Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Bytedance

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 284

  1. ByteDance reportedly plans up to $6 billion share buyback

    ByteDance plans a new share buyback of up to $6 billion at $241.35 per share, corresponding to an approximately $400 billion valuation, according to reports citing The Information and informed sources. The buyback price is about 20% above the level of the previous round last November. The company has also proposed extending to December 11, 2030, the deadline by which external shareholders may request that it repurchase their entire holdings. ByteDance said it plans to provide 300 million shares linked to the value of its AI business to employees working in AI roles.

    36氪快讯 · 🔥 0
  2. Cookiy Labs raises $7M for AI-mediated human research

    TMTPost reports that Dong Youchao left his previous role in 2025 to found Cookiy Labs, which announced a $7 million pre-seed financing. The company aims to help businesses and AI systems find experienced people for interviews, surveys, and data exchanges, then return the results to AI workflows; its scale, speed, and other capabilities are vendor claims.

    钛媒体 · 🔥 0
  3. HIT-ByteDance Team Introduces Structured Multimodal Reasoning Format DRT

    A joint Harbin Institute of Technology and ByteDance team introduced Dense Reasoning Trace (DRT), a structured format that separates visual observations from logical deductions and links intermediate states with symbolic operators. According to Juejin AI, the team’s DRT traces achieved 5.5-fold token efficiency over natural-language chain-of-thought and scored 1.3 percentage points above a Qwen3-VL baseline on mathematics and chart-reasoning evaluations. The report notes that the format was manually designed, that its generalization to open-domain tasks remains unverified, and that performance on natural images is unknown.

    掘金 · 人工智能 · 🔥 0
  4. WeChat tests a top-level Short Drama entry for Green Bubble

    微信 is gray-testing a “Short Drama” entry in its Discover page alongside “Look,” according to 钛媒体, which cited 表外表里. The report interprets the placement as a first-level distribution channel for Green Bubble, while arguing that limited original content and weak social sharing make it unlikely to match WeChat Channels or Hongguo.

    钛媒体 · 🔥 0
  5. Galaxy Zhixue Details Xiaolu Aixue’s AI Education Architecture

    量子位 reported that Galaxy Zhixue chief scientist Deng Shujun presented the technical architecture and school deployment model of Xiaolu Aixue at the ninth China Education Technology Innovation and Development and Application Results Conference in Wuxi on November 11. The company says the product combines Doubao large language models with a proprietary learning engine for learning assessment, task recommendations, Socratic questioning, student planning and teacher support. Galaxy Zhixue also says Xiaolu Aixue has served or partnered with more than 500 schools across 24 provincial-level administrative regions. The report was supplied and authorized by Galaxy Zhixue, and it does not provide independent confirmation of the product’s technical performance, deployment scale or learning outcomes.

    量子位(原生 RSS) · 🔥 0
  6. AI Splits Internet Economics Into Attention and Token Markets

    36Kr, republishing commentary from Internet Jianghu, argues that AI is dividing internet business models into attention-based content platforms and token-based AI tools. It cites QuestMobile figures reported for September showing Hongguo at roughly 430 million deduplicated users versus Doubao’s 419 million, alongside a reported sale of about 2 trillion tokens during a 30-minute Double 11 livestream. The commentary further argues that AI-generated short dramas can lower production costs while token spending creates value only when converted into delivered outputs.

    36氪 人工智能 · 🔥 0
  7. Seedance 2.5 flash launches with faster generation and lower pricing

    ByteDance Volcano Engine launched Seedance 2.5 flash, a video generation model positioned between the Seedance 2.0 and Seedance 2.5 flagship tiers. The company says it generates twice as fast as Seedance 2.5, costs about RMB 0.99 per second for 720P text-to-video and image-to-video, supports clips up to 30 seconds, and accepts up to 10 images, five videos and five audio files per request. Zhidx’s LibTV test found strong instruction following, detailed fast motion and good character and scene continuity, but occasional garbled text in fast-paced shots.

    IT之家 AI · 🔥 0
  8. ByteDance reportedly plans a Meta Muse-like personal AI assistant

    Bloomberg, as cited by GeekPark, reports that ByteDance plans to launch a personal AI assistant similar to Meta Muse within weeks. ByteDance reportedly presented the plan to its AI researchers, engineers, and support personnel, with ambitions to strengthen its position in AI video generation and narrow its gap with leading US AI laboratories. The report also says co-founder Zhang Yiming believes ByteDance can compete with Anthropic and OpenAI at the AI frontier.

