Essentyx didn't start as a product. It started as years of paying close attention to language in places where getting a single word wrong could cost a project.
Where the language affinity came from
For a long time, our work lived in the space where business and IT have to actually understand each other. Specifications, requirements, process descriptions: documents where the precise meaning of a single word decides whether a system gets built correctly or has to be torn down and rewritten six months later.
Years of that work teach you to read sentences for what they encode rather than how they sound. You stop trusting confident phrasing and start checking what is actually being said. That is where the affinity for language came from.
Once you see it, you can't unsee it
That same lens does not switch off when the laptop closes. We started noticing the same patterns everywhere: in news articles, social posts, headlines, even in everyday conversation. Certainty inflation, loaded framing, sentences that smuggle a position past you while looking like a neutral statement of fact. Once those patterns are visible, they are visible everywhere. You cannot unsee them.
For a while we assumed this was a niche problem, limited to a small subset of low-quality publishers. Reading more carefully made it clear the opposite was true. The patterns are everywhere. They are the default of how content gets written today, not the exception. That was the moment the idea was born.
The first prototype
That was about two years ago. The first version of Essentyx took a structural approach: pull the whole page, apply rules to locate the real article inside the navigation, ads and chrome around it, then operate on what was left. It worked well enough to prove the idea, but it was not yet a product.
Rebuilding it properly
Once we committed to putting this in the world, almost every early decision got revisited. The extraction logic, the model that does the rewriting, the way the extension behaves on the page: large parts were redesigned from the ground up. What survived from the prototype was the conviction that the problem was real. Almost everything else was rebuilt into something we would want a stranger relying on every day.
Where we are now
Watching it run on a real article is still satisfying. Not just the obvious cases, but the small ones, where a single sentence was doing more work than the meaning required and the engine quietly trimmed it back to what was actually said. The fact that it picks up even those subtle cases, sentences expressing more than just pure information, is what makes the cumulative effect of reading with it on feel different.
We are looking forward to seeing how the world adopts it and how much impact it has on the people who read and on the media that produces what they read.