docs/hrx: fix stray h1 headings (SEO); site.py: one h1 per page

[bong-water-water-bong]

The daily SEO check in #traffic-report flagged https://1bit.gg/hrx.html: it had 4 <h1> elements. Wrapped lines in docs/hrx.md that start with #36, #39, #40 render as headings (python-markdown needs no space after #).

• docs/hrx.md: those lines now start with "PR site: the 1bit.MONSTER template, a cleaned-up blog, three new posts #36" etc.
• tools/site.py: the build fails when a page has other than one <h1>, so this can't reach the site again. Verified: the unfixed doc fails the build (hrx.html has 4 <h1> elements), the fixed tree builds 41 pages, all with one h1.

🤖 Generated with Claude Code

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 36 - Partially compliant Compliant requirements: The site takes on the 1bit.MONSTER site's template The blog is cleaned up with new posts on the move to 1bit engine The template includes light modern-minimal tokens and their dark mode The template includes animated relay mark, top bar with light/auto/dark switch The template includes centered hero with the latest post, site cards, blog log, article column, and three-part footer Docs get a sidebar layout in the same tokens Pages include home, overview.html, docs.html, each doc, blog.html, each post, feed.xml, 404.html New posts include "1bit.MONSTER is now 1bit engine", "Prefill on HRX, decode on Vulkan, one KV cache", "The first week of 1bit engine" Blog cleanup includes listing 15 old posts worth keeping in blog/archive.tsv copyright.py covers *.js and *.svg and exempts images pages.yml rebuilds on blog/ changes Non-compliant requirements: None Requires further human verification: None 39 - Partially compliant Compliant requirements: Moves third_party/zinc from c50b4add3402 to upstream main b9123c649f81 Upstream commits include various performance improvements and fixes for ROCm and Vulkan Non-compliant requirements: CI has no GPU; requires manual verification on Strix Halo Requires further human verification: Manual verification on Strix Halo required for performance claims 40 - Partially compliant Compliant requirements: third_party/laya: 1e28ac20c089 -> 23a17522aa49 (NandhaKishorM/laya main) config/laya.json revision: aa8c91ca088e -> 55cf4c4ebb4e (convaiinnovations/laya) Non-compliant requirements: Requires scripts/fetch-laya.sh verification and Laya tests Requires further human verification: scripts/fetch-laya.sh verification and Laya tests required

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review SEO Compliance Check The new SEO compliance check in tools/site.py ensures that each HTML page contains exactly one <h1> element, which is important for SEO. This prevents accidental multiple <h1> tags that could be introduced by markdown rendering of lines starting with #<number>. The check is implemented using a regular expression to count <h1> tags and exits with an error message if the count is not exactly one. # one h1 per page (SEO): a line starting "#36 ..." in a doc renders as a stray h1 h1 = len(re.findall(r"<h1[\s>]", page, re.I)) if h1 != 1: sys.exit(f"{name}.html has {h1} <h1> elements; start no doc line with '#<number>'") Markdown Heading Fix The lines in docs/hrx.md that previously started with #36, #39, and #40 have been updated to start with "PR #36", "PR #39", and "PR #40" respectively. This prevents these lines from being rendered as <h1> elements in the HTML output, which would violate the SEO requirement of having only one <h1> per page. - PR #36 skips CONTs of fresh contiguous results in `src/models/zaya.cpp` (views keep theirs: on HRX their consumers take slower strided paths), bitwise the same output: +2.4%. - PR #39 runs the CCA convolution (Q/K concat, conv-state update, depthwise conv, grouped 2-tap conv and biases, ~35 graph nodes) as one dispatch (`zaya.cca_conv.decode_f32`): 72 -> 81 tok/s. - PR #40 runs the CCA query/key mixing and per-head norms (~30 nodes) as one dispatch (`zaya.cca_qk_norm.decode_f32`): 81 -> 90 tok/s. PRs #39 and #40 accumulate in their own order, so output is not bitwise the same; teacher-forced top-1 against the FP32 model stays 69/96 and batch-1 perplexity moves 33.31 -> 33.49 -> 33.45 (CPU backend

[context7]

Docs7 for 1bit-monster/engine

Result	Status	Action
Deployment	➖ Not used	—
Content review	⚠️ Did not finish.	View run

Commit 7e76fd7