Oprisa's Journal

August 17, 2025

“100% of the information you have about any business reflects the past, and 100% of the value of that business depends on the future.”  Bill Miller

 

In Today’s Weekly Journal

AI & Robotics

OpenAI is reportedly considering an investment in Merge Labs, an $850 million–valued brain–computer interface startup co-founded by Sam Altman, with plans to directly rival Elon Musk’s Neuralink.

Details:

  • Newton is a foundational AI “Large Behavior Model” that learns complex physics principles directly from raw sensor data—without any human guidance or pre-existing theoretical knowledge.

  • It not only predicts familiar systems like pendulums but also generalizes to unfamiliar situations—surpassing traditional, specialized AI models in tasks such as citywide power consumption forecasting.

  • Developed by Archetype AI, a team founded by ex-Google researchers, the project has secured $13 million in funding to date.

  • Newton represents a paradigm shift—moving from narrowly trained models toward a single, adaptable system capable of autonomous learning and interaction with the physical world.

Why it matters: This matters because OpenAI backing a Neuralink rival escalates the Altman–Musk rivalry and highlights the high stakes in brain–computer interface technology, showing Altman’s ambition to expand into frontier tech beyond AI

The location for the World Robot Conference 2025 in Beijing, China | Credit: World Robot Conference

Beijing has unveiled what is widely regarded as the world’s first Robot Mall—a four-story, fully integrated "4S" retail complex for humanoid and consumer-oriented robots. Launched alongside the 2025 World Robot Conference, this mall blends cutting-edge robotics with immersive consumer experiences.

Details:

  • The mall mirrors the automotive industry’s succesful “4S” framework, offering Sales, Services, Spare Parts, and Surveys all under one roof to support full lifecycle customer engagement.

  • Visitors can explore over 100 robot model from 40+ Chinese brands, including familiar names like Ubtech and Unitree.

  • Prices range from about 2,000 yuan (US $278) for basic models to several million yuan for advanced humanoids. A standout example: a life-sized, talking Albert Einstein robot priced at around 700,000 yuan (US $97,000).

  • Visitors enjoy hands-on interactions including robots playing soccer or chess, robotic pets, robot waiters and chefs serving meals in an adjacent themed restaurant, and performance zones featuring robot dancers and musicians.

Why it matters: It matters because China’s robot mall marks a major shift from experimental robotics to everyday consumer use, backed by massive government investment and designed to build trust while accelerating adoption and innovation in the sector.

OpenAI revealed that its reasoning model earned a gold medal at the 2025 international Olympiad in informatics (IOI), ranking 6th overall among human competitors and securing the top spot for AI in the world’s pre-college programming contest.

Details:

  • The AI faced off against leading student programmers from around the globe, tackling coding challenges under identical times and submission constraints as its human rivals.

  • OpenAI’s model functioned as a general-purpose reasoner, using only basic tools and no specialized fine-tuning for programming tasks.

  • The system reached the 98th percentile, a remarkable improvement from its 49% performance only a year earlier.

  • The same model also secured gold at the international Math Olympiad and AtCoder, highlighting its capabilities diverse and complex problem-solving challenges.

Why it matters: The doubling of its score demonstrates just how rapidly reasoning abilities have advanced over the past year. Human dominance in competitions is increasingly fleeting, and these milestones may pave the way for future models capable of breakthroughs in science, mathematics, physics, and beyond.

image by CCTV

A recent segment aired on China’s state broadcaster CCTV showcased the PLA’s newest innovation: four-legged armed robots—dubbed “robot wolves”—marching into a simulated battle alongside infantry. This dramatic display marks the combat debut of these advanced quadrupedal units, signaling a major leap forward in unmanned warfare.

Details:

  • Each 70 kg unit carries a QBZ-191 assault rifle, operating in coordination with soldiers and drones across challenging terrain.

  • Robots work in coordinated “wolf packs,” taking on roles like scouting, shooting, or carrying ammunition.

  • Their deployment reflects China’s push for human-robot collaboration, aiming to enhance combat efficiency and reduce human risk.

  • Robot wolves are equipped with AI-driven targeting system and multi-sensor suites — including cameras, lidar, and infrared — allowing them to detect enemies, navigate obstacles, and coordinate with other units in real time during operations.

