feat(marketing): surface AgentX across site copy - #756
Merged
cquil11 merged 3 commits intoAug 18, 2026
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
cquil11
marked this pull request as ready for review
August 18, 2026 20:39
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit bac3162. Configure here.
|
Found 2 test failures on Blacksmith runners: Failures
|
![Fix with [code]smith](https://pr-comments-assets.blacksmith.sh/codesmith/fix-with-codesmith-light.png)
Position the shipped AgentX agentic coding workload across core metadata, indexable copy, machine-discovery feeds, and localized Chinese surfaces without adding new promotional UI.\n\nCorrect stale public copy that described AgentX or the benchmark cadence as forthcoming/nightly, and add regression coverage for English and Chinese positioning.\n\n中文：在网站核心文案与元数据中突出 AgentX\n\n在核心元数据、可索引内容、机器发现订阅源及中文页面中准确呈现已上线的 AgentX 智能体编码工作负载，不新增促销式 UI。同步修正将 AgentX 描述为尚未上线、以及将当前测试频率描述为每夜运行的过时文案，并补充中英文回归测试。
Use agentic inference for category-level headlines, SEO titles, and labels. Keep AgentX as the name of the long-context, multi-turn coding scenario, workload, results, and methodology. Remove vague promotional and AI-writing-style phrasing, and point machine discovery to the /agentx route from #755. 中文：在类别层级的标题、SEO 标题与标签中统一使用“智能体推理”，仅在指代长上下文多轮编码场景、工作负载、结果与方法时使用 AgentX。删除空泛宣传及明显的 AI 写作套话，并将机器可读入口指向 #755 提供的 /agentx 路由。
Use the shorter category-correct heading in English and Chinese to preserve the non-scrolling overview layout at desktop, tablet, and phone widths. 中文：中英文统一采用更短且类别表述准确的智能体推理成本标题，确保总览页面在桌面、平板与手机宽度下不会出现横向滚动。
cquil11
force-pushed
the
agent/agentx-semantic-marketing
branch
from
August 18, 2026 21:21
bac3162 to
9050e1c
Compare
cquil11
added a commit
that referenced
this pull request
Aug 18, 2026
* docs(datasets): explain AgentX methodology Document the v1.0 trace pipeline, replay controls, interpretation guidance, and reconstruction limits on /datasets and /zh/datasets. Add E2E coverage for both localized pages. 中文：补充 AgentX v1.0 的 trace 处理流程、回放控制、结果解读与重建边界，并同步更新 /datasets 和 /zh/datasets；新增覆盖中英文页面的 E2E 测试。 * feat(agentx): publish canonical methodology routes Make /agentx and /zh/agentx canonical, preserve the former datasets URLs with permanent redirects, and update navigation, metadata, internal links, analytics, and tests. Add bilingual long-form methodology pages with source-grounded copy, 16 original figures, descriptive captions, and reproducibility links. 中文：发布 AgentX 规范路由与方法论页面。将 /agentx 和 /zh/agentx 设为规范地址，以永久重定向保留旧 datasets 链接，并同步更新导航、元数据、站内链接、分析事件与测试。新增中英文长篇方法论页面，使用 16 张原始图片、准确图注及可复现性链接。 * fix(agentx): correct statistic semantics Place each definition term before its value in the DOM on both AgentX methodology surfaces while retaining the existing value-first visual order. 中文：修正 AgentX 统计信息的语义结构。在两处方法论页面中，让每个定义术语在 DOM 中位于数值之前，同时保持现有的数值优先视觉顺序。 * fix(nav): defer desktop links to xl screens Keep the hamburger navigation active through 1024 CSS pixels so the six-link AgentX header does not overflow. Add component coverage at 1009, 1012, 1020, and 1024 pixels and retain an explicit xl desktop-bounds check. 中文：将桌面导航延后至 xl 屏幕。在 1024 CSS 像素及以下继续使用汉堡菜单，避免包含六个链接的 AgentX 页眉横向溢出；新增 1009、1012、1020 和 1024 像素的组件测试，并保留 xl 桌面边界检查。 * feat(marketing): surface AgentX across site copy (#756) * feat(marketing): surface AgentX across site copy Position the shipped AgentX agentic coding workload across core metadata, indexable copy, machine-discovery feeds, and localized Chinese surfaces without adding new promotional UI.\n\nCorrect stale public copy that described AgentX or the benchmark cadence as forthcoming/nightly, and add regression coverage for English and Chinese positioning.\n\n中文：在网站核心文案与元数据中突出 AgentX\n\n在核心元数据、可索引内容、机器发现订阅源及中文页面中准确呈现已上线的 AgentX 智能体编码工作负载，不新增促销式 UI。同步修正将 AgentX 描述为尚未上线、以及将当前测试频率描述为每夜运行的过时文案，并补充中英文回归测试。 * fix(marketing): distinguish AgentX from agentic inference Use agentic inference for category-level headlines, SEO titles, and labels. Keep AgentX as the name of the long-context, multi-turn coding scenario, workload, results, and methodology. Remove vague promotional and AI-writing-style phrasing, and point machine discovery to the /agentx route from #755. 中文：在类别层级的标题、SEO 标题与标签中统一使用“智能体推理”，仅在指代长上下文多轮编码场景、工作负载、结果与方法时使用 AgentX。删除空泛宣传及明显的 AI 写作套话，并将机器可读入口指向 #755 提供的 /agentx 路由。 * fix(overview): shorten agentic inference heading Use the shorter category-correct heading in English and Chinese to preserve the non-scrolling overview layout at desktop, tablet, and phone widths. 