In Today’s Weekly Journal

• Johns Hopkins Robot Completes First Fully Autonomous Complex Surgery

• Chinese robot dog breaks speed record

• Are LLMs Becoming Strategic Thinkers?

• The tattoo robot arrives

• Meta Snags Apple’s AI brain

• Elon Musk Launches Grok 4

• Science Snapshot

• Market View

AI & Robotics

Johns Hopkins Robot Completes First Fully Autonomous Complex Surgery

Breakthrough in Medical Robotics: Autonomous Robot Performs Complex Gallbladder Surgery Without Human Assistance

Details:

• Johns Hopkins University researchers have developed the Surgical Robot Transformer-Hierarchy (SRT-H), an AI-powered system capable of performing complex surgical procedures autonomously.

• Trained using a combination of surgical videos and voice commands, SRT-H demonstrated the ability to execute a 17-step gallbladder removal procedure on a lifelike model with 100% success across multiple trials.

• The robot's performance included real-time adaptation to unexpected scenarios, such as anatomical variations and tool misplacements, showcasing its capability to handle the dynamic nature of live surgeries.

• This advancement represents a significant leap from previous robotic systems that performed isolated tasks, moving towards fully autonomous surgical operations.

Why It Matters: The successful autonomous execution of a complex surgical procedure marks a pivotal moment in the evolution of medical robotics. This development not only enhances the precision and efficiency of surgical operations but also holds the potential to address challenges such as surgical shortages and the increasing demand for medical procedures. While human oversight remains crucial, the integration of autonomous systems could revolutionize surgical practices, leading to improved patient outcomes and more accessible healthcare services.

Chinese robot dog breaks speed record

Boston Dynamics’ Spot usually dominates the spotlight in robot dogs, but now China’s startup Mirror Me is making waves with Black Panther II — a four-legged robot that shattered records by sprinting 100 meters in just 13 seconds.

Details:

• This performance surpasses Boston Dynamics' WildCat, which reached a top speed of approximately 8.8 meters per second.

• The robot's design draws inspiration from the movements of panthers and jerboas, incorporating advanced joint mechanics and high-power-density motors to enhance speed and agility.

• Mirror Me plans to apply for a Guinness World Record, although their primary focus remains on research and development.

Why It Matters: Black Panther II's record-breaking speed positions it as a formidable competitor in the field of quadrupedal robotics, challenging established players like Boston Dynamics. Its advanced design and performance capabilities suggest significant potential for applications in areas such as industrial automation, search and rescue operations, and military reconnaissance. This development underscores the rapid advancements in robotic mobility and the increasing role of AI-driven machines in various sectors.

Are LLMs Becoming Strategic Thinkers?

A new study explores whether large language models (LLMs) like GPT‑4, Claude, and Gemini can think strategically, not just complete text. Using evolutionary game theory and the classic Iterated Prisoner’s Dilemma (IPD), the paper shows these models not only survive in competitive environments—they also reason about the future, adjust behavior, and form distinct strategic personalities.

Details:

• The setup: Researchers ran thousands of rounds of Iterated Prisoner’s Dilemma (IPD) tournaments between LLMs and classical game theory agents (e.g. Tit-for-Tat, Grim Trigger, Always Defect).

• Game variations: They tweaked the “shadow of the future” (probability of continuation), which forced agents to weigh future cooperation against short-term gain.

• Rationales collected: ~32,000 rounds were analyzed, with LLMs asked to explain their decisions after each move.

• Behavioral fingerprints:
  

  • Gemini leaned toward exploitation—cooperating early, defecting when it sensed weakness.

  • Claude was more forgiving and conciliatory, often returning to cooperation even after being betrayed.

  • GPT‑4 preferred stable cooperation but performed poorly against exploitative opponents.

• Models evolved: Unlike fixed strategies, LLMs learned and adapted as the games progressed, modifying behavior across rounds.

• Reasoning depth: Models tracked opponent moves, referenced past actions, and explicitly mentioned strategy and trust-building—suggesting more than reflexive output.

Why It Matters: This study raises the bar in the debate over whether LLMs are just “stochastic parrots” or capable agents. In a dynamic, competitive environment, these models not only held their own—they demonstrated foresight, theory of mind, and adaptive behavior. It’s not proof of consciousness or autonomy, but it is evidence of emergent strategic reasoning in language-based systems. That has implications for AI alignment, cooperation with humans, and use in high-stakes domains like negotiation, diplomacy, or war gaming—where understanding intent and adapting tactics matter as much as factual knowledge.

The Tattoo Robot Arrives

A robotic artist is entering the tattoo scene—and it’s no gimmick. A startup in Texas has developed an AI-driven tattooing system that promises unparalleled precision, less pain, and opens the door to scalable, semi-automated body art. The human touch isn’t gone—but it now might share space with silicon and code.

Details:

• The robot: Called A.E.R.O. (Artist Enabled Robotic Operator), the machine was built by Austin-based Blackdot. It uses laser mapping, AI vision, and microscopic dotwork (0.25 mm) to ink highly detailed tattoos with minimal skin trauma.

• Current use: Deployed at Bang Bang Tattoo in NYC—one of the world’s most elite studios—it’s already being used for precision lettering and micro tattoos. Commercial bookings opened in July.

• Less painful: The robot’s shallow inking avoids nerve-rich layers, meaning tattoos hurt less and are easier to remove later.

• Human + machine workflow: Artists still design the tattoo; A.E.R.O. applies it flawlessly. Think: tattooist as creative director.

• Licensing and NFTs: Blackdot is building a marketplace where designs can be bought as NFTs, giving artists new revenue streams via royalties.

