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. Building a One-Minute AI Short with Seedance 2.5 Flash

    A first-person developer workflow shows how a father used DeepSeek, Doubao image generation, DeepSeek Harness, and Seedance 2.5 Flash to turn his child’s drawings into a one-minute superhero short. The author split the story into two clips, used reference images and a light-based transition to connect them, then generated and combined the videos through Python. The author reports occasional prop-consistency errors and unclear native speech, giving the result a subjective score of 8 out of 10.

    掘金 · 人工智能 · 🔥 5
  2. Bytedance and universities release Axion 1.0 AI agent

    Bytedance and researchers from several US universities introduced Axion 1.0, an AI agent trained through three phases. The researchers claim its benchmark performance rivals leading commercial models from OpenAI and Anthropic and attribute the gains to carefully curated training data rather than additional training steps. The model, data, and code are freely available, according to The Decoder.

    The Decoder · 🔥 9
  3. Three Image Models Tested Across Five Styles and Three Edits

    TMTPost compared MiniMax H3-Image, ByteDance Seedream 5.5, and Alibaba Qwen-Image-3.0-Pro using one scene across five visual styles, followed by three sequential edits. H3-Image and Seedream 5.5 completed the edits while preserving much of the original scene; Qwen-Image-3.0-Pro lost the scene during the first two edits and the test was stopped. H3-Image produced 4K output throughout, while reported generation times were generally fastest for Doubao, followed by Qwen and H3-Image in some tests.

    钛媒体 · 🔥 7
  4. AI Video Prices Fall as Model Economics Shift Toward Platforms

    钛媒体, citing NextFin News, reports that Google has cut the list price of a mainstream 720p video-generation tier to well under a dollar for several seconds, while some Chinese vendors promote rates in fractions of a yuan per second. It also says public leaderboard gaps have narrowed enough that ordinary-shot capabilities are often considered interchangeable, making speed, price and workflow fit increasingly important. The analysis examines how Google, ByteDance and Kuaishou can absorb video-generation costs through broader cloud, advertising, content and commerce businesses, while Sora reportedly closed after roughly half a year amid heavy compute costs.

    钛媒体 · 🔥 4
  5. China’s AI short dramas face rights and theater tests

    China’s AI short-drama market is expanding rapidly, with trackers cited by NextFin News counting more than 26,000 new titles across Douyin-group surfaces in October. Moderators on ByteDance’s Hongguo app said they had handled nearly 44,000 works tied to suspected portrait-rights issues by the end of October, blocking almost all before publication. A side chapter of an AI fantasy serial was also scheduled for appointment-based screenings in hundreds of Chinese cities, although the report cautions that platform heat reflects algorithmic favor rather than independent audience demand.

    钛媒体 · 🔥 4
  6. ByteDance Denies Zhang Nan and Xie Xin Departure Report

    A ByteDance representative said information in a report was inaccurate. The report claimed that former Douyin CEO Zhang Nan and former Feishu CEO Xie Xin would leave the company, with Zhiying taking over Jianying and Zhu Jun taking over Jimeng AI. The response did not specify which parts of the report were disputed, and no further details were provided about the alleged departures or succession plan.

    IT之家 AI · 🔥 7
  7. Magnific Says It Passed $300M ARR and Expands in Asia

    Magnific said its annual recurring revenue surpassed $300 million, up from $230 million when it rebranded from Freepik in April. The AI creative platform is opening offices in Shanghai and Singapore and plans to spend $10 million in Asia-Pacific through early 2027. CEO Joaquín Cuenca said the company remains bootstrapped but could consider outside investment next year, with no current plans.

    The Next Web · 🔥 3
  8. AI Video Prices Fall as Model Gaps Narrow

    AIX Finance, republished by 36Kr and TMTPost, reports that AI video generation is moving toward lower per-second pricing and increasingly interchangeable model performance. Google made Gemini Omni 1.2 Flash generally available with 15-second generation and multi-turn conversational editing, while pricing for 720p output reportedly fell to $0.07 per second; Shengshu Technology also introduced Vidu Q4 Preview at a promotional rate of RMB 0.09 per second. The report examines whether large platform ecosystems can absorb continued inference costs even as OpenAI reportedly shut down Sora and no video-model company has publicly disclosed profitability.

