Daily TEA – Hedge Funds, Humanoids, and the Alien LLM Autopsy

DeepSeek war chest, Dubai crypto rules, safety apps, humanoid robots, alien LLM biology

Hello, dear TEA-mates — here’s what you need to know today.

1.🧠 DeepSeek Founder’s Hedge Fund Fuels AI War Chest

DeepSeek founder Liang Wenfeng’s hedge fund platform, Zhejiang High-Flyer Asset Management, delivered an average return of about 56.6% in 2025, ranking among China’s top quantitative funds and dramatically expanding his resources for AI development. The firm, which manages over 70 billion yuan (around 10 billion dollars), shifted from market-neutral to long-only equity strategies and generated some of the industry’s highest risk-adjusted returns, with performance and management fees estimated to exceed 700 million dollars—far above DeepSeek’s reported single-digit-million model budget. Read More (TipRanks).​

🫖 TEA For Thought: DeepSeek can keep developing advanced AI without relying entirely on outside investors, making it a very strong competitor in the global AI race.

2.🏙️ Dubai Hands Crypto Token Vetting to Licensed Firms

Dubai’s Financial Services Authority has revamped its crypto token regime by transferring responsibility for assessing token “suitability” to licensed firms operating in the Dubai International Financial Centre, rather than maintaining a central list of approved assets. Regulated companies must now document and regularly review their token assessments to meet anti-money laundering and suitability standards, while separate rules from the Virtual Assets Regulatory Authority still restrict privacy coins in most of Dubai outside the DIFC. Read More (CoinMarketCap Academy).​

🫖 TEA For Thought: This is definitely another milestone in cementing the UAE as a center for emerging tech, including crypto and AI.

3.🤖 Humanoid Robot NEO Uses Video AI to Learn New Skills

Robotics company 1X has upgraded its humanoid robot NEO with the “1X World Model,” a video-based AI system that lets the robot turn simple text or voice prompts into new physical actions it was not explicitly trained to do. By learning from online video and grounding that knowledge in real-world physics, NEO can autonomously plan and execute everyday tasks like making a lunch box or opening doors in unfamiliar environments and then improve over time using data from its own experience. Read More (Interesting Engineering).​

🫖 TEA For Thought: With LLMs, video/audio models, world models, and humanoid hardware converging, AGI may emerge in scenarios where robots learn by watching videos like humans and then iteratively improve through trial and error.

4.📱 ‘Are You Dead?’ App Becomes Viral Safety Tool For Solo Dwellers

A bleakly named Chinese app called Are You Dead? has surged to the top of the country’s paid download charts by targeting people who live alone and fear facing emergencies unnoticed. Users tap a large button every one or two days to confirm they are alive or else the app alerts a designated emergency contact, and despite criticism of its ominous name, it has spread internationally under the name Demumu while the creators consider a rebrand and explore versions tailored to elderly users. Read More (BBC).​

🫖 TEA For Thought: This product definitely found its product-market fit, becoming a striking phenomenon in the era of AI and digital loneliness.

5.👽 Scientists Study LLMs Like Alien Life-Forms

Researchers featured by MIT Technology Review are treating large language models as if they were unfamiliar biological organisms, using experimental techniques to systematically probe how internal “circuits” give rise to specific behaviors. By running controlled interventions and tracking how model states change, they aim to explain surprising or undesirable behaviors, highlighting that even leading experts still do not fully understand how today’s most powerful LLMs work. Read More (MIT Technology Review).​

🫖 TEA For Thought: Sometimes we forget that even expert scientists do not completely understand how large language models work.

Prompt Tip of the Day: The 30-Day Fast Learning System

Builds a long-term learning edge.

Create a 30-day learning faster plan. Break it into weekly themes: 
Week 1: Clarity 
Week 2: Understanding 
Week 3: Recall 
Week 4: Application 
Give daily learning tasks under 30 minutes.

TEAHEE Moment

Stay sharp, stay informed. See you tomorrow.

Follow along on X for more real-time sips of TEA: Twitter/X.