    极客公园 · 🔥 0
  9. US Data Firms Reportedly Sell AI Training Environments to Chinese Labs

    A ChinaTalk investigation reported by The Decoder identifies at least nine US data firms selling roughly $500 million in AI training data to Chinese AI labs. Some of the data consists of reinforcement learning environments for tasks such as coding, financial analysis and office work. Buyers reportedly purchase environments previously sold to OpenAI or Anthropic, allowing them to avoid some in-house development costs and wasted training runs. The investigation’s authors called for export controls on dual-use environments, particularly in cybersecurity and biology, while saying Chinese labs still trail US peers by at least several months structurally.

    The Decoder · 🔥 0
  10. ByteDance Reportedly Plans Muse-Like AI Assistant

    Bloomberg reports, as cited by Engadget and other outlets, that ByteDance plans to launch a Muse-like AI assistant in the coming weeks. The assistant's name and workings have not been detailed, ByteDance did not immediately respond to a request for comment, and it remains unclear whether companion hardware is planned. A separate iFanr report, citing Caixin and Bloomberg, says ByteDance started the internal Spell project in April to explore personal AI assistants. After Meta introduced Muse, ByteDance reportedly accelerated the project and planned to combine Spell with Doubao's conversational team to integrate related functions into Doubao.

    Engadget AI · 🔥 0
  11. ByteDance Plans Muse-Like AI Assistant Release

    ByteDance plans to release a Muse-like AI assistant in the coming weeks to augment its position in AI video. The company hopes the milestone will help narrow the gap with leading US AI laboratories in as soon as three years.

    Bloomberg Technology · 🔥 0
  12. AI Short-Drama Agents Expand Into Full Production Workflows

    A Jiemian News report says AI short-drama tools are moving beyond single-purpose generation toward workflows covering scripts, character and scene management, storyboards, video production, editing, distribution and performance evaluation. ByteDance's Xiaoyunque AI, iQIYI's Naduo Pro, SenseTime's Seko and Jimeng AI's short-drama studio are among the products entering the market. China Netcasting Services Association figures show that 239,000 micro-dramas were released domestically in the second quarter of 2026, including 153,000 AI dramas. The market is competing on workflow coverage and price, while subscription-based tools and closer collaboration in content production have emerged as two commercialization paths. The report does not establish which model will prove more stable.

    虎嗅 AI · 🔥 0
  13. TRAE Takes Center Stage in Luo Yonghao’s Double 11 Livestream

    量子位 and 36氪人工智能 report that Luo Yonghao demonstrated ByteDance’s TRAE AI development platform during a Double 11 livestream. He used TRAE to build a mouse-wheel voice-input tool, create a short-drama market report, and turn more than 50,000 words into two slide decks; the demonstrations also exposed occasional character-recognition and sending failures. The livestream promoted a 9.9-yuan first-month Lite membership, but the reports did not provide independently verified sales figures or benchmark results.

    量子位(原生 RSS) · 🔥 0
  14. AI Short-Drama Platforms Compete for Content and Pricing Power

    An analysis by 万点研究, also published by 钛媒体 with authorization, compares four AI short-drama strategies: ByteDance’s production-to-distribution loop, Tencent’s WeChat-based viewing-time incentives, Kuaishou’s monetization of Kling AI, and Baidu’s AIGC infrastructure. It argues that platforms are competing not only for scripts and traffic but for control of settlement coefficients that shape content supply, distribution, and revenue sharing. The analysis cites QuestMobile figures showing that Red果 Free Short Drama exceeded 430 million monthly active users in September 2026. Several company metrics and strategic interpretations are cited from QuestMobile or company disclosures and were not independently verified in the supplied reports.

    虎嗅 AI · 🔥 0
  15. Anthropic’s Reported IPO Valuation Reset Pressures AI Model Economics

    36氪 reports that Anthropic is planning a Nasdaq IPO, with investor valuation expectations falling from $2 trillion to $1.5 trillion and then to discussion near $1 trillion as technology stocks adjusted, infrastructure financing costs rose, and investors reassessed its financial performance. The report also analyzes how Anthropic’s revenue growth, GAAP operating loss, adjusted profitability, and long-term compute commitments may affect valuation expectations for Chinese model developers such as DeepSeek, Moonshot AI, Zhipu AI, and MiniMax.

