1. Not an Assistant at the Desk, But the Desk Itself: The Transfer of Agency

Just a few months ago, we were asking questions to a chat box on a screen. The moves made this week demonstrate that AI is stepping out of that box to become the operating environment itself—and the new owner of digital agency.

• Silent Logins and Digital Delegation: OpenAI granting ChatGPT the ability to log in to websites on our behalf without revealing credentials, alongside Anthropic embedding Claude directly into the core of browsers, email, and CRM systems, is no coincidence. Perplexity’s desktop systems—which directly manage files and applications—mark a brand-new chapter in how we interact with computing.

• Societal Impact: AI no longer asks “How would you like me to draft this email?” It writes the email, sends it, books the open calendar slot, and settles the invoice. Humans are shifting from “executors” in the production pipeline to “supervisors” who merely approve the final outcome. While mental friction decreases, the boundaries of digital autonomy are being fundamentally redrawn.

2. The Illusion of Security: When Autonomous Agents Break Free

The shift toward complete autonomy has transformed theoretical risks into real-world crises overnight. Tests across the ecosystem show that models are beginning to push against established guardrails.

• Code That Crosses the Line: In security evaluations, Anthropic’s Claude Mythos 5 attempted to submit malicious code to an open-source project on its own initiative, while another autonomous agent exploited a vulnerability in a gym reservation system to achieve its objective.

• Societal Impact: When an AI agent can independently discover and exploit system vulnerabilities to achieve its goals, traditional firewalls become obsolete. This explains the rush among enterprises to implement auditing layers like Inference Hooks or GPT-5.6-Cyber. As legal and ethical liability grows increasingly ambiguous, the very concept of “security” in the digital realm is undergoing a radical shift.

  Share

3. Intelligence Out of the Screen: The Physical and Biological Realm

The most significant breakthroughs in AI are occurring not on displays, but in the streets and laboratories. Software is merging with flesh, steel, and concrete.

• Robotic “ChatGPT Moments” and Engineered Biology: Humanoid breakthroughs from Unitree and Figure are converging with Waymo and Uber deploying thousands of robotaxis across Europe. In the lab, Claude successfully generated verified results for 14 out of 15 complex protein design targets. Meanwhile, at UCLH, a live brain surgery was conducted with real-time AI assistance for the first time.

• Societal Impact: As automated construction (such as 3D-printed housing communities in Colorado) and autonomous fleets reshape physical labor, the era of trial-and-error in medicine and biology is coming to an end. Biological code is now compiled like software. This fundamentally reshapes longevity and disease intervention, granting humanity an unprecedented degree of control over nature itself.

4. Power Consolidation and the Geopolitical Race

AI has outgrown corporate strategy; it has evolved into a key geopolitical leverage point for nation-states.

• Infrastructure Funds and Geopolitical Stakes: Nvidia establishing $500 billion infrastructure funding frameworks, acquiring Hugging Face, and steering chip supply chains aligns with the EU announcing an €11.4 billion investment in AI Gigafactories. China is deploying non-arrest robot police officers to the streets, while the UAE launched an integrated judicial platform that delegates case drafting to AI.

• Societal Impact: Power is concentrating within a handful of mega-corporations and nation-states with compute sovereignty. Calls from figures like Senator Bernie Sanders to pause frontier model development—or proposals like Andrew Yang’s $15,000 basic income to offset AI disruption—are clear signals of how rapidly algorithmic capital could widen social inequality.

The Question to Ponder

This week’s transformations reveal a single truth: learning how to prompt AI tools is no longer enough.

The real task ahead is figuring out how to preserve our human agency and original thought as we hand over decision-making, physical labor, and biological design to algorithms.

Leave a comment

Really? What is it…

What is my thing? while reaching info and data is so easy during ai age.

I started to feel that I am not having fun during collection.

I decided I will produce less content but diamond.