中文：中英文统一采用更短且类别表述准确的智能体推理成本标题，确保总览页面在桌面、平板与手机宽度下不会出现横向滚动。 * docs(glossary): define agentic inference terms (#757) Add a first-class Agentic inference category with AgentX, workload, replay, closed-loop, and subagent definitions. Keep the English and Simplified Chinese glossary surfaces, metadata, cross-links, and content-contract tests aligned.\n\n中文：新增“智能体推理”一级分类，并补充 AgentX、智能体编码工作负载、轨迹回放、闭环基准测试与子智能体定义；同步英文和简体中文术语表、元数据、交叉链接及内容契约测试。 * feat(agentx): highlight new AgentX results Add a localized NEW badge to the AgentX tab on desktop and mobile. Replace the Kimi K3 launch promotion with an agentic-results banner and modal that link directly to Agentic Traces, name the covered models, and use View results actions. Capitalize the AgentX Methodology heading and extend route, nudge, locale, and responsive coverage. 中文：为桌面端和移动端的 AgentX 导航标签添加本地化“NEW”徽标；将 Kimi K3 发布提示替换为智能体结果横幅与弹窗，直接跳转至 Agentic Traces，列出已覆盖模型，并统一使用“查看结果”操作；同时修正 AgentX Methodology 标题大小写，并补充路由、提示、本地化与响应式测试。 * fix(agentx): unify NEW badge sizing Render the AgentX navigation, launch banner, and launch modal badges through one shared 32 by 16 pixel component. Add an end-to-end geometry assertion so the three pills cannot drift apart again. 中文：AgentX 导航、发布横幅和发布弹窗统一使用同一个 32×16 像素徽标组件，并新增端到端尺寸断言，防止三个“NEW”徽标再次出现样式偏差。 * fix(agentx): optically center NEW badge labels Shift the shared badge label by one pixel to compensate for the uppercase font baseline while preserving the common 32 by 16 pixel pill. Extend the browser regression test to verify the label position in all three placements. 中文：将共享徽标文字上移 1 像素，以补偿大写字体的视觉基线，同时保持统一的 32×16 像素胶囊尺寸。扩展浏览器回归测试，验证三个使用位置的文字对齐。 * fix(agentx): center badges and upgrade methodology figures Center every NEW label exactly within its shared pill and remove the trailing letter-spacing imbalance. Replace ten methodology figures with the supplied 4200-pixel originals, add five new full/256k and replay variants, and extend bilingual captions and browser coverage to all 21 figures. 中文：将所有“NEW”文字精确居中到共享徽标中，并移除字距造成的视觉偏移。使用提供的 4200 像素原图替换 10 张方法论图片，新增 5 张 full/256k 与回放变体图片，并为全部 21 张图片补充中英文说明和浏览器测试。 * fix(ui): centre the NEW badge label inside its pill The 32x16 pill carried px-1.5, leaving a 20px content box for a label that inks ~23px at 10px bold. The glyphs overflowed that box and Chrome pushed them to the end, so `NEW` sat 6px from the pill's left edge and 2.7px from the right, with the W nearly touching it. Drop the padding to slack (px-0.5) and centre with flex on the pill itself instead of a nested full-width grid. That nested `w-full` label is also what hid the bug from the e2e check: its box stayed centred at 20px while the text spilled out of it, so assert on the rendered ink. Renders 32x16 with even 4.34px gutters in DM Sans and in every fallback tested (Arial, Verdana, Tahoma, Trebuchet, Impact, Courier New), so the shared badge size holds even when next/font's `display: optional` falls back. * fix(agentx): sharpen methodology figures
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Terminology and copy audit
Generic category copy now uses "agentic inference benchmark" or "agentic inference." AgentX appears only when naming the specific scenario, workload, results, metrics, or methodology.
Changed copy was audited against the indicators in Wikipedia: Signs of AI writing. The follow-up removes vague promotional claims, filler such as "insights," broad unsupported comparisons, repeated summary language, and unnecessary em-dash or parenthetical constructions. Concrete workload mechanics, hardware coverage, metrics, and data-availability qualifiers remain.
Stack
This PR is stacked directly on #755 and targets
agent/agentx-methodology. The parent branch supplies the canonical/agentx,/zh/agentx, and methodology routes plus permanent redirects from the former dataset URLs. This PR only updates semantic copy and machine discovery to use those canonical routes. Merge #755 before #756.Claim boundaries
This is a semantic copy and SEO change only. It does not add a popup, redesign the interface, change benchmark behavior, or expand the detailed AgentX methodology. AgentX availability remains qualified by supported or available data.
Validation
Latest stacked verification:
bun run fmtbun run lintbun run typecheck/agentx,/zh/agentx,/agentx/methodology, and/zh/agentx/methodology: HTTP 200/datasetsand/zh/datasets: permanent redirects to the canonical AgentX pagesllms.txt: canonical AgentX and methodology links verifiedEarlier full PR validation:
bun run test:unit: 3,952 passingbun run buildbun run security