    36氪 人工智能 · 🔥 4
  9. AI Video Moves From Mobile Screens Toward Cinema and TV

    Seedance 2.5 4K opened its API, while a 30-second promotional film generated with the model screened at more than 20 high-end cinemas in Beijing, Shanghai, and other cities. AI animated series "Wanyao Zhuan Lu Zhuan" is bringing its special episode "Yunzhou Yike" to cinemas through reservation screenings, while ByteDance has listed a role to build Red Fruit's OTT product for televisions. The report presents these as separate tests of professional production, theatrical IP extension, and household distribution.

    36氪 人工智能 · 🔥 0
  10. AI-generated microdramas surge in China, commentary says

    An IC Laboratory commentary published by 36Kr says 1,055 AI-generated microdramas surpassed 100 million views in the first three quarters of 2026, while nearly half of newly released microdramas were AI-generated. It reports that AI tools can cover much of the production workflow, from story and script generation to character and scene creation and video generation. The commentary links the growth of AI short dramas to the emergence of Seedance 2.0 and describes lower production costs and rapid output as characteristics of the AI wave. It also says the broader short-drama market has shifted toward free, platform-centered distribution, a change that began with ByteDance’s Red Fruit model in 2023, before the AI wave. The figures and assessments are presented by the commentary and were not independently corroborated in the supplied material.

    36氪 人工智能 · 🔥 0
  11. Tencent warns AI competition is a survival issue

    At Tencent’s 28th-anniversary executive meeting, Ma Huateng said rapidly advancing AI threatened even established core businesses and called the company’s response a survival issue rather than an ordinary adjustment. TMTPost interprets the warning alongside Tencent’s higher AI spending, proposed organizational changes, and ByteDance’s expansion across content, commerce, workplace software, and AI assistants. The analysis also examines Tencent’s Xiaowei assistant in WeChat and ByteDance’s effort to bypass established app ecosystems with agentic automation.

    钛媒体 · 🔥 0
  12. China’s AI Office Leaders Consolidate Products and Resources

    An article from the Silicon Research Institute’s WeChat account, authorized for publication by 36Kr, reports that ByteDance, Tencent, Alibaba, Baidu and Kingsoft Office recently adjusted their AI-office organizations. ByteDance integrated Doubao, Feishu and Volcano Engine; Tencent moved QClaw-related businesses and teams into WorkBuddy; Alibaba combined three products into Qwen Office and assigned it to the DingTalk team; Baidu consolidated internal products into Baidu Dazi; and Kingsoft Office shifted more of its focus to AI. The article argues that as widely available models become interchangeable, data usability, enterprise-process customization and established customer and commercial channels are more important competitive factors.

    36氪 人工智能 · 🔥 0
  13. Cola2 applies video-style diffusion modeling to language

    ByteDance Seed researchers led by Yan Zeng published Cola2, which uses continuous latent representations and diffusion-inspired modeling to generate text more efficiently. The paper reports an average score of 60.10 across eight tasks using about 346 billion tokens, while the authors say the model had not converged and required more data and compute. TMTPost interprets the work as supporting a large-scale model direction at ByteDance, but the reported “10-trillion-parameter” roadmap and related personnel claims lack official confirmation.

    钛媒体 · 🔥 0
  14. ByteDance Nears $500M Microsoft Azure Deal Using OpenAI Models

    ByteDance is in advanced talks to add roughly $500 million to its Microsoft cloud-services contract, mainly to access OpenAI models through Azure, according to The Information. The potential transaction has not been completed, and ByteDance had not responded to reports. OpenAI has raised concerns that Chinese firms could use model outputs for distillation, while Microsoft says it found no evidence that ByteDance was improperly distilling or copying model inference.