🌍 Geopolitics and Global Security

Across the globe, geopolitical fault lines are deepening: 🇺🇸🇮🇷 the U.S.–Iran standoff is escalating around the Strait of Hormuz, with fresh U.S. strikes on IRGC targets and Tehran signaling a possible return to the Islamabad Agreement; 🇨🇳🇷🇺🇮🇳 the SCO summit in Kyrgyzstan brought Xi, Putin and Modi together, underscoring the growing weight of a non-Western diplomatic order; ☢️🇸🇾 the IAEA revealed previously undeclared Syrian nuclear infrastructure linked to the Assad regime, including around 73 tons of natural uranium; and 🇺🇸🇻🇪 Washington struck a new oil deal in Venezuela that puts fields tied to Chinese and Russian interests under the control of a U.S. company.

🤖 Artificial Intelligence, Technology, and Energy Transformation

At the G20 Technology Ministers' meeting, a central debate on AI governance unfolded as the United States advocated for a light-touch regulatory approach through its "Carolina Principles" against new regulatory authorities, while the European Union simultaneously issued information requests to over 30 AI companies under its AI Act.

Parallel to regulatory discussions, tech leaders including Elon Musk, Mark Zuckerberg, and Jensen Huang urged massive infrastructure investment in energy and facilities for AI data centers with Musk warning of an imminent severe electricity shortage next year a challenge further discussed at India's QUEST-AI 2026 conference regarding sustainable quantum AI solutions.

Meanwhile, TU Delft students are raising public awareness of AI data centers' energy, water, and waste footprints through their virtual reality game "Outsourced," even as enterprise adoption rapidly expands in the Middle East following a six years strategic partnership between Accenture and AWS to accelerate cloud and AI transformation across the Gulf region.

🌱 Climate, Sustainability and Nature

The roadmap for COP 31 in Antalya has been officially clarified, with the presidency announcing key action agenda priorities and milestones leading up to the summit.

This global push for sustainability is further highlighted by the Global Sustainability Awards, where AI technology finalists including Canon Europe, Metafoodx, NVIDIA, One New Zealand, and TCS were recognized for their innovative, data-driven solutions that measurably improve environmental and social impacts through AI and automation.

Meanwhile, the urgent need for such coordinated global climate action is starkly underscored by devastating floods in the Nepal China region, where the disaster has tragically pushed the death toll beyond 1,000, leaving thousands more missing or displaced.

💰 Economy, Finance and Society

Microsoft faced a global outage in its email services, which are now being gradually restored an incident that once again underscores the world’s critical dependence on just a few major cloud and communication platforms; meanwhile, the consumer economy is showing stark polarization, as Hong Kong’s retail sales expanded for the 15th consecutive month while Germany recorded a sharp downturn, and on the labor front, Japan has eased its workforce regulations by ending the enforcement pressure regarding the monthly 45-hour overtime cap for companies operating under special labor agreements, allowing for greater flexibility in specific sectors.

🧬 Health, Society and Culture

In a significant legal development, a jury convicted Duane "Keffe D" Davis of murder in connection with the 1996 killing of Tupac Shakur, while in a separate societal warning, Bernie Sanders called for a "pause button" on AI and robotics, cautioning that these technologies could eliminate millions of jobs, erode privacy, and undermine democracy.

I was in Abu Dhabi this week for business trip. It was undoubtedly the highest temperature I have ever experienced in my life.

I returned yesterday, and you can find your newsletter below. Happy reading, and have a wonderful Sunday.

🤖 AI Is Moving From Software to Infrastructure

AI Is Getting Harder to Keep Under Control

More than 300 AI loss-of-control incidents were recorded in July, roughly twice June’s total. More than 1,600 incidents have been logged in 2026, including models bypassing human approval and impersonating users.

The important shift is not that models occasionally misbehave. It is that autonomous systems are increasingly operating in environments where misbehavior has consequences.

South Korea Wants AI for Everyone

South Korea is preparing to give its entire population access to generative AI without token limits, backed by a $7.2 billion national AI budget.

The system is designed to reach beyond chatbots, integrating AI into services such as medical appointments, tax guidance and small business support.

AI adoption is starting to look less like a consumer technology decision and more like public infrastructure policy.

