Handy 990
Technical Documents
In-depth looks at how Handy 990 works under the hood — the methods behind the practice.
Published · 2026-09-07
TOEIC Apps vs. Books and Ebooks: Which One Adapts Your Study Plan?
A good TOEIC app can offer a book's complete lessons, examples, and explanations, then adapt your study plan to your performance, goals, and available time. Paper still has practical advantages: fewer digital distractions, more room than a phone screen, and easy handwritten notes. Here is how the learning tools and reading experience compare—and why an app should try to understand what you need.
Read the article →Published · 2026-08-30
TOEIC study plans that plan themselves: how Handy 990 decides what you practice today
Open Handy 990 and the home screen lists today's work: a 23-minute smart session, 3 mistakes to review, 14 vocabulary cards, a 1/3-length listening mock. None of those numbers is random, and none of them is a black-box AI hunch — every one comes from a rule we can write down. This article opens the machinery: why the plan never asks for more time than you said you have, why the mock tests are a third of a full form, why your weakest Part gets the most minutes, and why the Parts you've already mastered don't disappear. Our position is simple: a plan you can read is a plan you can trust with three months of your life.
Read the article →Published · 2026-08-28
How Accurate Is a TOEIC Score Estimate? Four Test-Takers Sent Us Their Real Scores, So Let's Do the Math
The day the August 16 TOEIC results came out, four test-takers reported their real scores back to us — the first time Handy 990's score estimate has faced a real, graded exam in batch. The headline: listening was startlingly accurate (two of the four were off by exactly 5 points), reading carries three systematic biases that each have a name, and the estimate ranked all four people in exactly the right order. This post lays the four daily estimate trajectories against the four real score reports, and answers a harder question along the way: what does it actually mean when an estimate "converges"?
Read the article →Published · 2026-08-27
Your TOEIC Score Report Is More Than Two Numbers: What ABILITIES MEASURED Means, and How It Calibrates Your Estimated Score
Below the two big numbers on a TOEIC score report sits a row of bar charts most people glance at once and never again — ABILITIES MEASURED, ten percent-correct figures across ten skills. This article explains why those bars are the most information-dense part of the whole report, and what Handy 990 does when you enter your real result: it calibrates your estimated score — keeping the shape of strengths and weaknesses your months of practice established, and shifting only the overall level onto the official truth. We'll also own up to something: our first version got this wrong.
Read the article →Published · 2026-08-24
Memorizing TOEIC vocabulary with absurd stories: why a goat named Penalty beats flashcards
The hardest part of learning TOEIC vocabulary isn't understanding words — it's remembering them. Handy 990's new "Level Stories" module takes the two oldest tools in memory science — narrative structure and absurd imagery — and turns them into a feature: before you drill a vocabulary level, you read ridiculous stories built from it. Stories featuring a filing cabinet that wins the national lottery, a space heater that testifies at its own trial, and a goat named Penalty who saves an aluminum mill by screaming at a wall. This article explains why the nonsense is not decoration. It is the entire design — and it stands on some of the oldest, hardest evidence in memory research.
Read the article →Published · 2026-08-21
I Already Have ChatGPT. Why Would I Need a TOEIC App?
This is the question we get asked most, and it's a perfectly fair one. AI writes practice questions, explains grammar, and will chat with you until dawn. So what exactly is a dedicated TOEIC app selling? This essay is an honest answer — including an admission of the things AI genuinely does better than we do. The conclusion up front: AI is the engine, the app is the car. Having a superb engine in your garage does not mean you can drive it to the test center tomorrow.
Read the article →Published · 2026-08-19
Can learning speed reading save my reading section?
Everyone who has taken the TOEIC knows the moment: fifteen minutes left in the reading section, a full column of Part 7 passages untouched, and you switch to "finishing by guessing." The reflex is to go learn speed reading — after all, not finishing sounds like reading too slowly. This post takes that reflex apart: what traditional speed reading actually teaches, which parts survive scientific scrutiny, which action really stalls you in Part 7 — and the three drills we built into the web app accordingly.
Read the article →Published · 2026-08-15
It's 2026. Why Are We Still Memorizing Phonetic Symbols?
Kids around the world are still made to memorize phonetic alphabets before they're allowed to just listen to English. This essay wants to say, gently but firmly: phonetic notation was a brilliant invention of the printed-dictionary era — brilliant the way the pager and the fax machine were brilliant — and it has earned the same honorable retirement. When one tap on a phone plays a native speaker, shows you the mouth, and gives you AI feedback, why are we still teaching children a secret code for describing sounds instead of the sounds themselves?
