Filotto
Cloze drills, and a nightly chapter checked against the words you have proved you know before it is sent.
Live The drill is live at filottolang.com. Nightly chapters are paused, and the project is parked until October while the other two Italian products get the attention.
Graded readers hand every learner at a level the same book, which means the vocabulary is right for a statistical average and wrong for you. Filotto measures the words you personally have proved, then refuses to send you a chapter that fails against that list.
What it does
Filotto has two halves. The drill is cloze and reverse cloze over a large sentence corpus, scheduled so that you keep meeting words near the edge of what you can already do. The reader is a chapter written fresh each night and delivered as a link to your own page, where tapping a word gives you the gloss and, in the same motion, tells the system you did not have it.
A gate sits between the two. Before a chapter is sent it is measured against your known-word list, and only words you have proved count towards coverage. The statistical estimate of your vocabulary, useful as it is for placing you on day one, never gets spent as free vocabulary in prose. A chapter that scores below the floor does not go out; it is rewritten until it clears.
That rule is strict, and the rewriting is what it costs. It is also the whole promise. Because every chapter is checked against what you know before it is sent, what lands is readable when it arrives rather than aspirational, and a reader who has been let down by a graded book can tell the difference by the end of the first page.
Who it is for
For learners who want to read rather than drill, and who have been burned by a book that turned out to need a dictionary on every page. The drill is running today. The nightly reader is built, and its sends are paused.
The details
| Price | Free. |
|---|---|
| Where it runs | A browser, at filottolang.com. Nothing to install. |
| The drill | Cloze and reverse cloze, scheduled to meet you near the edge of what you can already do. |
| The reader | A chapter written fresh each night, delivered as a link to your own page. Sends are paused. |
| Sources | Sentences from Tatoeba, glosses from Wiktionary, frequencies from OpenSubtitles, each under its own license and credited on the page. |
Questions
What does the calibration do?
It shows you Italian words and asks which ones you do not know, mixing in invented words like crallarro so that guessing shows up as guessing. It takes a couple of minutes and it is how the first estimate of your vocabulary gets made.
What is the coverage gate?
Before a chapter is sent it is measured against your known-word list, and only words you have actually proved count towards the score. A chapter below the floor does not go out. It gets rewritten until it clears.
Why does the estimate not count?
The statistical estimate of your vocabulary is good enough to place you and not good enough to spend. A word the model thinks you probably know is a word that can stop you mid-paragraph, so proof is the only currency the gate accepts.
Why are the nightly chapters paused?
The reader had no readers. Sending generated chapters into an empty room proves nothing, so the sends are off and the project is parked until October. The drill stayed up because people use it.
Get Filotto
Log: Filotto
3 entries- 22 Aug 2026 note One known-word list per person
- 20 Aug 2026 note Filotto's coverage gate only counts words you have proved
- 20 Aug 2026 experiment Filotto's nightly chapters are sending
Get updates
Occasional notes on what's happening at Enginery. No spam, no marketing.