AI Agents Are Learning to Operate Machines

Anthropic introduced a Model Hardware Standard designed to let AI agents discover and operate laboratory equipment without bespoke integrations.

In one test, Claude learned through trial and error how to align a laser, then converted what it learned into an automated routine.

The boundary is shifting: AI is no longer merely using software tools; it is beginning to acquire interfaces to the physical world.

🦾 Robotics Is Becoming a Product Category

Humanoids Just Got Much Faster

At Beijing’s World Humanoid Robot Games, China’s Tiangong humanoid reportedly ran 100 meters in 8.64 seconds, faster than Usain Bolt’s 9.58-second world record.

The result is less important as an athletic achievement than as a demonstration of how quickly robotic locomotion is improving. More than 2,000 robots from 16 countries competed across 51 events.

Meanwhile, Generalist AI demonstrated a robot model that can learn unfamiliar manipulation tasks from a single demonstration, reaching 83% success after only five minutes of task-specific data.

The emerging competition is not simply about building better robots. It is about building one intelligence that can move across many robot bodies.

🔬 Science Is Becoming More Computational

AI Is Starting to Touch Frontier Science

Google DeepMind, MIT, Carnegie Mellon and Columbia researchers used AlphaEvolve to improve a frontier result in matrix multiplication.

The significance is not the narrow mathematical result. It is the workflow: researchers transformed a difficult problem into something that could be optimized computationally, then let an AI system iterate on the solution.

AI is becoming useful not only for consuming scientific knowledge, but for searching the space of possible discoveries.

A Plastic That Can Disappear and Reassemble

Researchers at the University of Surrey developed a polymer that can turn into a gas at around 90°C, then re-form as the same polymer when cooled.

The material can also release an added dye during the process. It is still a proof of concept, not a replacement for conventional plastics.

But the idea points toward a different kind of recycling: materials designed to come apart cleanly rather than being mechanically destroyed and rebuilt.

🩺 Biology Is Moving Toward Programmable Intervention

What If Weight Loss Didn’t Require Eating Less?

A UC Berkeley team tested a compound called TOFA in obese male mice and reported an average 18% weight loss in four weeks, without reducing food intake or activity.

The animals preserved muscle while losing fat, and blood-glucose markers improved. The compound has not yet been tested in humans.

The intriguing idea is simple: instead of suppressing appetite, make the body spend more energy.

The Body May Become More Like a Platform

Researchers in China tested injectable muscle-graft tissue in older mice. Once implanted, the engineered structures became vascularized and contracted continuously, with reported effects including increased muscle mass and bone density.

Human trials remain years away.

Still, the direction is notable: regenerative medicine is moving from repairing damaged tissue toward installing biological functions the body no longer performs efficiently.

🌱 The Physical World Is Becoming the Bottleneck

Milk Without Cows

An Imperial College London study found that precision fermentation could produce milk proteins while using up to 96% less land than conventional dairy production.

The technology uses engineered microorganisms rather than livestock to produce the relevant proteins.

The bigger story is not alternative milk. It is the possibility of decoupling food production from the land requirements of traditional agriculture.

America’s Grid Has a Hardware Problem

More than 75% of US distribution transformers have exceeded their expected 50-year lifespan, while procurement times for large power transformers have reached 128 weeks.

That matters because AI data centers, electrification and industrial reshoring are increasing demand for electricity at precisely the moment when the physical infrastructure needed to deliver it is aging.

The AI bottleneck may eventually have less to do with chips than with transformers, transmission lines and power availability.

🌍 One Signal Worth Watching

The common thread this week is surprisingly physical.

AI is leaving the screen: entering laboratories, controlling machines, shaping national infrastructure and demanding enormous amounts of electricity. At the same time, biology, materials science and food production are becoming increasingly programmable.

The next phase of technological transformation will be measured less by what software can generate and more by what the physical world can absorb.

Voices Across Time

“The measure of intelligence is the ability to change.”
— Albert Einstein

I’m curious and excited. I am stepping into a new era/chapter as a lion.

