AFTERTIME is the future, available to our users today. We do not wait for a technology to become familiar: we bring it to the point where you can simply use it.
01
The content lives in the phone, not on a server. On a plane, in the mountains and on roaming it works the way it does at home. That costs more to build — and is the honest trade: with no network our products do not turn into a splash screen.
02
Anything that can go out of date carries the date it was checked. Where we are not sure, it says "check this" instead of a number we invented. It costs us pretty promises and pays us back in trust.
03
We do not build a profile of you and we do not sell behaviour. The account exists for exactly one thing: so a purchase follows the person rather than the device.
Behind each of our products is not a stack of ready-made libraries but an envelope of our own: methods we design and test, computation that runs on the user’s device, and data we take from its source. Here is what it consists of.
We build on what is already demonstrated and reproducible: the forgetting curve and spaced repetition in learning, psychoacoustics in sound, atmospheric models in forecasting. We do not reinvent the science — we bring it to an applied form and state plainly where its limits are.
The curriculum, the difficulty scale, the order material is introduced in and the repetition algorithm are designed here, not borrowed. The same goes for verification: every unit of material carries a source, a date it was checked and a confidence level — an internal standard that everything we publish answers to.
Sound synthesis, repetition intervals, forecast processing and search all run on the device itself. That architecture costs more to build and buys three things at once: it works with no network, it has no latency, and personal data has nowhere to leak to.
Where a product depends on outside data we work with its origin directly — national weather services, official registries, native speakers — rather than with resellers. The source is named in the product and the user can check it.
A model here does one narrow job with a checkable result: explain the mistake in a sentence, make sense of a document, suggest a wording. We do not turn a product into a wrapper around a chat, and we do not pass a model’s answer off as an expert opinion.
One source of truth for the site and the app, typing and automated checks on every change, a history behind every edit. None of it is visible directly — it shows in how quickly a correction ships and how rarely something breaks.
Sign-in opens together with payments — we are wiring those up now.
The link arrives by mail and is good for 15 minutes. No password needed.
