X says it has rebuilt its web app from the ground up and is now shipping the faster version to users, part of an effort to reverse years of sluggish performance on desktop and mobile browsers. The company began rolling out the update “one country at a time” this week, starting with signed-in users, according to a blog post from principal engineer Yagiz Nizipli.

The headline claim is striking. In a before-and-after comparison run over a deliberately slow connection, Nizipli said the old app left visitors staring at the X logo for more than 22 seconds before the first post rendered. On the new app, he wrote, that first post appears in 2.7 seconds.

“Open a profile on X.com over a slow connection. On the old web app, you'd stare at the X logo for more than 22 seconds before the first post appeared. On the new one, it shows up in 2.7.”
— Yagiz Nizipli, principal engineer, X

What actually changed

Beyond the load-time numbers, the most consequential architectural change is consolidation. According to the blog post, the new web app “replaces the three servers we used to run for people, search crawlers, and link previews with one.” That detail matters more than it first appears.

Running separate rendering stacks for human visitors, search engine crawlers, and the bots that generate link previews is a common pattern at large consumer platforms, but it is expensive and hard to keep in sync. Each stack needs its own deployment pipeline, caching layer, and debugging story, and small differences between them produce the odd inconsistencies users notice when a shared link looks different from the page it points to.

Collapsing those into a single server means one code path to maintain, one set of caches to warm, and presumably a smaller infrastructure bill — a priority that has been visible across X since the company dramatically reduced its headcount following Elon Musk’s 2022 acquisition.

Independent scrutiny is thin so far

The performance figures released so far are the company’s own, drawn from a throttled-connection test rather than a broad field measurement. That is standard practice for launch posts, but it leaves open questions about real-world results: how the app performs on mid-range Android handsets, on congested mobile networks, or for logged-out visitors who currently see a heavily restricted version of X.

The rollout itself is deliberately gradual. Deploying “one country at a time” for signed-in users lets engineers watch error rates and latency before a global switch, but it also means the experience will be inconsistent for weeks. Security researcher Jane Manchun Wong, known for reverse-engineering unreleased features in consumer apps, posted about the update as it began appearing — a signal that the change is already visible to technically inclined users watching X’s client bundles.

A platform rebuilding in public

The web app rewrite fits a broader pattern. Since the rebrand from Twitter to X, the company has been re-architecting core systems with a much smaller engineering organization while simultaneously pushing an “everything app” agenda spanning payments, creator monetization, and long-form content. Performance work is one of the few areas where the results are measurable and the claims are falsifiable, which may explain the willingness to publish specific numbers.

Historically, Twitter’s web client was a frequent target of criticism for heavy JavaScript bundles and slow time-to-first-content, particularly on older devices in emerging markets where bandwidth is expensive. Those markets are also where X has bet on growth, making speed a commercial issue as much as an engineering one: users on slow connections are the most likely to abandon a page before it paints.

Why this matters beyond X

Web performance has become a competitive battleground across the industry, with search engines rewarding fast pages through ranking signals and app-makers treating load time as a retention metric. For a company that derives revenue from advertising and subscriptions, a 20-second wait is not merely an annoyance — it is inventory that never loads, impressions that never count, and users who never sign in.

  • Load time: X claims a drop from 22 seconds to 2.7 seconds on a slow connection for the first post on a profile.
  • Architecture: three server stacks — user traffic, search crawlers, link previews — merged into one.
  • Rollout: country by country, signed-in users first.
  • Verification: figures are self-reported and not yet independently benchmarked.

What to watch

The next markers will be whether the new client reaches logged-out visitors and crawlers without regressions, whether third-party measurement firms confirm the gains on real networks, and whether X publishes field data rather than lab numbers. Until then, the company’s claim is best read as a credible engineering direction paired with marketing-grade benchmarks — a familiar combination in platform announcements.

For users, the practical test is simple: load a profile on a bad connection and count. If the first post arrives in under three seconds, X will have delivered on the rare promise that is easy for anyone to check.