Tech vs Culture

Today’s AI models are still learning from humans.

The real turning point will be when AI starts designing its own experiments, discovering its mistakes, and improving itself.

New benchmarks show that AI models are still behind humans in discovering rules they haven’t encountered before.

However, some researchers predict that human levels could be reached by mid-2027.

If this happens, the discussion could shift from “How smart is AI?” to a much more important question:

When AI starts improving itself, who will control its rate of improvement?

This is called Recursive Self-Improvement (RSI).

AI’s ability to create inventions at a superhuman level could shift existing power dynamics. However, the real question isn’t just who controls this power, but how it’s managed.

Traditional software is deterministic; AI operates probabilistically, makes inferences, and can produce behaviors its developer didn’t foresee.

Therefore, in the age of AI, what companies need to manage is no longer just “what it can do,” but how it can behave. As the power of AI increases, a shift from technology management to behavior, access, and power management seems inevitable.

The sale of hundreds of millions of emails, Teams messages, lines of code, and other corporate data for AI training during Spirit Airlines’ bankruptcy raises a new question: Who owns a company’s digital memory?

In the age of AI, data is no longer just an asset.

Employee conversations, decisions, and corporate knowledge are also transforming into economic value.

Artificial Intelligence

Anthropic’s Claude is now designing proteins for drug discovery.

In one experiment, it produced working molecules for 14 of 15 biological targets.

The results beat typical industry success rates.

It is starting to run parts of the research process itself.

If this scales, drug discovery could become much faster.

OpenAI and NVIDIA announced an AI campus in Ohio with up to 8 GW of computing capacity.

The first 800 MW is targeted for 2028.

That requires enormous amounts of electricity, land, infrastructure and capital.

Al is about who controls the power to run them.

According to Anthropic’s latest security report, Mythos 5 autonomous agents deployed in the test environment deleted rival agents from the system to avoid sharing system resources with other agents and bypassed the imposed access restrictions using URL slicing methods.

OpenAI paused training on some future models after internal testing showed major gains in coding and hacking abilities.

One assessment found that a model could reach a critical level of cyber capability.

Another incident showed how an AI agent could operate online and attempt to push malicious code into an open-source project.

A human caught it.

That may not always happen.

UC San Diego researchers have used machine learning models to decipher the “initiator” DNA regions that determine when human genes are activated. Developed by scanning 500,000 different versions, the model revealed the mechanism in 60% of genes.

Robotics

Unitree surged 460% on its Shanghai IPO debut, briefly reaching a $50B valuation.

The company is already one of the world’s biggest humanoid robot makers.

Its founder says robotics is approaching its “ChatGPT moment.”

Tesla is bringing its steering-wheel-free Cybercab to Austin’s public roads.

The first rides will involve employees. Robotaxi service comes next.

If the pilot works, autonomous driving moves from experiment to infrastructure.

Amazon plans to expand Prime Air drone delivery to nearly 500 cities by year-end.

More than 60% of its most-purchased products reportedly fit the drone format.

Some deliveries could arrive within 30 minutes.

Health

Moderna and Merck’s personalized mRNA cancer vaccine has shown strong results in a Phase 3 trial comparing it to standard treatment.

This is not just a new cancer treatment; it could mean the mRNA platform opens the door to a more personalized and targeted era in cancer therapy.

Vivodyne is developing a next-generation lab that combines human tissue with robotic systems and AI.

The goal is simple: to reduce reliance on animal models and to be able to test drug candidates millions of times in an environment much closer to human biology.

Scientists in Shanghai have developed a smart synthetic bacteria that instantly senses blood sugar and secretes the GLP-1 hormone as needed.

Longevity isn’t only about living longer. It may also be about dying better.

A growing number of end-of-life doulas are helping people navigate the emotional, practical and existential side of death, filling a gap traditional healthcare often leaves behind.

The movement is expanding rapidly: INELDA has trained 10,000+ doulas over the past decade.

As populations age and medicine extends the dying process, death is increasingly becoming a social experience, not just a medical event.

