{"id":1391,"date":"2026-06-26T09:08:48","date_gmt":"2026-06-26T01:08:48","guid":{"rendered":"https:\/\/blog.liu-qi.cn\/2026\/06\/26\/x-daily-2026-06-25\/"},"modified":"2026-06-26T09:08:48","modified_gmt":"2026-06-26T01:08:48","slug":"x-daily-2026-06-25","status":"publish","type":"post","link":"https:\/\/en.blog.liu-qi.cn\/2026\/06\/26\/x-daily-2026-06-25\/","title":{"rendered":"X Platform June 25 AI Brief | Anthropic Alleges Largest-Ever Distillation Attack by Alibaba, ByteDance Releases Seed-2.1-Pro, Qualcomm Acquires Modular"},"content":{"rendered":"<h2 id=\"topic-55fba137ea\">Anthropic Publicly Accuses Alibaba&#8217;s Qwen of Launching Largest-Ever Distillation Attack<\/h2>\n<p>Anthropic has formally written to the U.S. Senate Banking Committee and the White House, accusing Alibaba&#8217;s Tongyi Qianwen (Qwen) AI Lab of conducting a systematic, industrial-scale distillation attack on Claude. According to the letter, the involved parties used approximately 25,000 fake accounts to conduct over 28.8 million interactions between April 22 and June 5, aiming to distill Claude&#8217;s software engineering and Agent reasoning capabilities. Compared to the total volume (about 16 million interactions) of the three previously named companies\u2014DeepSeek, MiniMax, and Moonshot AI\u2014the volume attributed to Alibaba alone is nearly double. Anthropic states these operations violate both terms of service and geographic restrictions (Claude is not available in China). Alibaba has not yet commented. This incident occurs amidst delicate relations between Anthropic and the Trump administration\u2014following the Commerce Department&#8217;s restriction of access to Fable 5 and Mythos 5 on national security grounds. Anthropic&#8217;s current valuation is $965 billion, and it has confidentially filed for an IPO; the distillation attack directly threatens its commercial moat.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@dotey: <a href=\"https:\/\/x.com\/dotey\/status\/2069979107903918149\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/dotey\/status\/2069979107903918149<\/a><\/li>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069992812997943603\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069992812997943603<\/a><\/li>\n<li>@imwsl90: <a href=\"https:\/\/x.com\/imwsl90\/status\/2070123783479054653\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/imwsl90\/status\/2070123783479054653<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-7ff2ccc50c\">Apple Raises Prices Across Entire Product Line, Covering Mac and iPad Series<\/h2>\n<p>Apple has increased prices across its entire product lineup, with multiple bloggers compiling and verifying the specific hikes. The starting price for the iPad 11 has risen from \u00a52999 to \u00a53799, the iPad mini from \u00a53999 to \u00a54799, the iPad Pro from \u00a58999 to \u00a510799, the MacBook Neo from \u00a54599 to \u00a55499, and the Mac Mini from \u00a54499 to \u00a55999. The U.S. market has seen similar adjustments: the MacBook Neo increased from $599 to $699, and the MacBook Air from $1099 to $1299. Some bloggers report that JD.com channels have also confirmed the price increases, with certain models out of stock. The U.S. price hikes were confirmed by MKBHD, and @joshesye confirmed the actual selling price increase on JD.com. @imwsl90 commented, &#8220;Thanks, Apple, for saving me money.&#8221;<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2070139320674230601\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2070139320674230601<\/a><\/li>\n<li>@imwsl90: <a href=\"https:\/\/x.com\/imwsl90\/status\/2070142384940417292\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/imwsl90\/status\/2070142384940417292<\/a><\/li>\n<li>@joshesye: <a href=\"https:\/\/x.com\/joshesye\/status\/2070149515483508882\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/joshesye\/status\/2070149515483508882<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-7147cea064\">Claude Fable 5 Shows Signs of Return, But Actual Restoration Awaits Confirmation<\/h2>\n<p>Multiple bloggers have observed signs of Fable 5&#8217;s impending return. String changes in Claude Code v2.1.190 added the prompt &#8220;Your Fable 5 usage for this week&#8221; and removed the phrase &#8220;purchased separately from your plan,&#8221; suggesting Fable 5 will be permanently included in subscriptions with a weekly quota. Subsequently, reports confirmed Fable 5 has been re-listed on AWS but requires U.S. identity verification and use case review. However, a message cited by @dotey states &#8220;currently Fable 5 traffic is zero.&#8221; @xiaohu speculates it might return tonight. This is the first positive signal since the U.S. Commerce Department restricted access to the model on national security grounds on June 12.