Read the article →Published · 2026-08-11
How *Not* to Emotionally Blackmail Your Users — a Breakup Notice from a Duolingo Veteran
Duolingo has turned emotional blackmail — that's the term the internet itself settled on — into a business triumph, marching tens of millions of people back into the app every single day. I'm a 2,871-day-streak heavy user and long-time paying subscriber, so this isn't an outsider's cheap shot; it's first-hand testimony from a victim. This article looks at how the tactics work, why "it works" doesn't mean "it's right," and why Handy 990 chose to go the other way — we may even remove streaks entirely.
Read the article →Published · 2026-08-07
Beyond "here's where you're weak": how the study-direction recommendations work
You've reviewed hundreds of questions and your mistake list keeps growing — but what should you actually practice next? Handy 990's study-direction recommendations push weakness analytics one step further: instead of just drawing a radar chart, the app ranks the two or three things most worth doing right now, each with a one-tap shortcut into the matching practice. This article explains how the ranking works, and why setting a goal is where it all starts.
Read the article →Published · 2026-08-06
How Many Words Do You Need for TOEIC? We Did the Math — and Decided Not to Give You a Number
It's probably the question we get asked most, and the internet's answers run from five thousand to "20,000 must-know words." So we counted every single word in six full mock exams — 1,200 questions — and reached two conclusions. First: the number everyone quotes ("score X needs Y words") cannot actually be computed — not by us, not by anyone. Second: a different question turns out to have a surprisingly clear answer. Vocabulary is finishable, and the finish line is closer than you think. This post lays out the whole calculation, all the numbers, and how they changed the vocabulary features in the app.
Read the article →Published · 2026-08-04
What Moneyball Teaches Us: Don't Buy Players, Buy Wins — and Don't Count Questions, Count Points
The 2011 film Moneyball is nominally about baseball, but it's really about something more general: with honest measurement, analytics and engineering can be applied to almost anything — assembling a poor team on a rich league's budget, or pushing your TOEIC score up. This post is about the three things we shamelessly stole from the movie: redefine the goal, find the undervalued asset, and outsource the "what should I practice" thinking to data. Two clips are embedded below with English subtitles on by default — watch a movie, sneak in some listening practice.
Read the article →Published · 2026-08-04
In Defense of Drilling: Wittgenstein, Linguists, Memory Science, and AI Take the Stand
In East Asia it has a name — 刷題, "grinding through practice questions" — and a reputation to match: cram-school culture, exam-factory learning, the reason people "score high but can't hold a conversation." Real language learning, the story goes, happens somewhere else. This essay is a serious defense of drilling. Taking the stand: one of the twentieth century's most important philosophers of language, several linguists who actually counted, a century of memory science — and one surprise witness, the large language model itself. Verdict up front: drilling is not a detour around language learning; short of moving to an English-speaking country, it is the closest thing there is to the direct route.
Read the article →Published · 2026-08-04
Part 7 difficulty is measured, not guessed — and once you see the structure, you know how to answer
A technical explainer for the curious. The difficulty series runs one post per Part, published in order; this is the seventh and last. It takes on the hardest part of TOEIC to pin down — Part 7, reading comprehension — and explains how we turned its difficulty into something measurable, how we calibrate that ruler against a commercially published mock-exam set, and (most useful to you) how seeing exactly *how* a question was made hard tells you where to spend your time. No statistics background needed.
Read the article →Published · 2026-08-04
Part 6 difficulty comes down to "did you actually need the passage?" — a test that covers up the text
A technical explainer for the curious. The difficulty series runs one post per Part, published in order; this is the sixth. Part 5 difficulty focuses on how the four options are arranged, while Part 7 difficulty is scattered across the entire passage. Part 6 sits in between, and its difficulty comes from a place that is easily overlooked: <strong>does this blank actually require you to read the passage?</strong> This post explains how we turned that question into a measurable number, the method we used so this measurement wouldn't just be "one person's feeling," what the results were (the first measurement looked very bad for us, so we rewrote the bank based on it), and—most useful to you—how knowing that Part 6 contains two completely different types of blanks tells you how to manage your time on the exam. No statistics background needed.