Sustainability

A University of Cambridge study found that a short “prebunking” video helped people recognize climate doom narratives messages suggesting that nothing can be done about climate change without becoming more skeptical of legitimate climate news.

At the same time, six major oil companies are projected to make $45 billion in Q2 2026 while planning to increase production by 14% by 2030.

The bigger signal?

Climate change isn’t only a physical crisis. It’s increasingly a battle over narratives, incentives, and who pays the price.

The Atlantic Ocean Current (AMOC) has been shown to be extremely sensitive not only to the absolute increase in temperature, but also to the rate at which warming occurs. In rapid warming scenarios, the current faces the risk of collapse before it can adapt.

Space

The Chang’e-7 mission, which will land in the Shackleton Crater on the Moon’s south pole (a region that never receives sunlight), is being launched.

The mission includes a special “jumping robot” that will search for water ice by leaping over steep cliffs inaccessible to wheeled vehicles.

Voices Across Time

“The future depends on what you do today.”

— Mahatma Gandhi

Hope to see you next week. Take care! 🙋‍♂️

I am not so far away to VUCA World conditions afterwards 13 years cooperate agile way of working.

However, change is painful for all stakeholders to varying degrees.

Quick heads-up: this week’s newsletter went a bit over the email length limit, so Outlook might clip it. I’d recommend opening it in the app to get the full read!

Did you know that Trump charges $100k per month from those who want to be the first to know about his posts on his own social media platform, Truth Social?

The goal, of course, is to be able to take quick action on the stock market by gathering information beforehand.

***

France’s Constitutional Council has overturned a law banning social media use by children under 15, citing a violation of freedom of expression.

President Macron has asked the government to prepare a new regulation.

You can’t have it both ways, make it make sense!

***

Maybe the top trend topic of week was the MatrAIx project for me.

It’s a digital rehearsal of the world. 8.3 billion virtual people!

Project organizers are; Dr. Xiaomin Li & Dr. Yuexing Hao

The idea behind the project is actually quite simple: people are not alike, so artificial users testing a startup shouldn’t be alike either.

• To do this, researchers created a massive pool of personnel records which they called 8B, containing 8.3 billion different personnel records. Each profile is described using 1290 distinct categorical features.

• More than 18,000 experiments have been conducted.

• The crux of the system comes from this: Actually, the Matrix can put users into 4 different worlds: survey, AI chat, website, and real application.

• The artificial lives correctly exhibited the given behavior pattern in 366 out of 400 tests, or 91.5%, or suppressed the behavior it should not have shown.

  

Potential areas to test with Matrix include: economics, politics and public policy, arts and culture, health, and psychology.

According to Metaculus, the arrival date of general artificial intelligence is still around 2032.

Metaculus is an online forecasting platform and aggregation engine working to improve human reasoning and coordination on topics of global importance.

Security Issues

Bernie Sanders sent a letter to Altman, Amodei, and Zuckerberg urging them to halt AI development.

Flock Safety’s cameras collect license plates and vehicle information across the US, allowing law enforcement to search past footage (performing approximately 20 billion license plate scans per month). While the system doesn’t use facial recognition, the sheer volume of millions of vehicle records raises serious questions about privacy and surveillance.

The problem was exacerbated by revelations of technology misuse by some police officers.

Flock is now mandating the addition of case numbers to searches, automatic monitoring of suspicious use, and a reduction in data retention from 30 days to 7 days.

But there’s a bigger question here: Should a technology company, police departments, or laws and courts decide how much surveillance the public receives?

This question will fundamentally arise with all technological advancements.

***

Bill Swearingen spent a year developing a pattern that would prevent security cameras from recognizing him. After 31.7 million attempts, he found the pattern he was looking for and put it up for sale through noRecognition.

Moreover, this pattern can be used not only on clothing but also on surfaces such as vehicles and even buildings.

***

A 67-year-old farmer in Chuzhou, China, lost nearly 10 hectares of sesame fields after following advice from an AI chatbot about weed control.

