ByteDance is a global technology company best known for developing and operating AI-driven content and short-form video platforms, including TikTok, Douyin, and Toutiao, serving billions of consumers worldwide. It also offers enterprise and B2B products such as Lark, BytePlus, and e-commerce infrastructure, and is actively investing in large language models and generative AI research through its ByteDance Seed team.
- Meta Goes All-In on Muse
- 🔥AI Labs Asked to Slow Down While NVIDIA Keeps Shopping - Here's the State of the Market in September 2026
- OpenAI Discloses More “Concerning” Model Behavior
- Dylan Patel – Anthropic & OpenAI will have most of the world’s compute by 2028
- 179_ 蒸馏风暴：一场无人公开谈论的技术竞赛
- 🔥Record Investments, Slowing Economy - Here's the State of the Market in August 2026
- The Memo - 14_Aug_2026
- Weekly TOP Startup _ VC news of August 13, 2026
- Google's AI Brain Drain, SpaceX's Huge Quarter, Airtable's 90% Collapse, US Data Fuels China AI
- AI Micro Dramas, Generative Media, and the Future of Creativity
- Chinese AI Models Are Increasingly Displacing OpenAI and Anthropic
- ⛽ Trading compute
“For whole-body teleoperation, a PICO 4 Ultra provides tracked human body poses”
Source→“Nvidia-backed Nscale... did not name ByteDance in its main S-1 even though the Chinese tech giant accounted for 73% of its revenue last year”
Source→“Duoshan (多闪) — ByteDance's earlier attempt at a Snapchat-like ephemeral social product”
Source→“today nothing's going on with me and my friend, I'll just come to Roblox to hang out”
Source→“Kuaishou — Extended case study of a company that survived competition against a much larger, resource-advantaged ByteDance because of deep, hard-to-migrate social network effects ('Laotie culture')”
Source→“This is probably not something Bytedance would do... a more common Bytedance way of doing things is: this data is good, so it keeps growing”
Source→“Maopaw — Liang's AI role-play/interactive-story app built inside ByteDance's Flow department, launched March 2024.”
Source→“new entrants shouldn't necessarily rush to build this: it's more important to first prove the experience is worth scaling before obsessing over paid acquisition infrastructure”
Source→“Xiang Liang (向亮) — ByteDance engineer who reportedly self-taught recommendation systems from a book he wrote himself”
Source→“They had an idea that short form video... can create a whole different experience that you watch and you have a meaningful connection before you go on to the next one.”
Source→“raised a $290 million round at a $1.5 billion valuation... ByteDance retains a 56% stake”
Source→“ByteDance Seed being the one outlier where they have significantly more than that.”
Source→“Zhang Yiming himself mentioned: if you go down the distillation path, at most you can continuously approach the other party — it is very difficult to achieve true surpassing.”
Source→“The new AI Data and Security department's head is Wang Yinglei, Adam Wang. He previously led TikTok's live streaming and reported to both Wen Jia and Alex.”
Source→“Zhang Yiming, in an internal meeting at Seed, mentioned that he does not allow ByteDance to do this distillation.”
Source→“ByteDance advances a colossal new AI model, intensifying global competition.”
Source→“FT reports ByteDance is "training a Mythos-scale, 10-trillion parameter model" — placing it at or near the cutting edge of global model scale.”
Source→“ByteDance advances a colossal new AI model, intensifying global competition.”
Source→“ByteDance is building at the frontier of model scale, not playing catch-up... FT reports ByteDance is 'training a Mythos-scale, 10-trillion parameter model'”
Source→“ByteDance is training a Mythos-scale, 10-trillion parameter model.”
Source→“Our base model follows GR-Dexter [47], a π0-like flow-matching VLA for dexterous manipulation.”
Source→“SiMDex uses only ~1.49M mined samples (<5% of the pool) yet improves the overall success rate from 47.7% to 61.1%”
Source→“ByteDance developed GR-Dexter (the base VLA model used in this paper) and the ByteDexter hand”
Source→“SiMDex retrieves only ~1.49M samples (under 5% of a 32M-sample egocentric video pool) and improves overall dexterous manipulation success from 47.7% to 61.1% — a +13.4 point gain”
Source→“Corresponding author with direct contact email provided (xuwei.robotic@bytedance.com). Leads the dexterous manipulation effort at ByteDance, including GR-Dexter and the ByteDexter hand.”
Source→“Author of both GR-Dexter (the base VLA) and the ByteDexter teleoperation paper [48]. This person is building both the hardware interface and the learning system”
Source→“ByteDance developed GR-Dexter (the base VLA model used in this paper) and the ByteDexter hand... This paper signals ByteDance is building a full dexterous manipulation stack — hardware, VLA model, and now data curation infrastructure.”
Source→“Seedance 1.5 Pro produces implausible hand scales, yielding a poorer interaction prior" (Section 4.2, Q3). Table 5 shows Seedance 2.0 significantly outperforms 1.5 Pro, especially on contact-rich tasks like Open Drawer (69.2% → 84.6% SR).”
Source→“On the optimizer used in V4, Kimi made important improvement contributions. V4's residual connection improvement MHC's original version came from ByteDance Seed's HC, and later Kimi was inspired by MHC and developed Tension Residues — though only proposed this spring, it's already being used in the latest K3.”
Source→“ByteDance alumni have an easier time fundraising, regardless of whether they have a research or product background. I think that comes from recent historical inertia — ByteDance was China's strongest tech company over the past five to eight years, with the strongest monetization and organizational strength.”
Source→“We are of course leveraged on the model as an application layer — we are inherently selling the model. Because your value is built on top of the model.”
Source→“We saw yesterday Seedance is now doing this with Seedance 2 Fast, which is much cheaper and faster.”
Source→“I think the trend we see with all of these models is that over time, the quality that you can get for a certain price goes up... I think Google did this with VO3 when they released VO3 Fast. And we saw yesterday Seedance is now doing this with Seedance 2 Fast, which is much cheaper and faster.”
Source→“The largest deals crossed asset classes, from DeepSeek, ByteDance, Moonshot AI and StepFun in Greater China.”
Source→“InternVLA-A1 is the strongest external baseline on UniVTAC at 67.1%, but N0-VTLA beats it at 83.1%. On NeoSim dual-arm tasks, InternVLA-A1 collapses to 1.0% (Tables 1–2)”
Source→“ByteDance is either the biggest or the second-biggest customer of Google Cloud for a reason, and they're getting many, many Blackwell from them, right? And the same with Oracle and the same with Microsoft.”
Source→“Chinese officials have held meetings with Alibaba, ByteDance, Z.ai, and other top tech companies about potentially restricting overseas access to China's most advanced AI models, as Beijing moves to treat homegrown AI as a national-security asset.”
Source→“ByteDance is making inroads in Hollywood with the quality and low cost of its Seedance AI video creation tool.”
Source→“ByteDance is making inroads in Hollywood with the quality and low cost of its Seedance AI video creation tool.”
Source→“X Square Robot…closed four consecutive financing rounds culminating in a Series C at a $2.8+ billion post-money valuation. Investors included IDG, HongShan, Xiaomi, Meituan, Alibaba, and ByteDance.”
Source→“MGX has written massive checks to the foundational AI models, including Anthropic. It also was part of ByteDance's deal to divest its U.S. operations.”
“Tether followed OpenAI at $500B valuation. A stablecoin issuer now ranks above ByteDance.”
Source→“What an incredible group of companies. And look at the diversity, SpaceX, Stripe, Anthropic, Databricks, Revolut, ByteDance, Android.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.