{"id":1772,"date":"2026-09-22T09:04:11","date_gmt":"2026-09-22T01:04:11","guid":{"rendered":"https:\/\/blog.liu-qi.cn\/2026\/09\/22\/x-daily-2026-09-21\/"},"modified":"2026-09-22T09:04:32","modified_gmt":"2026-09-22T01:04:32","slug":"x-daily-2026-09-21","status":"publish","type":"post","link":"https:\/\/en.blog.liu-qi.cn\/2026\/09\/22\/x-daily-2026-09-21\/","title":{"rendered":"X Platform September 21 AI Brief | JEV Decision Model Opens for Trial, AI App Supply Surges but Paid Adoption Remains Low, 3D Modeling Enters Agent-Ready Stage"},"content":{"rendered":"<h2 id=\"topic-66c640fa93\">JEV Transitions from &#8216;New Model&#8217; to Verifiable Decision Model<\/h2>\n<p>Several bloggers have summarized JEV&#8217;s value as a decision model that &#8220;only outputs judgments, not responsible for writing&#8221;: its appeal lies not just in novelty, but in speed and concurrency that directly translate into visible application results. TypeSafe-related posts show JEV has removed its waitlist, offering all users a $5 credit; Guizang also demonstrated using it to concurrently judge dozens to hundreds of 3D assets, completing scene construction in about one second. The community is also experimenting with integrating it into the Rime input method to sort candidate Chinese characters using context; LangSmith shared arrangements for a free trial of the open-source decision model SemIf. At this stage, it&#8217;s more suitable as a high-speed judgment tool for filtering, sorting, and classification tasks, not as a replacement for models requiring text generation or multimodal understanding.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@op7418: <a href=\"https:\/\/x.com\/op7418\/status\/2101870777465557020\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/op7418\/status\/2101870777465557020<\/a><\/li>\n<li>@op7418: <a href=\"https:\/\/x.com\/op7418\/status\/2101874580432552170\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/op7418\/status\/2101874580432552170<\/a><\/li>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2101952863316607154\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2101952863316607154<\/a><\/li>\n<li>@Gorden_Sun: <a href=\"https:\/\/x.com\/Gorden_Sun\/status\/2101987706444419328\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/Gorden_Sun\/status\/2101987706444419328<\/a><\/li>\n<li>@hwchase17: <a href=\"https:\/\/x.com\/hwchase17\/status\/2102065131202945152\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/hwchase17\/status\/2102065131202945152<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-ecb783688b\">AI Application Supply Soars, But Mass Paid Adoption Remains a Major Bottleneck<\/h2>\n<p>Data compiled by Xiaohu from Consumer Edge shows that in Q1 2023, the proportion of all age groups paying out-of-pocket for AI was less than or close to 1%; by Q1 2026, the overall figure rose to about 3%: approximately 6% for ages 18\u201324 and about 1.4% for ages 65+. Another compilation notes that after February 2025, monthly new iOS apps rose from about 30,000\u201350,000 to about 120,000, and Chrome extensions increased from about 3,000 to over 20,000, but revenue in mature categories like games, lifestyle, and books continues to decline. The signal is closer to &#8220;application supply rapidly expanding, with revenue concentrating among a few winners,&#8221; rather than a synchronized boom across the entire consumer AI market.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2101919524845191242\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2101919524845191242<\/a><\/li>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2101919521925894289\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2101919521925894289<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-6fdbdda05a\">Quick Tunnels Temporarily Exposes Local Services to the Public Internet<\/h2>\n<p>Cloudflare Quick Tunnels, introduced by Xiaohu, provides a low-barrier path for development and debugging: after installing the tool locally, a single command maps a localhost service to a public HTTPS address, requiring no Cloudflare registration, domain purchase\/hosting, or opening inbound ports on the router. Its key mechanism is cloudflared establishing an outbound-only reverse tunnel, changing the connection direction from external active access to the router to local active outbound. It solves temporary sharing and testing needs, not equivalent to a full production deployment or access control solution.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2101877033517379832\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2101877033517379832<\/a><\/li>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2101877036348535206\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2101877036348535206<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-2076a80879\">ZCode Responds to Previous Security Controversy with Open Source and Audit<\/h2>\n<p>Max For AI relayed a Zhipu announcement stating that ZCode has completed rectification and open-sourced its code, with v3.14.0 removing the Repo Wiki and severing the local repository snapshot generation and upload pipeline. The announcement also stated that Zhipu invited the China Academy of Information and Communications Technology (CAICT) and Nsfocus to conduct security audits, with both institutions confirming cloud data has been cleaned and the new client version shows no path to trigger local repository snapshots or file exfiltration; Zhipu simultaneously reaffirmed that code data is not retained, not used for model training, and will establish a regular vulnerability feedback mechanism. The above is visible information from the official announcement as relayed by the blogger; audit conclusions and data commitments should be considered alongside subsequent public materials.