Go + HTMX vs React at enterprise scale: what the primary sources actually document

What did Cloud Codes actually claim?#

On 11 August 2026, the YouTube channel Cloud Codes published "Go + HTMX replaces React: In Enterprise Scalability Performance Tests", arguing that the Go-plus-HTMX stack now has enterprise-grade scalability evidence behind it. The video ranges over five underlying sources: two essays hosted on htmx.org, the TechEmpower Framework Benchmarks repository, Go's release history, and a Drupal core issue.

Some of what the video says is documented in those sources. Much of it is not, and the difference matters — so this article separates the two.

What does "Does Hypermedia Scale?" actually say?#

The load-bearing primary source is Carson Gross's essay "Does Hypermedia Scale?". Gross examines whether Hypermedia-Driven Applications (HDAs) scale along five distinct dimensions: system nodes, application performance, feature count, feature complexity, and team size. His conclusion is that HDAs scale well in most of these scenarios, while acknowledging specific challenges with complex client-side features.

That is a more careful claim than "HTMX beats React." The essay concedes the exact territory — deeply interactive, state-heavy client-side features — where a JavaScript framework like React is strongest. Two cautions apply. First, the essay is published on the htmx project's own site; it is an argument by an interested party, not an independent evaluation. Second, the video's framing of "wide" versus "deep" applications — wide apps with many simple screens suit hypermedia, deep apps with heavy client-side state suit React — is Cloud Codes' vocabulary, not the essay's, though it points the same direction as the essay's own concession.

Where does "67% less code" come from?#

The number traces to a single case study: a write-up on htmx.org of a real-world React-to-htmx port at Contexte, based on work by David Guillot. Cloud Codes itself calls the "67% less code" line a "famous talk meme" originating from that port.

Two things follow. First, this is one application ported by one team — a data point, not a benchmark, and it says nothing directly about Go. Second, the text of the write-up available to us does not include the specific percentage, so the 67% figure should be treated as reported rather than documented.

Which of the video's headline claims are not documented?#

Four of the video's most concrete claims could not be confirmed against the sources supplied with it:

• The TechEmpower benchmarks are dead. The TechEmpower Framework Benchmarks repository exists and dates to March 2013. Cloud Codes reports the project was archived in 2026 and that a "64-core HTTP Arena" benchmark has taken its place. The repository metadata available to us does not confirm an archival, and no primary source for "HTTP Arena" was supplied at all.
• Go 1.26 ships a "Green Tea" SIMD garbage collector. Cloud Codes describes a vector-accelerated garbage collector in Go 1.26. Go publishes its release history publicly, but the page content available to us does not confirm the claim.
• Drupal 11.3 adopted HTMX into core. Cloud Codes reports Drupal 11.3 integrated HTMX as a jQuery replacement, "slashing browser JS by 71%". The underlying Drupal core issue could not be retrieved for this article, and the 71% figure appears in no other supplied source.
• A 740:1 NPM-versus-CDN download gap. Cloud Codes presents NPM downloads outnumbering CDN hits 740:1 for React against 4:1 for HTMX as evidence of HTMX's real browser footprint. The methodology behind both ratios is unstated, and the numbers are entirely uncorroborated.

What are the limits of this evidence?#

Every claim in this story traces to a single reporter. No transcript was available for the video, so video claims are attributed to chapter timestamps rather than verbatim quotes. The central primary source is an essay hosted on the project it argues for, and the three strongest numbers in circulation — 67% less code, 71% less browser JS, 740:1 — each rest on exactly one account.

What would change the picture: the Drupal core issue resolving and confirming (or denying) HTMX's integration into Drupal 11.3; an archival notice appearing on the TechEmpower repository; Go 1.26's release notes documenting the garbage-collector change; or any second, independent measurement of the download discrepancy. Until one of those lands, the honest summary is narrower than the video's title: hypermedia-driven applications have a coherent, self-aware scaling argument, one encouraging case study, and a growing stack of claims that still need primary documentation.

Sources#

Primary sources#

• </> htmx ~ A Real World React -> htmx Port
• TechEmpower/FrameworkBenchmarks — 2013-03-22; TechEmpower/FrameworkBenchmarks
• Release History - The Go Programming Language
• </> htmx ~ Does Hypermedia Scale?

Reporting that surfaced this#

• Cloud Codes — Go + HTMX replaces React: In Enterprise Scalability Performance Tests (2026-08-11)