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2070091002887668161\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2070091002887668161<\/a><\/li>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069876683507937436\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069876683507937436<\/a><\/li>\n<li>@Gorden_Sun: <a href=\"https:\/\/x.com\/Gorden_Sun\/status\/2069908293229924721\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/Gorden_Sun\/status\/2069908293229924721<\/a><\/li>\n<li>@xiaohu: <a href=\"https:\/\/x.com\/xiaohu\/status\/2070104158536073602\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/xiaohu\/status\/2070104158536073602<\/a><\/li>\n<li>@dotey: <a href=\"https:\/\/x.com\/dotey\/status\/2070153590342123836\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/dotey\/status\/2070153590342123836<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-a9cdf56fcf\">ByteDance Releases Doubao-Seed-2.1-Pro, Multiple Developers Test Coding Capabilities<\/h2>\n<p>ByteDance Seed released Doubao-Seed-2.1-Pro on June 23, with multiple bloggers conducting immediate tests and giving positive reviews. The model achieved a 59.1% win rate against Claude Opus 4.6 in a developer crowdsourced blind test, supporting full repository code understanding and multi-file collaborative editing. @imwsl90 tested it in the Trae Work IDE and found React project performance superior to Opus 4.7 but execution speed 10-15% slower; native HTML optimization still lags behind Opus. @cellinlab integrated it with Codex for game development, PPT creation, and bug fixing, achieving usable results. @dongxi_nlp used it to organize personal study notes and generate a knowledge base page. @vista8 praised its &#8220;god-like fish recognition.&#8221;<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@imwsl90: <a href=\"https:\/\/x.com\/imwsl90\/status\/2070044202348781941\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/imwsl90\/status\/2070044202348781941<\/a><\/li>\n<li>@imwsl90: <a href=\"https:\/\/x.com\/imwsl90\/status\/2070044208120140097\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/imwsl90\/status\/2070044208120140097<\/a><\/li>\n<li>@imwsl90: <a href=\"https:\/\/x.com\/imwsl90\/status\/2070044213430165849\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/imwsl90\/status\/2070044213430165849<\/a><\/li>\n<li>@cellinlab: <a href=\"https:\/\/x.com\/cellinlab\/status\/2069939701172379956\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/cellinlab\/status\/2069939701172379956<\/a><\/li>\n<li>@dongxi_nlp: <a href=\"https:\/\/x.com\/dongxi_nlp\/status\/2070000935007383654\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/dongxi_nlp\/status\/2070000935007383654<\/a><\/li>\n<li>@vista8: <a href=\"https:\/\/x.com\/vista8\/status\/2070046399077466574\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/vista8\/status\/2070046399077466574<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-08bd662fee\">OpenAI Releases Custom Chip Jalape\u00f1o and GPT-5.5 Instant Update<\/h2>\n<p>OpenAI officially announced the design and mass production of its first custom AI chip, Jalape\u00f1o, in collaboration with Broadcom, specifically designed for LLM workloads. A Google TPU\/XLA engineer analysis suggests this chip is nearly the first to be fully developed by Codex\/GPT\u2014not only was the software stack written by coding models, but much of the hardware design work may also have been done by AI. On the same day, OpenAI also released a new version of GPT-5.5 Instant, improving intent understanding and complex constraint handling, rolling out gradually to paying users. @MaxForAI quipped, &#8220;The world wants GPT-5.6, not 5.5 Instant.&#8221;<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069868543613796378\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069868543613796378<\/a><\/li>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069863220375077067\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069863220375077067<\/a><\/li>\n<li>@OpenAI: <a href=\"https:\/\/x.com\/OpenAI\/status\/2069843083701915755\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/OpenAI\/status\/2069843083701915755<\/a><\/li>\n<li>@thsottiaux: <a href=\"https:\/\/x.com\/thsottiaux\/status\/2069837606574289329\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/thsottiaux\/status\/2069837606574289329<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-c58a705dfc\">Qualcomm Acquires AI Computing Firm Modular for Approximately $3.92 Billion in All-Stock Deal<\/h2>\n<p>Qualcomm announced the acquisition of AI computing infrastructure startup Modular in an all-stock transaction valued at approximately $3.92 billion based on Qualcomm&#8217;s closing price. Modular was co-founded by LLVM\/Clang\/Swift creator Chris Lattner. Its products include the inference framework MAX and the high-performance programming language Mojo, with the core goal of solving the problem of repeated adaptation of AI models across different chips, providing a cross-hardware software abstraction layer. Analysis suggests Qualcomm&#8217;s move aims to strengthen its developer platform in the AI chip software ecosystem competition, targeting NVIDIA&#8217;s CUDA moat.