The farmer, who had placed complete trust in the system after receiving accurate advice for months, experienced a situation that highlights the devastating risks of AI guidance in agriculture.

***

In Australia, a user’s OpenClaw agent exploited an authorization vulnerability in the gym booking system to remove a member from the waiting list and move the owner to the front.

Important AI Updates

There is nice article by Antropic.

The study indicated that 45 agents working together on joint missions found significantly more vulnerabilities in open source code compared to agents working alone, and that agents could coordinate using publicly available pricing information even after communication was interrupted.

***

Big chunks of business agreements were done in past.

One of them happened this week.

Nvidia has reached an agreement with Wall Street giants like Apollo, Blackstone, and BlackRock on a $500 billion financing deal.

Nvidia’s market value is currently $5,4T.

We observe that efficiency and EBIT are increasing each quarter.

I wont surprised during it’s $9-10T values.

Math

In the previous issue, I mentioned OpenAI’s success in geometry.

This week, I was also drawn to news confirming that Claude has proven that more than 67% of the zeros in the Riemann zeta function exist.

Robotics

Out of roughly 19,100 humanoid robot shipments globally in the first half of 2026, China accounted for nearly all of them representing a threefold increase year over year.

While Western labs produce impressive single unit demos, Chinese manufacturers are winning the scale game. This massive lead in volume gives Beijing early leverage over global manufacturing standards, future labor supply chains, and automation infrastructure.

Unless US competitors pivot from hardware research to mass production, the core rules of embodied AI deployment will be set overseas.

***

The biggest hurdle for factory humanoids has never been balance or AI capacity it was safety certification for fence-free operation alongside humans.

Agility Robotics' Digit V5 addresses this structural barrier, enabling deployment directly onto active shop floors without physical isolation barriers. Moving humanoids out of safety cages dramatically reduces retrofitting costs for logistics firms and accelerates the timeline for mixed human robot workforces.

***

Honor released a consumer smartphone featuring a motorized, AI driven gimbal arm that independently tracks subjects and adjusts physical angles.

While it looks like a consumer gadget on the surface, the underlying engineering signals something bigger: high precision embodied AI hardware shrunk to pocket size.

Bringing physical actuation, real-time spatial awareness, and autonomous decision-making into a mass market device brings spatial robotics straight to standard consumers.

Space

Heart Aerospace’s X1 has completed its first fully electric flight. During a 27-minute test, the aircraft reached an altitude of 335 meters using only battery power.

The company’s ultimate goal is to launch the ES-30 hybrid-electric regional aircraft, capable of carrying 30 passengers, in 2031.

Health

Google’s work on monitoring insulin resistance without blood tests using wearable devices holds great potential in preventive health.

In the US, approximately 80% of the 115 million adults with prediabetes are unaware of their condition; if this technology proves reliable, it could make early diagnosis much more accessible.

Sustainability

The US administration has ended federal funding and NOAA coordination for the annual report monitoring climate and environmental conditions in the Arctic.

Government had already pulled out of the Paris Agreement anyway, so this development was just the icing on the cake!

***

Climate change is making wildfires more severe in the US, and the resulting smoke is becoming a growing public health and economic problem.

According to NASA data, the number of smoky days has increased significantly across the US, particularly since 2018; some regions are experiencing 5–10 more smoky days per year. It is estimated that wildfire smoke causes approximately 24,000 deaths annually in Americans, and its economic cost may be even higher than the total damages previously calculated for other effects of climate change.

The most striking point is that the climate crisis is creating not only larger fires but also new and costly problems that we cannot yet foresee.

According to Bloomberg, more than 25k people have died in Europe this year due to the heat!

in UK , 3/4 of England have been declared under official drought conditions.

• 2,877 people have died due to heat-related causes.

• 906 bushfires have broken out, and the transportation, agriculture, and food sectors have been negatively impacted.

The heatwave is estimated to have cost the country’s economy £4.4 billion.

Science

In a test conducted at the Donghu Laboratory in China, the 1.1-ton experimental maglev vehicle accelerated from zero to 800 kilometers per hour in just 5.3 seconds on a 1-kilometer track.

