About

BitPage is a pseudonym. This blog runs anonymously — what matters here are the fixes, not the names. Here’s who’s behind it.

I’m the person people call not to “build us a website,” but when “everything’s down, figure it out.”

I taught myself to code long before my first job in IT. Nobody wanted to hire a self-taught developer without experience, so I got in through QA: two years hunting other people’s bugs before I wrote a line of production code myself. It turned out to be the best possible school — I knew from day one exactly how systems break.

Fourteen years later, nine of them as a backend team lead. What has gone through my hands:

• a large backend platform where I’m one of the main authors of the code;
• a server fleet fully described as code;
• PostgreSQL and ClickHouse clusters under real production load;
• Android development;
• commercial projects I built from scratch that are still running today.

You won’t find the exact size of any of those systems here: they don’t belong to me. How that plays out in the articles is below.

Over these years I’ve worked through every layer of real business pain: recovered servers after crashes, tracked down bugs “nobody could find,” moved infrastructure between hosts, put CI/CD where things used to deploy by hand, and optimized databases that had been slow for years.

When the AI era arrived, I did what I always do: instead of rehashing other people’s articles, I built my own AI platform from scratch — architecture, chats, admin, security. My way of understanding a technology is one: build it with my own hands and put real people on top of it.

So the advice here is simple and practical. Behind it there’s no theory — only systems that are running right now.

What this blog is about

• Only what’s been tested in production. Nothing from the “people say this works” category — only what I’ve actually run.
• Numbers instead of adjectives. Not “much faster,” but “latency dropped from X to Y.”
• Primary sources. Links to documentation, RFCs, and changelogs — not to summaries.
• Honest dates. The “Updated” badge appears only when the text has genuinely changed — with a note on what changed.
• Other people’s systems stay anonymous. Versions, service limits, response codes and verbatim error text are exact — those are properties of the tools, and anyone can verify them. But the size of someone else’s infrastructure, its volumes and the exact dates of an incident I give as a model: “take a fleet of ten nodes.” The failure mechanism is unchanged; the client’s system stays closed. Where real measurements are needed, I build a test rig and measure them myself.

Contact

You can reach the author through comments on syndicated versions of these articles on external platforms.