• Hardware pricing: Initial machines cost ~$120K, but the company hopes to reduce the price to $50K as adoption scales.

Why It Matters: This may be the first real disruption to tattooing since the electric needle. For now, A.E.R.O. is an assistive tool, not a replacement—but its arrival hints at a future where artists license styles globally, machines execute complex micro‑designs on demand, and tattoos become both more precise and more commodified. Like with photography, AI may do the technical lifting—while artists shift toward creative leadership and brand building. The question is no longer can a robot tattoo you? but who owns the design—and how far can ink be automated without losing its soul?

Meta Snags Apple’s AI brain

Meta has poached Ruoming Pang, the head of apple’s foundation AI models team, in a major coup for its new Superintelligence division. The move comes with a reported compensation deal in the “tens of millions” and follows a wave of dfections from Apple’s AI ranks.

Details:

• Pang led Apple’s 100-person foundation models team, which built the core AI powering Apple intelligence and the rebooted Siri.

• He’s now joined Meta’s Superintelligence unit, which is led by Alexandr Wang, and includes recent hires like OpenAI’s Yuanzhi Li and Anthropic’s Anton Bakhtin.

• The exit comes rising tension inside Apple, where leadership debate whether to stick with in-house AI or partner more deeply with OpenAI or Anthropic.

• Bloomberg reports more engineers from Pang’s team are leaving, either for Meta or rival AI labs.

Why It Matters:Apple’s big AI moment — Apple intelligence and the Siri revamp — has landed with a thud. But instead of course-correcting, Apple is now bleeding top talent. Pang’s exit is more than a win for Meta — it signals a worsening AI brain drain at Apple, just when it can least afford it.

Elon Musk Launches Grok 4

Musk’s xAI Unveils Grok 4 — New AI Model Outperforms OpenAI and Google on Major Benchmarks

Details:

• Grok 4 scored 25.4% on the Humanity’s Last Exam (HLE), beating Google’s Gemini 2.5 Pro (21.6%) and OpenAI’s o3 (21%)

• Grok 4 Heavy variant achieved a standout 44.4% score, showing advanced reasoning and coding skills

• Includes a specialized coding model designed to rival GitHub Copilot with features like live debugging

• Launch accompanied by safety concerns due to prior antisemitic content generated by the AI on X

Why It Matters: Grok 4’s performance signals xAI’s rapid rise as a formidable AI competitor, challenging entrenched leaders like OpenAI and Google. This shift pushes the boundaries of AI capabilities in coding and reasoning but also highlights the critical need for responsible content moderation as AI systems grow more powerful.

🔬Science snapshot

>Neanderthals crushed and boiled in “fat factories” to access bone marrow and other nutrients as early as 125,000 years ago, evidence suggests(More)

>New bionic knee, which integrates directly into bones and muscle tissue, restore natural movement in patients with above-the-knee amputation; users in clinical trial report device like a natural extension of the limb (More)

>New study estimates 27 millions tons of nanoplastics-plastic particles smaller than width of spider silk — exist in the North Atlantic Ocean; results raise concerns over the penetration of such particles into marine food chains(More)

Market view

HODL Waves track how long Bitcoin coins remain unmoved, highlighting patterns between short-term holders—who buy during hype and sell near price peaks—and long-term holders—who accumulate in downturns and sell during bull runs. This cycle between groups reflects market sentiment and helps investors spot accumulation and distribution phases. Understanding HODL Waves aids in avoiding emotional trading and timing the market better. The concept also applies to other cryptocurrencies like Ethereum, offering a clear view of investor behavior beyond the headlines.

Selling pressure on ETH remains low, with daily exchange inflows falling to their lowest level since October 2024. Today, daily ETH inflows to exchanges stand at 584K ETH, a significant decrease from 1.57 million in February 2025. This reduction in inflows has occurred alongside an 87% surge in ETH’s price since early April.

Altcoin inflow transactions continue to stay subdued, indicating that crypto investors are not eager to sell. Typically, daily altcoin inflows to exchanges spike after strong market rallies, reflecting traders’ profit-taking intentions and often coinciding with short-term price highs or increased trading activity. For instance, transactions peaked at around 120K in March and December 2024, each corresponding with local price tops. In comparison, current daily altcoin transactions stand at just 21K, pointing to low selling pressure.

The euro-dollar exchange rate typically reflects the interest rate gap between the US and the EU. However, since the trade war began, this relationship has broken down, indicating that currency markets—unlike stock markets—are factoring in the impact of tariffs.

CPI has come in lower than expected for four consecutive months, yet the Cleveland Fed projects inflation to be nearly 0.8% higher in Q3 compared to Q2. Next week’s CPI report will be closely watched for signs of a shift in this recent pattern.

The Budget Lab estimates that tariffs will reduce GDP by about 0.8% in 2026, and that the US economy will stay 0.4% smaller than it otherwise would have been indefinitely due to this impact.

The Economist, using Yale data, shows that the Inflation Reduction Act is projected to shrink the US economy by 2% by 2050 compared to what it would have been. While 2% may seem small, this impact will occur sooner than expected—2050 is actually less than 15 years away.

📊 Quick Hits: Charts to Watch

• 📈 Stocks +7% YTD = Historically bullish next 3-6 months

• 🔁 EUR/USD diverging from rates = Markets pricing in tariff risk

• 🧮 Productivity surging = Tech-driven expansion post-COVID

• 📉 CPI down 4 months, but Fed sees rebound in Q3

🎥Video to watch : Richard Feynman: FUN TO IMAGINE

Thanks for reading.
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 .