Skip to content
LangBytes.

LangBytes · How LangBytes works

From one phrase you met to a lesson and a review schedule: how the LangBytes loop works

· 4 min read

The short version

You meet a phrase in real life, capture it, confirm what you need to get done, and LangBytes builds a short lesson around that outcome. It then schedules review, and after you use the language for real you log how it went. Each step is small on purpose, and nothing is generated until you have confirmed the outcome.

1. Capture

An encounter is anything you saw, heard, read, wanted to say, failed to understand or expect to face again. You can type or paste it, or share it from a photo or from your phone. LangBytes cleans up the text and notes the language and a likely situation. It does not teach anything yet.

2. Confirm the situation and outcome

LangBytes proposes whether the encounter is useful, which family of situations it belongs to (dining, transit, work) and what you may want to achieve. These are suggestions, not facts. You confirm or correct the situation and the outcome, for example "order a bowl of ramen and pay". An encounter with no confirmed outcome stays an encounter and produces no lesson.

3. The lesson

The lesson is aimed at your confirmed outcome and has four parts:

  • Your moves: what you say.
  • Their moves: what the other person is likely to say.
  • Critical details: numbers, names and conditions to catch or say exactly.
  • Repair: what to say when the exchange breaks down, and when to use each line.

New language is limited. Each new item must ride on exactly one of your moves, and the lesson says which move and why it was picked, so you meet a few new things in context instead of a long list. A daily cap bounds how many enter one lesson.

4. Practice and review

Practice comes from the lesson, in five kinds: recognize their move, say yours, choose the right repair, handle a different phrasing, and run your moves in order with no hints. Your ratings feed a spaced-repetition scheduler based on FSRS, which sets each card's next review for when your predicted recall falls to the retention you choose. These ratings measure review only. They do not claim you can handle the situation.

5. Use it and log how it went

After a real conversation, you log what happened: it worked, it did not, or you only tried it. You note which move held or failed and add a short debrief. LangBytes shows the move to work on next, with a repair line from the same lesson. "Likely ready to try" is derived from your review history, not typed in, and a single lesson in a single sitting never makes a situation count as ready.

What this does not claim

  • Reviewing cards is not fluency. A review card says you have met a way to say something, not that you can say it in the situation.
  • Product activity is not progress. Counts of lessons and reviews help you spot gaps. Progress is what happens when you use the language for real.
  • Efficacy is not proven. LangBytes has not published learner-outcome results, and the method is still being tested.

Sources