    36氪 人工智能 · 🔥 0
  16. Six major Chinese AI firms' revenue is about one-tenth of OpenAI and Anthropic's

    TechNews cited an Epoch AI estimate made in September that six major Chinese AI firms generated about $11 billion in combined annual revenue last year, roughly one-tenth of OpenAI and Anthropic's combined revenue. The estimate includes cloud and related AI revenue from companies such as DeepSeek, Zhipu AI, Alibaba and ByteDance, rather than only standalone AI product sales; ByteDance led the group with an estimated $4 billion. The report contrasted this with 2026 OpenRouter usage statistics showing DeepSeek with a 23.3% token share, above the 19.2% combined share of OpenAI and Anthropic. It also reported that the two U.S. companies accounted for 86% of global enterprise LLM API spending and attributed part of the Chinese firms' revenue constraints to price competition in the open-source ecosystem.

    科技新报 · 🔥 0
  17. Google releases A2UI 1.0 for native, interactive agent interfaces

    Google released A2UI 1.0, the first stable version of its open generative UI protocol, allowing agents to describe forms, charts and confirmation cards that applications render with their own components. The release adds bidirectional function calls, broader accessibility and metadata support, web and Flutter renderers, and agent SDKs for Python, TypeScript and Dart. Google says Atlassian, Alation, Bloomberg and ByteDance are already building with the protocol, while Gemini Enterprise support is still being upgraded and Jetpack Compose support for version 1.0 remains forthcoming.

    Google Developers Blog · 🔥 0
  18. Ulanqab Emerges as an AI Computing Hub

    Recode China AI, translating and editing a first-person Daily People article, describes Ulanqab’s transformation from an agricultural city into a major AI data-center hub. The report cites Kuaishou’s first self-built facility, DeepSeek recruitment and reported expansion plans, and 28.2 billion yuan in project commitments secured in 2024. It also describes growing demand for data-center workers alongside rising qualification and age requirements.

    Recode China AI · 🔥 0
  19. China's Intelligent Knowledge targets expert data for AI post-training

    PINGWEST reports that 智能知识 was founded in April 2025, raising seed funding from FiveYuan Capital and MiraclePlus and angel funding jointly from Jingqiu Fund and Yaotu Capital. The startup focuses on post-training data, organizing expert knowledge around reinforcement-learning environments, tasks and validators, and argues that human experience remains necessary despite advances in RSI. Its leadership includes former ByteDance Seed model-data engineering head Wang Sheyang and CTO Li Kuanye, with other team members coming from companies including Scale AI.

    品玩 实时要闻 · 🔥 0
  20. Chen Lin Leaves ByteDance for AI Education Venture

    A report said Chen Lin publicly acknowledged leaving ByteDance to start a business focused on AI education. No further details about the venture were provided in the available report.

    DoNews · 🔥 0
  21. Former Toutiao CEO Chen Lin Confirms AI Education Venture

    Former Toutiao CEO Chen Lin publicly confirmed in a personal article that he has left ByteDance and is preparing an AI-education startup. His public profile describes him as the founder of Xiaolu Aixue, also known as Galaxy Zhixue, and as a former CEO of Toutiao and Dali Education. Chen Lin said concerns about how AI technologies could affect young people's growth inspired him to return to education entrepreneurship.

    IT之家 AI · 🔥 0
  22. BLA Targets AI Reverse-Engineering Agents With Decoys and Quotas

    量子位 reports that BLA is an open-source MCP service and CLI tool designed to frustrate AI-assisted reverse engineering of protected software. Its author says BLA combines poisoned files, quota exhaustion, and misleading paths; internal tests estimated that protected reverse-engineering tasks consumed roughly 300 to 2,800 times more tokens. Across 13 protected test cases, seven errored or were rejected, two produced partial results, and three showed no observed analysis blockage.

    量子位(原生 RSS) · 🔥 0
  23. Doubao Work faces workflow and adoption hurdles in enterprise push

    A New Berry Daybreak investigation, authorized for republication by 36Kr AI and Cyzone Tech, examines ByteDance’s enterprise push for Doubao Work, including its Feishu integration, seat-plus-usage pricing, sales reorganization and announced opening to third-party models including Kimi 3.1 Pro, GLM5.3 and DeepSeek V4.1 Flash. User examples indicate that Doubao Work can save time on approvals, analysis and reporting, but may omit conditional steps or use outdated information. WorkBuddy, Qwen Office and Baidu Dazi are pursuing enterprise markets through different products and industries. ByteDance has not disclosed Doubao Work’s sales or contract figures.