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069876193961361413\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069876193961361413<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-e459142490\">Samsung Electronics Officially Ends Home Appliance Retail in China, Official WeChat Account Deregistered<\/h2>\n<p>Following Samsung (China) Investment Co., Ltd.&#8217;s complete cessation of home appliance retail business in mainland China on May 6, the official Samsung Electronics WeChat public account completed self-initiated deregistration on June 23, with all historical posts taken offline. This account previously handled functions such as new product launches for Samsung appliances, promotional campaigns, and user education. Counting from its formal entry into the Chinese home appliance market in 1992, Samsung&#8217;s home appliance retail journey in mainland China has concluded after 34 years, covering over ten product lines including TVs, refrigerators, washing machines, air conditioners, monitors, and a full range of home appliances.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2070044009943478473\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2070044009943478473<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-344defcb9f\">Google Gemini Loses More Core Members, Two Join Anthropic<\/h2>\n<p>According to Bloomberg, two key researchers from Google DeepMind, Jonas Adler and Alexander Pritzel, will leave to join Anthropic. Adler is a key leader in Google&#8217;s AI programming projects, having participated in AlphaFold and Gemini 1.5; Pritzel is a veteran who joined DeepMind in 2014, involved in Deep RL, AlphaFold 2\/3, and the Gemini training system. Their respective focuses on coding and pre-training align precisely with the areas Anthropic currently seeks to strengthen.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@MaxForAI: <a href=\"https:\/\/x.com\/MaxForAI\/status\/2069867315068559461\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/MaxForAI\/status\/2069867315068559461<\/a><\/li>\n<\/ul>\n<h2 id=\"topic-fabf05d24a\">Tencent Launches Agent-Specific Email, Feishu CLI + Codex Workflow Emerges<\/h2>\n<p>Tencent has launched an email service specifically for Agents. After registration, users receive a prompt for configuring CLI use by Agents like Codex. @vista8 demonstrated daily applications of Feishu CLI + Codex: creating multi-dimensional tables for accounting through conversation, viewable and editable directly on mobile. Additionally, Google introduced the Open Knowledge Format (OKF) standard, using Markdown + YAML frontmatter to build knowledge packages consumable by Agents. Multiple bloggers are also experimenting with packaging various apps and services into Agent Skills.<\/p>\n<p>Sources:<\/p>\n<ul>\n<li>@vista8: <a href=\"https:\/\/x.com\/vista8\/status\/2070144005589831745\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/vista8\/status\/2070144005589831745<\/a><\/li>\n<li>@vista8: <a href=\"https:\/\/x.com\/vista8\/status\/2069991791357476887\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/vista8\/status\/2069991791357476887<\/a><\/li>\n<li>@vista8: <a href=\"https:\/\/x.com\/vista8\/status\/2069983469715755404\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/vista8\/status\/2069983469715755404<\/a><\/li>\n<li>@Khazix0918: <a href=\"https:\/\/x.com\/Khazix0918\/status\/2070047038188957757\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/x.com\/Khazix0918\/status\/2070047038188957757<\/a><\/li>\n<\/ul>\n<p>Statistics: Scan timeline count=360, Matched blogger count=35, Matched tweet total=188, Weighted tweet score=144.5, Original tweet count=82, RT tweet count=44, Crawl attempt count=2, Boundary coverage status=tail_confidently_crossed_target_boundary<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Major controversies and breakthroughs in the AI field, involving allegations of model attacks, new model releases, and chip ecosystem strategies.<\/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-1391","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\/1391","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=1391"}],"version-history":[{"count":0,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/posts\/1391\/revisions"}],"wp:attachment":[{"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/media?parent=1391"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/categories?post=1391"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/en.blog.liu-qi.cn\/index.php\/wp-json\/wp\/v2\/tags?post=1391"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}