- Horizon AI
- Posts
- Perplexity Wants to Buy TikTok đź’°
Perplexity Wants to Buy TikTok đź’°
How to use RunwayLM's "Expand Video" tool
In partnership with


Welcome to another edition of Horizon AI,
The AI search startup laid out its bid to buy and rebuild TikTok, introducing significant changes to enhance the platform with AI and make it more transparent.
Let’s jump into it!
Read Time: 4.5 min
Here's what's new today in the Horizon AI
Perplexity AI Makes a Bid to Merge with TikTok U.S.
Tencent’s Hunyuan-T1 AI Reasoning Model Rivals OpenAI's o1 Capabilities
AI Tutorial: How to use RunwayLM's "Expand Video" tool
AI Tools to check out
The Latest in AI and Tech đź’ˇ
AI Findings/Resources
AI News
PERPLEXITY
Perplexity AI Makes a Bid to Merge with TikTok U.S.

The AI search startup Perplexity just proposed a bid to acquire (and transform) TikTok, claiming they’re uniquely positioned to do this “without creating a monopoly.”
Details:
Perplexity intends to rebuild TikTok’s algorithm in American data centers under U.S. oversight, making it “transparent” and open source.
Plans include integrating Perplexity’s search tools, adding multilingual translation, and enabling AI citation capabilities for videos.
It also proposes upgrading the AI infrastructure using Nvidia Dynamo technology to scale TikTok's recommender models 100x while achieving faster inference speed.
Finally, it would offer personalization for users who connect their Perplexity and TikTok accounts, which the startup claims “could help surface more relevant content on TikTok.” Similarly, content preferences on TikTok could inform more personalized answers on Perplexity.
ByteDance has been reluctant to sell TikTok’s U.S. operations, even as the 75-day delay on the TikTok ban signed by President Trump is set to expire soon. Meanwhile, Perplexity faces stiff competition from tech giants like Oracle and Microsoft, as well as investor groups.ity faces stiff competition from tech giants like Oracle and Microsoft, alongside investor groups
TOGETHER WITH PROTON
Google and Apple can't be trusted with your private data
Big Tech profits off you without your consent, so take back control of your privacy with Proton Pass. Your data is end-to-end encrypted so no one can access it except you. Keep your passwords, credit cards, and personal information at hand and share securely with friends and family, even if they're not using Proton.
With Proton Pass you can:
Store unlimited passwords
Share safely with friends and family, even if they don't have Proton accounts
Create 10 email aliases to hide your personal email address
Get free access to our VPN, mail, and drive service
TENCENT
Tencent’s Hunyuan-T1 AI Reasoning Model Rivals OpenAI's o1 Capabilities

Tencent’s latest reasoning model, Hunyuan-T1, can match OpenAI's best reasoning models, according to the company.
Details:
Hunyuan-T1 delivers strong performance on most benchmarks, scoring 87.2 on MMLU-PRO—second only to OpenAI's o1—and 96.2 on MATH-500, just behind Deepseek-R1.
Tencent used curriculum learning to gradually increase task difficulty during training and incorporated a self-reward system for iterative improvements.
The model’s Transformer Mamba architecture processes long texts twice as fast as traditional models, making it more efficient.
This release follows Baidu's recent introduction of its own o1-level model and Alibaba's before that, all pursuing open-source strategies. Hunyuan-T1 is accessible through Tencent Cloud, with a demo available on Hugging Face.
AI Tutorial
How to use RunwayLM's "Expand Video" tool

This new feature allows you to transform videos into new aspect ratios by generating new areas around your input video.
Go to the RunwayML and sign in.
Select "Generate Video" and choose the Gen-3 Alpha Turbo model.

Click on “Expand Video”
f you upload a horizontal video, the model will automatically default into turning it into a vertical video.
You’ll have a few options:
a) Click the Generate button without providing a text prompt.

b) Provide a detailed text prompt telling the model what you’re hoping to see in the parts of the frame that aren’t visible. For example: “A man standing in a cave with a molten lava pool in the foreground and rock formations overhead.”

c) Provide the model with an outpainted image of the first frame of your video

The feature works best when:
Your subject is close to the center of the frame
There are no graphics or text present in the video
You use a text prompt to help guide the generation.
AI Tools to check out
🔥 ChatLLM Teams - Abacus.AI: One AI assistant for you or your team, offering access to state-of-the-art LLMs and features such as web search, video generation, and more.
đź“Š Basejump: AI data analytics platform.
đź“˝ Eddie AI: ChatGPT for video editing.
âś… Fontjoy: Generate font combinations with deep learning.
đź‘” Nowadays: AI-powered corporate event planner.
AI Findings/Resources
🎮 Why Anthropic’s Claude still hasn’t beaten Pokémon
đź‘€ Meet the humans building AI scientists
👨‍💻 Here’s how I use LLMs to help me write code
🎤 An Interview with OpenAI CEO Sam Altman about building a Consumer Tech Company
The latest in AI and Tech
The tool allows users to check if their work appears in LibGen, a massive archive of pirated books, scientific papers, and articles reportedly used to train language models.
The company plans to integrate cameras into Apple Watch models to analyze surroundings and provide relevant information, with camera placement varying by model.
X user Jonah Manzano, known for testing new social media features, spotted a “Write with Meta AI” prompt on Instagram, offering AI-generated comment suggestions.
Google is adding new AI features to Gemini Live that let it see your screen or smartphone camera and answer questions in real time. The update comes nearly a year after Google first showed the “Project Astra” technology behind it.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻‍💻