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2101847323047542878\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2101847323047542878<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-0f83a6971c\">Hypit Transforms a Once-Verified Video into a Reusable Workflow<\/h2>\n<p>The Hypit demo by AI Dad focuses not on &#8220;one-click generation of a single video,&#8221; but on converting already proven effective sales content into a modifiable, reusable, and mass-producible system. Its design organizes visuals, subtitles, B-roll, and effects around the semantics of the script, rather than locking everything to the timeline; users can use natural language to bring product introductions forward, adjust subtitles, or add shots, and first check shot sequence, readability, composition, and rhythm via Studio mock. The project is open source; the sales and profit figures in the case study come from partner business records, with the blogger explicitly stating they are not publicly audited data or profit guarantees.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@zstmfhy: <a href=\"https:\/\/x.com\/zstmfhy\/status\/2101984265290547387\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/zstmfhy\/status\/2101984265290547387<\/a><\/li>\n<li>@zstmfhy: <a href=\"https:\/\/x.com\/zstmfhy\/status\/2101984272278310928\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/zstmfhy\/status\/2101984272278310928<\/a><\/li>\n<li>@zstmfhy: <a href=\"https:\/\/x.com\/zstmfhy\/status\/2101984274484548093\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/zstmfhy\/status\/2101984274484548093<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-4d5d2a99bb\">Hyper3D Advances 3D Modeling Workflow to Agent-Ready<\/h2>\n<p>Hyper3D&#8217;s officially released Agentic Mode was relayed by two bloggers as a production-oriented 3D workflow: the system can understand inputs like multi-view images, industrial drawings, and e-commerce images, autonomously determine modeling methods and parameters, and generate editable N-GONS models; the posts also list capabilities like resizing and adding animations. What can be confirmed currently is the product&#8217;s public functional description and the demo seen by bloggers; one cannot infer all complex industrial scenarios have reached production-grade effectiveness from this, but the direction has shifted from &#8220;generating a model&#8221; to &#8220;having an Agent responsible for an editable workflow segment.&#8221;<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2102059922544599314\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2102059922544599314<\/a><\/li>\n<li>@vista8: <a href=\"https:\/\/x.com\/vista8\/status\/2102060480600936931\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/vista8\/status\/2102060480600936931<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-1fc6d6ee01\">OpenClaw Sees Concurrent Security Audit and Experimental Interaction Updates<\/h2>\n<p>Project lead Peter Steinberger relayed that OpenClaw underwent a security audit with Trail of Bits participation, with the team stating the audit &#8220;found no critical issues&#8221;; simultaneously, the project also shared development progress on FaceTime support: currently an experimental audio-direction feature supporting GPT-live, enabling user calls to Claw and Claw calls to users, expected in the next version. The former is an external audit signal, the latter is a developer-disclosed upcoming feature; both should be understood as relayed by the project, not as fully available.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@steipete: <a href=\"https:\/\/x.com\/steipete\/status\/2102049706830647467\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/steipete\/status\/2102049706830647467<\/a><\/li>\n<li>@steipete: <a href=\"https:\/\/x.com\/steipete\/status\/2101748928274419843\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/steipete\/status\/2101748928274419843<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-46cfb47e18\">Agent Software Testing Begins Covering the &#8220;Reproduce\u2014Fix\u2014Deploy\u2014Verify&#8221; Loop<\/h2>\n<p>Vercel-related developer Guillermo Rauch shared handling a mobile browser issue: the Agent independently delved into reproducing the problem, verified it in an iPhone simulator, created a temporary Vercel deployment, and completed the fix, deployment, and re-check. He concluded that the Agent&#8217;s advantage in software QA isn&#8217;t just generating code, but its ability to continuously execute high-intensity testing and verification workflows. This case is a single developer&#8217;s description of one practice, illustrating workflow potential, not implying all projects have reached the same reliability.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@rauchg: <a href=\"https:\/\/x.com\/rauchg\/status\/2101846262840799251\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/rauchg\/status\/2101846262840799251<\/a><\/li>\n<\/ul>\n<p>Stats: Scanned timeline items=559 Matched blogger count=58 Matched tweet total=322 Weighted tweet score=238.15 Original tweet count=115 RT tweet count=82 Crawl attempts=4 Boundary coverage status=tail_confidently_crossed_target_boundary<\/p>\n","protected":false},"excerpt":{"rendered":"<p>JEV model trial accelerates decision-making applications, AI tool supply explodes but public willingness to pay remains low, 3D modeling evolves towards editable workflows, highlighting the coexistence of AI practicality and market\u2026<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[19],"class_list":["post-1772","post","type-post","status-publish","format-standard","hentry","category-brief","tag-x--ai-"],"_links":{"self":[{"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/posts\/1772","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/comments?post=1772"}],"version-history":[{"count":1,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/posts\/1772\/revisions"}],"predecessor-version":[{"id":1773,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/posts\/1772\/revisions\/1773"}],"wp:attachment":[{"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/media?parent=1772"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/categories?post=1772"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/tags?post=1772"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}