feat: the 1.3 announcement, fresh benchmark numbers, and mobile fixes

[devol-bot]

Everything the 1.3 release needs on the site, plus three mobile faults found
while reading the post on a phone.

Do not merge until 1.3.0 is on npm. The post says the release is out and the
benchmark numbers come from code that is on main in the framework repo but not
yet published.

The announcement post

content/blog/3.announcing-1.3.md. Sections run in the order the release
matters: the cheaper request path, the queue module, devtools rebuilt as a
telemetry consumer, then the telemetry all three sit on. Introspection and the
logger hooks come after it, because both open by referring to telemetry and read
as footnotes to it.

The performance section reports 1.2.1 against 1.3 from a dedicated run,
alternating the two versions on one machine, four rounds per route:

route	1.2.1	1.3
GET /	101,729	102,004	unchanged
GET /id/:id?name=	96,949	100,095	+3.1%
POST /json	89,598	91,817	+2.5%

The plain route not moving is the honest tell rather than a weak spot: it takes
no arguments and its path is static, so neither of the changes behind those
numbers can reach it. The section also carries both costs, the bundle growing
from 267.6 KB to 294.5 KB and the route matcher compiling on first use.

Blocked on a video. public/video/devtools/Queues.mp4 does not exist yet
and the Queues paragraph links it. There is a TODO comment directly above the
tag. The video element will render empty until that file lands.

The benchmark section

Fresh numbers from results/results.md in vercube/benchmarks, a full run of the
suite on one machine. Vercube leads five of the six cells: all three workloads
against the JS routers, and plain text and JSON body against the frameworks that
also ship a container.

Two changes to the shape. Rows are split into the two questions a reader
actually has - how it compares to a bare router, and how it compares to what
they would otherwise pick - because one list mixing both answered neither. And
the "Fastest" badge is computed per group rather than pinned to Vercube, which
now matters: Rikta leads the container group on the query route.

Still open, and worth a decision before this ships. Three node targets are
faster than Vercube and are not on the chart: uws (169,005 on plain text),
hyper-express (124,616) and ultimate-express (163,799). All three are
uWebSockets, so a native binding rather than a JS framework, which is a
principled reason to keep them apart - but the page does not say so, it simply
omits them, while linking the results file where anyone can find them in a
minute. Either the caption should name the exclusion or the chart should include
them in a group of their own.

Mobile fixes

Three separate faults, all below the md breakpoint.

• No gutter on the article body. max-w-3xl mx-auto is wider than a phone,
  so mx-auto had nothing to centre and the text ran edge to edge. px-4 now
  matches the !p-4 of the title block, so the two left edges line up.
• The header collided with the logo. The nav is absolutely positioned with
  its lower edge around 62px and the post container started at mt-8, so the
  date line rendered under the logo mark. Real top padding at every breakpoint
  now, not only at xl.
• Body text sat on the bright halftone. The backdrop fades to transparent on
  the right, which is fine while the text column is on the left, but on a phone
  the column spans the full width. A scrim below md backs it.

Plus one that was not mobile-specific: inline code could not wrap. Nuxt UI
gives it inline-block, which makes a chip unbreakable, so @vercube/telemetry
pushed itself onto a line of its own. Inline with overflow-wrap now. This one
applies to the docs as well, which had the same problem, so that is worth a look
on a dense page.

The hero

Vertical rhythm only. With min-h-[78vh] plus padding the code window started
about a hundred pixels below the fold, so the first screen was a title card and
then nothing. At 60vh its top edge lands inside the first screen and the fold
crops it. Also a denser scrim under the headline, because the display face is
built from dots and so is the field behind it; five staggered entrances
collapsed into two beats; and a focus ring on the install command, which had
none.

Layout, type scale and copy are untouched. An earlier pass moved the code beside
the headline and rewrote the slogan; both were reverted.

Checks

pnpm test:lint clean. pnpm test:tscheck reports the same two pre-existing
OgImageComponents errors as main does, in app/pages/blog/[...slug].vue and
app/pages/blog/index.vue, neither touched by these commits. Verified against a
dev server: all three posts, the blog index, the home page and a docs page
return 200. Not verified visually at phone width - I have no browser here, so
the mobile fixes were made by reading the layout.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}