Skip to content
Manifesto

The deck writes
itself.

This page is the thesis behind Glotta: what it takes for granted, where it would be defensible, what is holding it back, and the three things that could disprove it entirely. It is written to be measured, not argued about.

“Learning a language is a problem of time, not of willpower.”

The line the whole thing started from

The idea

The moment you learn is a moment you already have.

Study apps ask for twenty minutes a day that nobody has, and teach sentences somebody else chose. Meanwhile, that same person spends two hours a day writing messages: real sentences, theirs, in real situations.

Glotta sits right there. It is a keyboard: you use it instead of the system one, you type normally, and the translation appears above the keys. If you need it, one tap swaps it in and you send that. You have not opened an app, you have not studied anything.

And every sentence you write becomes something to learn from — not because you saved it, but because you used it.

Glotta does not compete on content. It competes on friction, which is zero.

The language-app market is large, crowded, and has an enormous drop-off rate precisely because it asks for dedicated time. It does not need better content: it needs to ask for nothing.
The potential

The defensible position is the data.

After a month, Glotta knows which sentences that person actually uses, in which situations, and which ones they still get wrong. No course can know that. It is the kind of advantage that grows on its own.
Week 1It knows you write a lot in the evening, and you miss possessives.
Month 1It knows which twenty sentences you actually need, and in what order.
Month 3It knows which mistakes come back despite the reviews.
Month 6It holds a deck no course could rebuild for you.

The natural directions, nearest first

01

More languages per person

Today, one pair only. People living abroad use three, and use them on the same day.

02

Two-way conversation

Translating what you write and what you receive. Half the conversation does not go through Glotta today.

03

Register and tone

Formal or informal before you send. In languages with a polite form it is the difference between the right message and an embarrassing one.

04

A personal glossary

Names, nicknames, work terms no model can guess — and that turn up in every real conversation.

05

Professional pairs

People working in English as a second language write emails all day. It is the same product, with an audience that pays more readily.

The brake

Not technical, but commercial

The cloud engine costs money on every request. The subscription has to cover it, and the shared key currently compiled into the app has to move behind a service of our own before anything is sold.

What you need to know before believing it

Three things could disprove all of it.

They need measuring rather than arguing about. If the answers are yes, the rest is execution.
01

Do people install a keyboard?

It is the highest hurdle in this category of product. Asking someone to replace their keyboard is asking a lot, and no amount of design makes it a small request.

02

Do they come back to review?

Passive exposure is comfortable, but only active recall produces memory — and active recall requires somebody to open the app of their own accord.

03

Are the collected sentences really the right ones?

The central assumption is that frequency of use identifies what is worth learning. It is plausible, and it is exactly that: an assumption.

Project status

What is already real, and what is not

The app exists, it runs, and it is on TestFlight. What follows is the honest summary, including the two columns that do not look good.

Works, today

  • Translation in the keyboard, on-device and with optional cloud models
  • Correction within one language and one-tap replacement
  • Collecting and deduplicating sentences, with a usage counter
  • Spaced-repetition review on your own sentences
  • Grammatical analysis with conjugations and a contextual chat
  • Simulated conversation, widget, and a paid tier with five features
?

Not proven yet

  • That people come back. The review loop is new and nobody has used it for a week straight.
  • It is the only question that really matters, and it does not have an answer yet.

Not there

  • Sync between devices. Everything lives on a single phone.
  • For an app whose value is accumulation, that is a serious gap to close before asking anyone for months of trust.

Want the mechanism rather than the thesis?

The “How it works” page opens the five steps, the three surfaces and the engine underneath.

How it works →