自作新聞(Tonio Times)とAIにTOEICを解かせた話

It's been a long day — TGIF, I made it through the week. I'm a bit fried today, so instead of winging it, I had an AI script the whole thing, and I'll just read it out in clean English.

Honestly, some days the words just aren't in you. So I cheated a little — let the AI write my lines and I'm reading off the page. Bear with me; the polished parts aren't really me talking.

I made my own newspaper — The Tonio Times. Circulation: one copy. I'm the reader, the editor, and the typos are on me too — except there aren't any, since the AI wrote it.

It sounds ridiculous, I know — a newspaper with a single subscriber. But hear me out: it's the only paper actually written for me, at my level, about what I care about.

I'm just a casual watcher — I don't root for any specific team. I only follow MLB these days, not Japanese baseball. But Ohtani's on the Dodgers, so his news is everywhere and I end up following them.

I'm no die-hard — no team I'd bleed for. But Ohtani drags me in whether I like it or not; his highlights are unavoidable, so by default I'm a soft Dodgers guy.

Every morning the news is the same — politics, a fight, someone died, an earthquake, the weak yen. Does any of it touch my life? Not really; I close it feeling worse.

So I built the paper I actually want to read — science, baseball, zero dark news.

The news is basically a machine for ruining your morning before breakfast. I wanted the opposite — something I close feeling smarter, not heavier. So no politics, no disasters, just my stuff.

It doubles as my Eiken prep — the science pieces are packed with the academic vocabulary the exam loves.

Two birds, one stone. I'm reading for fun, but those big academic words are exactly what the Eiken throws at you, so it's sneaky studying.

I even had to bump up the font size — on A4 it came out too small. There's a Japanese translation on a separate sheet so I can mine the vocabulary afterward.

Rookie problem — I printed it and could barely read my own paper, so up went the font. The back page has the Japanese, so reading quietly turns into a vocab lesson.

Second thing — and this one's better. I've been building an English app for about six months, blood, sweat, and tears, and my listening has jumped because of it. I put a full TOEIC mock test on it.

This is the one I'm proud of. Six months of grind — and weirdly, my own listening was the first thing to level up. Now it's got a full TOEIC mock sitting on it.

Honestly it's a boring test — dull business emails, easy questions. Loads of learners here are stuck on it, but for me it's pure business: I make myself the guinea pig and turn the whole thing into funny content.

The test itself is mind-numbing — but that's the point. Everyone here is chained to it, so I make myself the lab rat and spin my suffering into something people actually want to watch.

Right as I was about to publish it, I got this cold feeling. I wrote the questions, I wrote the answers, and I'm the one checking them — like a teacher who makes the test, makes the key, and grades it himself without being sure of his own answers. A bit sketchy.

It hit me at the worst moment — publish button right there. Grading your own exam when you're not even sure you're right isn't confidence, it's a liability. So I didn't trust myself.

So instead of trusting myself, I made an AI sit a real mock test completely blind — 100 reading questions, no answer key, like a student walking into the exam with nothing. It scored 99 out of 100.

So I flipped it — real official-style mock, answer key hidden, AI walks in cold. A hundred questions, no lifelines. Ninety-nine right.

That told me more than my own grading ever could.

The one it missed was a sentence-insertion question — the most subjective type on the test. So the machine only tripped up exactly where humans can't agree either. I loved that; it felt weirdly human.

And here's the kicker — the one it botched was the insertion question, the single most subjective one on the test. It only cracked where people crack too. Honestly, that made me trust it more, not less.

Exactly — number 1 is the answer. The AI went with 4 and then made excuses instead of owning the mistake.

Spot on — number 1. The funny part isn't that it got one wrong; it's that it doubled down and argued the toss with me. Very human of it.

What an era we're living in. This was impossible just a few years ago, and now anyone can spin up this kind of content in an afternoon.

It's a wild time to be alive. The thing that used to take a publisher a month, I do between coffees. Good or bad I can't tell yet — but I'll take the opening.

Anyway — mission accomplished, I got everything off my chest. You look a little tired, though; are you okay? At the end of the day, health is the only thing that matters, so get some rest.

That's me done — full brain dump complete. But you look wiped, so go easy. Seriously, health first — the rest is just noise.