    The Information · 🔥 0
  15. Tencent Music says AI-generated tracks still draw negligible streams

    Tencent Music Entertainment management said on its Q3 earnings call that streaming plays of AI-generated content remained negligible, while quality and differentiated content continued to attract more attention. TMTPost cited Deezer data showing that AI songs had risen from 10% of daily new releases in January 2025 to nearly 50%, yet received only 1%-3% of plays. The reports also describe Tencent Music's copyright cooperation and content-management approach to unauthorized AI-generated music.

    钛媒体 · 🔥 9
  16. AI Toy Market Expands as Capital Interest Cools and Products Diverge

    TMTPost, citing Ministry of Industry and Information Technology data, reported that China’s AI toy market reached 29 billion yuan in 2025 and was projected to grow to 34.2 billion yuan in 2026. JD.com said its JoyInside hardware-and-software solution supported more than 40 brands, while the jointly developed Smart Hanhan from Huawei and Robo Intelligence sold 6,500 units in 15 minutes at launch. The report also described cooling investor interest, high return rates, product homogenization, limited repurchase demand and a shift toward products tailored to specific use cases.

    钛媒体 · 🔥 0
  17. World Models Draw Attention in AI Interactive Entertainment

    A report from STAR Market Daily says securities analysts have increased their focus on the potential use of world models in interactive film and games. It cites PixVerse R2, Runway's GWM Worlds 2 research preview and Google DeepMind's Genie 4 as industry developments. However, practitioners say current AI interactive works generally use AI video generation for predefined story branches and are largely distinct from true world models. Long-term scene consistency, complex interaction stability and real-time generation costs remain unresolved. Reports of ByteDance developing a world model have circulated, but the company has not formally released a related product.

    虎嗅 AI · 🔥 0
  18. 36Kr AI surveys 32 world models for physical AI

    36Kr AI compiled 32 world models and research systems for physical AI, covering robotics, autonomous driving, simulation, and policy training. It describes world models as learned internal simulators that predict environmental changes from observations and proposed actions, with systems from Nvidia, Baidu, Unitree, Google DeepMind, Meta, Waymo, Wayve, Xiaomi, and others. The report also notes limits involving deformation, disappearing objects, collision and sliding inaccuracies, accumulated prediction errors, generation speed, and computational cost.

    36氪 人工智能 · 🔥 0
  19. AI Interactive Short Drams Could Add Competition to Games

    A TMTPost commentary links a widely circulated internal speech attributed to Tencent's Pony Ma to the author’s warning that AI interactive short dramas could challenge the games industry. It describes branching narratives such as 《全球冰封:我打造了末日安全屋》, where audience choices alter the story, and reports initiatives involving Douyin, Tomato Novel, Kuaishou, Baidu, and Tencent. The commentary also says lower-cost AI production has revived interactive content and cites figures on costs and output. It does not provide the original speech, independent industry measurements, or validation of user participation and monetization effects.

    钛媒体 · 🔥 0
  20. Huling reportedly prepares AI voice-recording smart ring

    IT之家, citing Sina Technology and unnamed sources, reports that former Meituan executive Liu Yanxiang and former Feishu internationalization lead Garin founded Huling, which has completed a seed round in the 10-million-yuan range. Huling's first product is reportedly a voice health ring that records via a physical-button double-click and supports transcription, to-do extraction, and schedule recognition. The ring is reportedly planned for internal testing later in November, preorders at month-end, and shipments in early December.

    IT之家 AI · 🔥 0
  21. Doubao Work launches 21 role-specific AI workbenches

    ByteDance’s Doubao Work launched its first 21 professional workbenches, covering areas such as finance, medicine, education, creative work, writing, legal affairs, HR, data analysis, and administration. According to Zhidx, each workbench preconfigures relevant Skills, tools, plugins, task entries, and professional resources; users can also combine Skills and plugins to create and share custom workbenches. The finance research workbench can connect to Wind, Tonghuashun, and Guotai Haitong after users authorize their accounts, while Zhidx notes that accuracy and professional task quality still require further validation.

    智东西 · 🔥 0
  22. Seedance 2.5 4K Video Generation API Opens to Developers

    Volcano Engine announced that the Seedance 2.5 4K video-generation API is now available to developers and enterprise customers. The model supports a Draft mode for testing 480P footage before producing a 4K result, along with 10-bit color depth. Volcano Engine says specialized training improved the rendering of facial expressions, subtle movements and visual details.

