Skip to content
RU

Research: Runet Analytics in Numbers

We publish quarterly benchmarks based on millions of scans through Enterno.io tools. All data is collected from real user checks, aggregated and anonymised.

HTTP & APIs 17

AI coding assistants in 2026: what was measured and what was invented

Figures "from the Stack Overflow 2026 survey" circulate widely, but the results of that survey do not exist: the questionnaire opened on 23 June 2026 and the page returns 404. Ever…

Read →

What sets the cost of an embedded model: 61,977 production calls

Between May and August 2026 we made 61,977 calls to language models on live user requests. Of those 48.6% were served from cache and never reached a provider — a share that rose fr…

Read →

Website and API response times in 2026: 1.4 million measurements

Between March and August 2026 our system ran 1.92 million checks. Across the successful ones the median response time is 95 ms and the p95 is 966 ms — a tenfold spread.Which is exa…

Read →

Bots and AI crawlers in 2026: how many, and who gets disallowed in robots.txt

Are bots 35.5% of traffic, or 61.9%? Cloudflare gave both answers for the same week of August 2026. The difference is not the source but whether you count all requests or only page…

Read →

Container runtimes in 2026: what was measured, what was surveyed, and what is missing

Figures for "containerd’s share in 2026" circulate, but no measurement stands behind them: the last telemetry is November 2023 and the series has been discontinued. The CNCF survey…

Read →

CSS frameworks in 2026: Bootstrap, Tailwind and the large remainder

Surveys of CSS framework share usually argue about the ratio of Bootstrap to Tailwind. Our measurement of 399 home pages shows the dominant quantity is different: about two thirds …

Read →

Database driver popularity in 2026: a measurement across two registries

We queried driver download counters directly from npm and PyPI. PostgreSQL is first in both ecosystems — 64.6 million a week in JavaScript and 530.4 million a month in Python, ahea…

Read →

Docker image sizes in 2026: a measurement, and why no trend exists

We queried official image sizes through the Docker Hub API on 27 August 2026. python:3.13 weighs 393.7 MiB; its alpine variant, 16.1. A twenty-fourfold difference.Growth of the sam…

Read →

On-device model inference in 2026: the platform is ready, the use is not there

WebAssembly is supported by 95.96% of browsers and WebGPU by 83.99%. The platform for on-device computation is ready and available to nearly everyone.Meanwhile GPU work is submitte…

Read →

Headless CMS in 2026: why the share cannot be measured from outside

We parsed 401 home pages and found signs of a headless content management system on eleven. Against roughly a third for traditional systems.It is tempting to conclude that "headles…

Read →

JavaScript framework popularity in 2026: four sources, four different answers

React runs on 6.1% of websites, was named by 85% of surveyed developers, and pulled 670 million npm downloads in a month. None of these numbers converts into another.Below: what ea…

Read →

Language model provider shares in 2026: why there is no single answer

By website visits the leader is OpenAI at around 53%. By AI crawler traffic it is Google at 24.8%. Among developers, Anthropic at 39%. By client downloads in Python it is OpenAI ag…

Read →

How many dependencies an npm package and project has: four samples and live measurements

Between 49 and 56% of npm packages have no dependencies at all, and the registry median is zero. Four independent samples say so, including our own control. The mean has not moved …

Read →

What people actually install to build RAG in 2026: a two-registry measurement

No survey exists on "the share of teams using RAG pattern X" — but download counters do. We queried them directly, and the first thing visible is that the two most popular packages…

Read →

Russian AI services in 2026: three months of operation in figures

Between May and August 2026 we made 30,449 calls to YandexGPT on live user requests. The median response is 3,793 ms, with 0.14% empty responses.Mean latency rose from 2.2 to 4.0 s…

Read →

Web server shares in 2026: why nginx is 14% and 44% at the same time

Within a single Netcraft report for July 2026, nginx’s share is 14.45%, 20.37% and 43.92%. All three are correct — they answer different questions. W3Techs, the same month, says 31…

Read →

CMS shares in 2026: WordPress, Tilda and Bitrix worldwide and in Russia

"WordPress holds 58.9% of the market" and "WordPress runs on 40.7% of sites" are two numbers from one source on one date, and both are true. The difference is the denominator — and…