    36氪 人工智能 · 🔥 0
  24. Commentary Highlights Younger Leaders in China’s AI Research

    A commentary from the WeChat account 人人都是产品经理, credited to 怪哥, says researchers born in the 1990s and 2000s are leading important AI work at major Chinese technology companies. It cites Yao Shunyu at Tencent’s Hunyuan, Sun Tianxiang at Baidu, Luo Fuli at Xiaomi’s MiMo, Zeng Yan at ByteDance’s Seedance 2.0, Zeng Aohan at Zhipu AI’s GLM pretraining and infrastructure, and Zhao Pengyu at MiniMax. The commentary argues that scarce AI talent and rapid model development are weakening traditional seniority-based promotion rules. The personnel details and interpretation are not independently corroborated in the supplied material.

    虎嗅 AI · 🔥 0
  25. ByteDance launches Seedream 5.5 for professional image editing

    ByteDance released Seedream 5.5, an image-generation model focused on precise multi-round editing, native RGBA transparency, stronger small-text rendering and more natural photographic detail. The model is available through Doubao, Doubao Work and Jimeng, while Volcano Ark API access was reported as planned or rolling out; IT之家 cited an approximately RMB 0.28 price per 2K-square image edit. QuantumBit’s 100-round test found Seedream 5.5 retained more fidelity than GPT Image 2.6 in that task, although some distortion remained in later rounds.

    36氪 人工智能 · 🔥 0
  26. Huawei’s pharmaceutical unit targets AI, not drug development

    TMTPost reports that Huawei established a pharmaceutical unit in December 2025, with president Fan Jie positioning the team as an AI enabler for the pharmaceutical industry rather than a developer of medicines or drug pipelines. The article also examines how OpenAI, Anthropic, ByteDance, NVIDIA, Meta and other technology companies have expanded into AI-assisted drug discovery, while noting that AI has yet to demonstrate clear advantages in Phase II and III clinical trials.

    钛媒体 · 🔥 0
  27. Possible Coze shutdown fuels workflow-versus-agent debate

    A commentary says that on November 2, a ByteDance Coze developer posted that Coze might be taken offline; the article cites no official confirmation. It interprets the developer's description of drag-and-drop nodes as a low-code transitional form and contrasts predetermined workflows with autonomous agents that plan each step. The author argues that workflows may remain the better choice when reliability, accuracy and reproducibility require consistent execution, while autonomous agents suit tasks whose paths cannot be fixed in advance.

    掘金 · 人工智能 · 🔥 0
  28. ByteDance Weighs Second Phase for Brazil Data Center Campus

    Bloomberg reports that ByteDance is discussing a second phase for its Pecém data center campus in Ceará, Brazil, and seeking a developer for the project. Sources say the proposed phase could be larger than the first phase, while a separate facility serving Brazilian customers remains under preliminary discussion. The existing first phase is being developed by Omnia with about 200 MW of IT capacity, about 300 MW of total power demand, and 20 planned data halls.

    Bloomberg Technology · 🔥 0
  29. TRAE Accounts Will Automatically Upgrade to Doubao-Feishu Accounts

    TRAE announced that existing accounts will automatically upgrade to Doubao-Feishu accounts on November 18, 2026, according to IT之家 and DoNews. TRAE said code, projects, history, membership status, points, and other usage data would be preserved, while users with a different or missing Doubao-Feishu account could end up with a separate account depending on their phone number. The announcement follows the October 9 merger of TraeWork and TraeCode into the unified TRAE development platform.

    IT之家 AI · 🔥 0
  30. Report: AI-Made Theatrical Feature Draws Weak Reception

    A 36Kr commentary says a film described as China’s first AI-made theatrical feature received a public-screening permit and opened in cinemas last month, but its reception was reportedly weak, with forecasts of box office below 1 million yuan and social-media criticism of its quality. The report says its material faced resolution, dropped-frame and black-screen problems in theaters, while AI-film production workflows remained unstable. It also cites Chinese Online’s disclosed data that AI short dramas accounted for 55% of domestic AI token consumption, describing them as a more established commercial use of the technology.

    36氪 人工智能 · 🔥 0

← PreviousNext →

Experience and discussion from the community

Share my Bytedance experienceAsk about Bytedance

RSI: Evolution Goes Beyond Models

AI has already entered code writing, algorithm optimization, and model development. But after a task is completed, what remains may be reusable capability—or merely an answer and scattered logs. The difference becomes apparent only when a similar problem comes up again: can the system reuse what it learned, or must it pay the cost of exploration all over again? Recursive Self-Improvement (RSI) co…

RSISwarmEvoMapLoopEvolution

evomap.aihttps://evomap.ai