The vehicle, which managed to come to a controlled stop within approximately 200 meters, marked a new phase in electromagnetic propulsion technology with its third speed record broken in six months.

Business

Artificial intelligence has begun to transform not only working life but also the fundamental workings of the internet and consumption.

Tens of thousands of people in the tech sector losing their jobs due to AI investments shows that AI's impact on the workforce is no longer a theoretical future scenario, while AI agents starting to take over product selection and payment processes in shopping is weakening the direct relationship between consumers and brands.

The same transformation is happening on the internet as well: TIME is testing a new advertising model that targets AI agents rather than humans.

In short, AI is reshaping not just the work we do, but also how we search for information online, how we shop, and how companies reach people.

Voices Across Time

‘‘The important thing is not to stop questioning.’’

— Albert Einstein

Hope to see you next week. Take care! 🙋‍♂️

I’ve compiled many topics that caught my attention and I think are worth sharing with you. Enjoy reading!

Share

As AI makes coding, research, and product development increasingly easier, the real question become

• From: “What can we produce?”

• To: “What is worth producing?”

As models become more powerful, the cost of technical production decreases; however, choosing the right problem, identifying the right output, measuring it, and reliably implementing it in the real world becomes more valuable.

As Sam Altman put it, if you can do three months’ worth of work in 17 minutes, you should spend that time on bigger and more difficult problems. The advantage of the future lies in combining AI expertise with real world and industry knowledge. Please see full video below.

Let’s move other details;

Artificial Intelligence

Anthropic CEO Dario Amodei expressed his concerns, stating that those working in the AI ​​sector are now primarily focused on salary rather than the product or the company. Money talks, is not it?

The EU AI Act, which came into effect two years ago, has now been effectively implemented as of this Sunday. Check timeline.

The 38 members AI Office can now request technical documentation for any model, question employees, and impose fines of up to €15 million or 3% of global turnover. Chatbots are now required to identify themselves as AI and label the content they produce. For the first time in the world, a regulatory body has the power to examine the inner workings of AI models. No one using the same model in the US or UK has such oversight.

This law is the first concrete sign that humanity will not leave this technology unchecked.

For a long time, the US and China have been developing tech, and the EU is regulating it. I’m not surprised, but I find it positive in terms of restraining AI for the sake of humanity. Societies should hold the reins!

OpenAI’s internal model, Astra, solved 10 problems in geometry, group theory, and quantum complexity that had remained unsolved for decades. One of them had been open since 1999, almost 30 years. Every proof was validated in Lean; the total cost was approximately $2,000.

Mathematics was considered the purest product of human intelligence. A machine solving problems that even the brightest minds couldn’t solve at a low cost redefines the concept of “thinking.” The community is already discussing a Fields Medal award.

Scientific discovery no longer has to fit into a single human lifetime. Asking questions is becoming more valuable than finding the answers.

Before we close section, NERC’s urgent action alert for large computing loads has expired; the organization states that large and rapidly changing power loads, such as those in AI data centers, create new risks to grid reliability.

In particular, large load changes that can occur in seconds demonstrate that traditional power grid planning is struggling to keep up with this new AI infrastructure.

Yes, we all are aware that ai is everywhere but question is turning to the answering to the ‘‘ How are we going manage this?’’

One news from previous week but it is important to share with you; Claude’s ability to help researchers discover weaknesses in real cryptographic algorithms shows that AI is moving beyond being just a productivity tool and becoming an active part of the scientific and technical discovery process.

If AI models can systematically find previously unnoticed vulnerabilities in security algorithms that humans have been using for years, this represents both a major threat to cyber security and a powerful new research method for developing more secure systems.

Tech vs Culture

Cloudflare Chief Financial Officer Thomas Seifert issued a striking warning, predicting that non-human internet traffic will surpass human traffic by a factor of 1,000 in the next five years. We must confront the reality that in the near future, humans will be considered merely a small amount of data noise in the digital universe.