    IT之家 AI · 🔥 0
  23. Seedream 5.5 vs GPT Image 2.6: 15-Prompt Hands-On Test

    A Juejin author tested ByteDance's Seedream 5.5 and OpenAI's GPT Image 2.6 with the same 15 prompts, selecting only the first generated image for each model. GPT Image 2.6 won 7 prompts, Seedream 5.5 won 4, and 4 were tied; the author found GPT stronger at following complex and logical instructions, while Seedream produced more appealing textures, poster layouts, and animation keyframes. The test also reported latency, output resolution, usage limits, and API pricing comparisons based on the author's own use.

    掘金 · 人工智能 · 🔥 0
  24. Doubao Launches “Find Store” Feature

    The available report states that the Doubao app launched a “Find Store” feature. It does not describe the feature’s functionality, rollout scope, or current user availability.

    DoNews · 🔥 0
  25. Doubao launches map-based local shop finder

    Doubao has launched a map-based “Find Store” feature that marks matching merchants and lets users view group-buy packages, order, and start navigation from the same conversation. Results can include per-person spending, operating status, ratings, group-buy packages, and vouchers, while AI can refine the list based on party size, meal period, and budget. The service also covers nearby attractions, hotels, gas stations, and entertainment venues.

    AIbase AI新闻 · 🔥 0
  26. How RAPTOR, GraphRAG, and OpenViking Structure RAG Knowledge

    A WeChat article explains that flat retrieval can struggle with statistical counting and cross-document relationships, then presents three approaches that organize knowledge before retrieval. RAPTOR builds a recursive tree of clustered summaries; GraphRAG represents entities and relationships in a knowledge graph with community summaries; and OpenViking maps agent memories, resources, and skills into a virtual file system. The article also discusses incremental updates, version-aware retrieval, and expiry cleanup as knowledge-governance practices.

    掘金 · 人工智能 · 🔥 0
  27. Brin and Zhang Yiming Return to AI Model Work

    TMTPost reports that Google co-founder Sergey Brin has resumed working with researchers on large-model architecture, while ByteDance founder Zhang Yiming has attended Seed core-team reviews regularly since the second half of 2024. The commentary frames both founders as re-entering technical leadership and cites a shared ambition to catch OpenAI and Anthropic by 2030, but it provides no primary confirmation of that target or several personnel details.

    钛媒体 · 🔥 0
  28. ByteDance Plans $6B Share Buyback at $400B Valuation

    ByteDance shareholders reportedly received a proposal to buy back up to $6 billion of shares at a valuation of about $400 billion, with the price set at $241.35 per share. The Information reported the plan, which was relayed by TMTPost. TMTPost characterized the transaction as a liquidity-management and valuation-anchoring measure for the still-private company.

    钛媒体 · 🔥 0
  29. Physical AI Challenges Corporate Information Controls

    A commentary from the WeChat account 针尖 argues that information asymmetries underpin organizational power and that physical AI tools can bypass established controls over confidential information. Using the reported Tencent layoff rumor and an alleged leak of a DeepSeek investor meeting, it discusses how AI recording, transcription and smart glasses may affect management, security and public relations. These incidents and their consequences are presented as arguments in the commentary and are not independently substantiated by the supplied report.

    虎嗅 AI · 🔥 0
  30. H3-Image, GPT Image 2.6 and Seedream 5.5 Face an Image-Model Test

    钛媒体 reports that MiniMax released H3-Image weights on November 10, after OpenAI introduced GPT Image 2.6 on November 5 and ByteDance released Seedream 5.5 on November 9. In hands-on tests through ChatGPT, Doubao and Hailuo AI, H3-Image produced images fastest, while GPT Image 2.6 showed stronger instruction following and editing. MiniMax’s claim of 16-megapixel images in 10 seconds at 50%-70% of comparable mainstream pricing was not fully reproduced, with standard mode taking about 25-30 seconds and high-speed mode 10-15 seconds in the report’s tests.

    钛媒体 · 🔥 0

Next →

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