Read →

Security 5

DNS & Domains 4

SSL / TLS 4

Other 17

How bot traffic changed by 2026: one honest trend and five irreconcilable measurements

Between December 2025 and August 2026 the human share of HTML page requests fell from 47% to 38.1%. This is one of the few comparisons in the field built on the same metric from on…

Read →

CDN on Russian websites in 2026: rarer, and not Cloudflare

We requested headers from 180 domains and identified CDNs from those headers and from IP address ownership. In .ru a CDN was found on 24.6% of the domains that answered, in .com on…

Read →

HTTP response code distribution in 2026: why every such statistic is wrong

We hold 39,660 checks, and 98.93% of them carry response code 200. We could publish that as a headline figure — and it would be untrue about the web while the data is entirely true…

Read →

The npm supply chain in 2026: what was measured, and what is spam

In December 2025 npm revoked every long-lived token — the measure the whole effort after that year’s attacks was built around. It worked. And two of the three worm campaigns of 202…

Read →

Cross-platform passkeys in 2026: what changed with the specification

On 25 August 2026 WebAuthn Level 3 reached W3C Recommendation status — the first such step in more than five years. It brings flags by which a site can distinguish a credential tha…

Read →

WebGPU in 2026: browser support against real-world use

WebGPU is fully supported by browsers accounting for 83.99% of global page views. Meanwhile 0.004% of Chrome page loads actually submit work to a graphics processor.The gap between…

Read →

Delivery network latency in 2026: measured from a Russian server

We measured the path from a Moscow server to eighteen public addresses and broke time to first byte into its parts. Russian networks answer in 40–72 ms, foreign ones in 65–133.So t…

Read →

IPv6 in Russia in 2026: why four measurements give four different answers

For Russia, Google shows 59.94% IPv6 penetration, APNIC 3.10%, Cloudflare 7.2%. A twentyfold gap between two measurements of the same quantity is nobody’s mistake — it follows from…

Read →

Cookies on home pages in 2026: how many are set immediately, and with which attributes

We requested 600 home pages once, without a single click, and looked at which cookies that response already sets. In the .ru zone 59% of sites do so — against 31% for .org.This is …

Read →

Core Web Vitals in Russia in 2026: country-level CrUX data and its limits

By CrUX field data for July 2026, Russia is above the world on desktop (69.26% against 60.04%) and below it on mobile (51.63% against 53.21%). A year ago it led on mobile — the gap…

Read →

Mail authentication in 2026: SPF and DMARC across zones

On SPF prevalence the .ru zone trails by nothing at all: 85% of mail-accepting domains publish the record — exactly as in .com.But the strict mode that actually rejects a forgery i…

Read →

Mail deliverability in the Russian web in 2026: who runs the mail, and why it matters

In the .com zone two companies run the mail for 37% of domains. In .ru the largest player takes 20%, and the market is split across five services.But something else matters more: 2…

Read →

HTTP/3 across CDNs in 2026: who actually turns it on

Of the 104 hosts in our sample advertising HTTP/3 support, 62 are served by Cloudflare. Every other CDN and hand-configured server combined accounts for 42.The difference is decisi…

Read →

How often a site needs checking: 140 monitors and 489 incidents measured

We looked at 140 active monitors and 489 closed incidents across five months. An unwelcome coincidence emerged: 40% of monitors poll once an hour, while one outage in four lasts fi…

Read →

Open ports in 2026: 1,042 checks run by users themselves

We did not scan third-party hosts on our own initiative — instead we counted 1,042 checks that users ran themselves, choosing the host and the port set.The result is calmer than th…

Read →

Redirect chains in 2026: data from 806 checks

Across 806 checked addresses: 48% answer directly and another 35.4% after a single hop (http → https or www). The mean is 0.74. 2.9% take three or more, and one address in the samp…

Read →

www versus bare domain and canonical URLs in 2026: a 600-host measurement

The choice between www and the bare domain does not affect ranking — what matters is that a choice was made and is held to. We looked at how it is made in practice: in the .ru zone…

Read →

Want to suggest a topic? Open an issue on GitHub or email support@enterno.io.