Why it matters: Integrating AI-driven targeting and multi-sensor systems gives the robot wolves true battlefield autonomy — enabling them to navigate, identify threats, and act with minimal human guidance. This not only increases their effectiveness in complex environments but also accelerates the shift toward fully autonomous combat operations, raising both strategic advantages and ethical concerns.

NVIDIA is pushing the boundaries of robotics with its innovations, introducing new Omniverse libraries, Cosmos physical AI models, and enhanced AI computing infrastructure. These tools aim to accelerate the development and simulation of intelligent robots, bridging the gap between virtual testing and real-world applications.

Details:

  • NVIDIA introduced Omniverse libraries to streamline robotics simulation and development.

  • The new Cosmos physical AI models allow robots to learn and interact more realistically in virtual environments.

  • NVIDIA’s platform offers advanced AI computing infrastructure to handle complex robotics workloads.

  • These tools are designed to accelerate the transition from virtual simulations to real-world robot applications.

Why it matters: This matters because it gives developers and researchers the tools to train and test robots in highly realistic virtual environments before deploying them in the real world, reducing costs, speeding up innovation, and improving safety.

🔬Science snapshot

>Scientists develop platform allowing proteins to evolve thousand of times faster than in nature; approach expected to accelerate new drug discovery for a wide variety of diseases (More)

>Study suggests oxytocin, also referred to as the “love hormone,” plays a role in the selectivity of social connections; oxytocin-deficient rodents took up to a week to form bonds, compared to 24 hours for control group (More)

>Physicists use sound waves to store quantum information; new approach can store quantum memory for 30 times longer than conventional techniques (More)

>Remains of Dennis Bell, an Antarctic researcher who disappeared in 1959, discovered by Polish team studying a receding glacier on King George Island (More)

Market view

I’m not a financial analyst—just someone who digs through the daily market noise and pulls out the insights worth knowing. Think of it as me doing the heavy lifting so you can get the highlights that actually matter.

Markets are acting oddly disconnected from traditional economic signals: inflation is ticking up, deficits are ballooning, and a recession looms — yet Treasury yields are flat, and stocks keep hitting record highs.

Some expert, like Aswath Damodaran, believe the market has shifted from predicting the future to reacting in the moment after years of failed forecasts. Others think it’s still predicting — just forecasting strong gains ahead thanks to expected interest rate cuts, potential tariff repeals, and higher corporate earnings. Investor behavior may also be driven more by fear than risk, with the AI boom adding urgency. For some, the greatest danger isn’t economic collapse but missing out on gains before AI potentially reshapes or even takes over the economy.

CHARTS.

The M2 money supply is rising again, fueled by increased bank lending and government deficit spending — trends likely to continue. According to the FT, the AI boom has sparked “ a race among private capital providers” to issue bigger and more frequent loans for building data centers.

The capital expenditure race kicked off with ChatGPT, and it’s still in its early stages. The big four hyperscalers are projected to pour $350 billion into data centers in 2025, followed by another $400 billion in 2026.

Cash flow won’t be enough — Morgan Stanley projects hyperscalers will invest $2.9 trillion in data centers over four years, with less than half of that covered by their cash generation.

Producer prices jumped in July, indicating companies may be starting to pass tariff costs on to consumers. Goldman Sachs estimates that through June, consumers absorbed only 22% of these costs while businesses absorbed about two-thirds through reduced margins. By October, Goldman expects the balance to reverse, with consumers bearing roughly two-tirds of the burden.

For the first in a decade, wage growth for low-income workers has fallen below that of high-income earners. As the Wall Street Journal puts it, “The era of big raises for low-paid workers is over.” if that’s true, it was a remarkable run while it lasted.

@&.2% of stocks in the S&P 100 trade at a P/E of at least 50. Just one trades 10x.

Ethereum (ETH) is nearing a new all-time high and has entered a bullish cycle against Bitcoin, with the ETH/BTC ratio above its 365-day moving average.

Investor focus is shifting toward ETH, with rising ETF allocations, higher open interest, and stronger spot trading volumes compared to Bitcoin.

However, increased ETH inflows to exchanges and its relative value approaching historical overvaluation levels suggest potential profit-taking and short-term resistance.

If you found this valuable, consider sharing or forwarding it to a friend who needs to be ahead of the curve. 🧠📩

📬 Until next time,

— The Macro Scratch team, with AI-powered research and support.