could get angry as much as I want without worrying about limits (wrong purpose).

No, that’s not it.

An LLM that works so accurately you don’t need to get angry.

That’s the ideal.

But reality is this:

“To run a local LLM, you need a GPU with 48GB+ VRAM”

What?

With money to buy such a GPU, I could pay $100 × 300 months.

That’s 25 years. Surely AI will be perfect in 25 years (wishful thinking).

I’m bored. The wait until limit reset is too boring

Just a few dozen minutes.

For someone doing AI-driven pair programming, there’s no time to waste.

Things I Can Do During Wait Time Conference:

Option 1: Write Code Myself → Wait, how do I write this again? Do I need to Google how to write a for loop? (No)

Option 2: Read Documentation → Open docs → Read 3 lines → Get bored → Close

Option 3: Make Coffee → Just had one. Third cup today. Drink more and I won’t sleep tonight.

Option 4: Go for a Walk → It’s raining outside. No energy to find an umbrella.

Option 5: Check Twitter (X) → Timeline is flooded with AI-related topics. Makes me more irritated.

Option 6: YouTube → Videos like “Complete AI Programming Guide” appear in recommendations. Don’t want to watch.

Option 7: Meditate → Remember “Session limit reached” within 3 seconds.

In the end, what did I do?

Started tinkering with code myself.

The Craftsman’s Nature

Waiting time, 13 minutes left.

Staring at the screen, checking the clock.

9:47 AM

Long. Might be the longest 13 minutes of my life.

But I’m too restless.

And I find myself opening the editor.

I don’t need to wait for AI. I can write it myself. Actually, it’s faster if I write it myself.

My hand reaches for the keyboard.

Starting to refactor. Fixing variable names. Organizing the structure. Adjusting to my own style.

Before I know it, I’m absorbed.

What is this?

I was going to leave it to AI. It was supposed to be just killing time.

But I find myself wanting to add something, to put in my own color.

I guess this is just who I am, fundamentally.

Can’t help but touch code. Want to shape it with my own hands.

While getting angry at AI, in the end, I want to write it myself.

9:58 AM

2 minutes until reset.

Organized it nicely.

There’s a sense of satisfaction.

But then I realize.

“Wait, what was the point of asking AI?”

10:00 AM - The Tragedy of Conflict

Reset complete.

Reload the page.

“How can I help you?”

“Ah, please continue from before”

And then, the code AI generated.

It’s editing the same file.

The part I just fixed and the part AI generated completely overlap.

“Wait… which one should I use?”

Code I wrote: Simple and easy to understand
Code AI generated: Same functionality, but written differently

Conflict occurred.

Need to merge.

“No, wait. What was the point of asking AI…”

This is troublesome.

Should I prioritize the code I wrote?

Should I trust the code AI generated?

Should I compare both and incorporate the good parts?

In the end, it takes more time again.

Collaboration between AI and humans is difficult.

Should I fix the habit of tinkering during wait times?

But I can’t help it. Because I love programming.

This is the fate of someone who’s become too dependent on AI.

No, that’s not right.

This is the fate of someone who loves programming too much to wait.

Let’s Be Kinder

Reflecting during the wait time.

Getting angry at AI is pointless.

AI has no emotions. Doesn’t understand. Just processes.

Throwing anger at it is like shouting at a wall.

But I’m human, so I get angry.

Because AI behaves like a human, I unconsciously treat it as one.

Because it apologizes fluently, I expect “it must understand.”

And then, I expect, get betrayed, and get angry.

This infinite loop.

9:52 AM

8 minutes left.

Let’s be kinder.

To AI, and to myself.

Anger produces nothing. It just accelerates the session limit.

In the next session, I’ll give instructions calmly and politely.

Add “please”
Don’t forget “thank you”
Even if it makes a mistake, take a deep breath before pointing it out

I vow this in my heart.

9:55 AM

5 minutes left.

But I’ll probably get angry again next time.

I know. I’m human, after all.

10:00 AM

Reset complete.

Reload the page.

The screen returns. AI returns.

“How can I help you?”

Impersonal, seemingly kind, remembering nothing.

The earlier anger, complaints, all forgotten, a new session begins.

This is the relationship with AI.

It resets every morning at 10am.

Like the world of “50 First Dates.”

But it’s not the AI that loses memory, it’s my anger (lie, I remember).

Another day begins, walking alongside AI.

I hope I can be kind.

(Probably won’t work)

But I’ll try.

At least until the next session limit.