Read the article →Published · 2026-08-04
Part 5's difficulty hides in the four options — the four measured patterns, and how to answer each
A technical explainer for the curious. The difficulty series runs one post per Part, published in order; this is the fifth. In the Part 7 post the difficulty was spread across a whole passage; Part 5 is the exact opposite — one sentence, four options, and the difficulty sits almost entirely in <strong>how the options are arranged</strong>. This post explains how we turned that into something measurable, how we calibrate it against a commercial mock set, how we measured our own instrument, found it wasn't accurate enough, and abandoned a change we had planned — and, most useful to you, what to do with each pattern once you can see it, and how many seconds it deserves. No statistics background needed.
Read the article →Published · 2026-08-04
The Difficulty in Part 4 Is Not in the Monologue, but in the Questions: Four Measured Question Types, and How to Answer Each One
A technical note for the curious. This difficulty series has one article for each Part; this is the Part 4 article. In Part 5, difficulty centers on how the four options are arranged. Part 4 is almost the opposite: the same monologue, paired with different questions, can differ by several difficulty levels. This article explains how we turned “what makes a question difficult” into something measurable, how the answer we originally expected measured out <strong>backwards</strong>, what we saw after comparing against a commercial practice-test set, and, most useful to you, once you understand the four question types, how to listen for each one and when to let go. No statistics background required.
Read the article →Published · 2026-08-04
What makes Part 3 difficult is “who said what” plus “the answer being reworded” — what the measurements showed, and the two things we changed because of them
A technical note for the curious. This difficulty series covers one Part at a time, in order; this is the third article. Part 5 concentrates its difficulty among four options, while Part 7 spreads it across an entire text. Part 3 dialogues sit between the two: a dialogue lasting less than 90 seconds comes with three questions, and the difficulty comes both from what happens while you listen and from how the options are written when you read them. Here, we measure our question bank item by item against all six tests in a commercial practice-test book — dialogue length, number of turns, number of speakers, and question-type mix — to explain where we match and where we are deliberately harder. The measurements exposed two gaps — our questions were too shallow, and the answers were too easy to guess — and we have now closed both: we rewrote 575 questions and 434 sets of options, and wrote another 108 new dialogues. No background in statistics required.
Read the article →Published · 2026-08-02
Part 2 difficulty: the hard part isn't understanding it, it's that understanding it isn't enough — three ways a question-response item is made hard
A technical note for the curious. One post per Part in the difficulty series; this is the Part 2 one. Question-Response is only 25 items, and the question plus all three options exist only as sound — not one word is printed. This post explains where the difficulty actually sits: not in whether you understand the English, but in the step between "I understood that" and "that counts as an answer". It ends with the part most useful to you: what to listen for in the first word, and when to raise your guard. No statistics background needed.
Read the article →Published · 2026-08-02
Part 1 Difficulty: it lives in a single instant of the verb — five ways to make photo questions hard, and how to listen for each
A technical explainer for the curious. Each Part gets its own post in this series; this is the Part 1 post. At just six questions, Photographs is the shortest part of the test, and the one most often dismissed as "just looking at a picture and listening to a sentence." This post explains where its difficulty actually lives — not in vocabulary, but in whether the verb describes a completed state or an action in progress — then compares our bank against 36 questions from a commercial mock set (yielding a confirmation we didn't expect), and ends with the most useful part: what to look at before the audio starts, and which word to listen for first once it plays. No statistics background needed.
Read the article →Published · 2026-08-01
Difficulty Overview: what does "as hard as the real exam" actually mean? — how we measure it, what we compare against, and what we don't claim
A technical explainer for the curious. This is the overview of our difficulty series; each Part then gets its own post. This one covers what all the later posts share: the formal definition of difficulty in testing, why nobody — us included — can measure it directly, what we use as an external reference, and the difference between what we are willing to claim and what we are not. Once you've read this, each Part's post can go straight to where that Part's difficulty actually comes from. No statistics background needed.
Read the article →Published · 2026-07-24
How Handy 990 helps you learn vocabulary: from frequency ranking to memory algorithms
Every vocabulary book in the world shares one urban legend: the first word is "abandon" — and that is roughly where most people abandon it. That isn't your fault; it's a scheduling failure. Working from A to Z was always an overestimate of human nature. This article walks through the machinery we built instead: each piece solves one specific problem, and together they form a loop.
Read the article →Published · 2026-07-23
How Handy 990 estimates your TOEIC score
A technical explainer for the curious. It assumes you've seen a little probability and statistics — you know what a probability, an average, and a standard deviation are — but nothing beyond that. Everything else is built up from scratch.
Read the article →