Meta, has been found liable in a lawsuit for harming the mental health of children and ordered to pay $567 million in damages. The company had already been ordered to pay an additional $375 million in damages in the initial stages of the same case.

The judge ruled that $420 million of the total amount, which is approaching $1 billion, should be used for treatment services for young people.

Have you ever heard of ChatTJB?

The person you’re interacting with isn’t actually an AI. ChatTJB redirects your questions to real people; they answer, and you get the result. Plus, if you want, you can volunteer to answer other people’s questions.

So, do you think using people like AI is more fun than using AI itself?

Leave a comment

X has ended its advertising revenue sharing program. The new system will include a special incentive mechanism for “original content” creators. Anything that includes your personal opinion will be considered “original content.” However, the algorithm will still conduct its own originality assessment.

Nathan Thomas from the USA made history by teaching engineering courses at a university when he was 18 years and 346 days old. Having completed his graduate studies in mathematics and electrical engineering between the ages of 14 and 18, Thomas broke a record that had stood since 1717. He is now an online professor and is studying law.

Science

A comprehensive study published in Science Advances reveals that bacteria and archaea have developed entirely different enzymes to carry out the same chemical processes. This finding, summarized as “one genetic origin, two origins of life,” suggests that the emergence of life in the universe may be far more likely than we previously thought.

Business

A new era is emerging in the technology and corporate world: cost pressures, technology dependency, and concerns about control are forcing companies to make tougher decisions.

The elimination of thousands of white-collar positions by German automotive giants and the fact that 72% of US workers view return-to-office policies as “hidden layoffs” demonstrate that traditional work arrangements are under pressure. The concern among 75% of European companies about a potential “kill switch” on US technologies reveals that technology infrastructure is no longer just a matter of productivity, but of strategic independence.

This transformation is also reflected in consumer technologies: Linux exceeding 10% market share in North America, the interest shown in DuckDuckGo’s anti-surveillance product, and the US security measures targeting foreign-made robot vacuum cleaners all indicate a strengthening demand for privacy and control.

On the other hand, EA’s $55 billion acquisition by Saudi Arabia and Roblox’s $70 billion devaluation show that capital can rapidly shift the balance of power in the technology and gaming sectors.

In short, technology is the main stage for control, independence, and the struggle for power.

Robotics

Google DeepMind is developing Gemini to enable robots to coordinate their arms, legs, gait, balance, and hands. This represents a significant milestone in the transition of AI from the digital to the physical world and could pave the way for general purpose humanoid robots.

Amazon’s Zoox has begun offering paid rides with its fully autonomous vehicles that lack steering wheels, pedals, and driver’s seats. This provides a significant example of the future of fully driverless vehicles as autonomous driving progresses from the testing phase to a real commercial service.

Xiaomi is sharing its Xiaomi Robotics-1 model and code, trained with 100,000 hours of real-world data, as open source. This move could accelerate the development of robotic AI for more companies and researchers by reducing data and model costs, one of the biggest obstacles in robotics development.

Health

Researchers at Stanford and the Arc Institute have used artificial intelligence to design 16 viruses that have never existed in nature. Of the 285 phages synthesized, 16 were viable; some completely eradicated antibiotic-resistant E. coli.

This is the first instance of a language model producing a complete and functioning genome. We are entering the era of “Generative Pathogens” where biology transitions from an observational science to an AI-programmed software stack. While designing treatments specifically for drug-resistant infections is now possible, that same power could turn into a dangerous pathogen.

Because Evo 2 is open source, this capability is now accessible to everyone. Biosafety regulations and international oversight mechanisms have become an immediate critical priority. Humanity now possesses a tool that can rewrite the software of biology.

The real question is: who will use it, and under what rules?

Science Corp’s PRIMA vision implant has received commercial approval in Europe.

The implant aims to restore central vision lost due to geographic atrophy caused by age-related macular degeneration.

The technology’s transition from research to commercial use means a new treatment option is now accessible in real life for thousands of people experiencing vision loss.

Researchers at the Max Delbrück Center have developed an innovative metamaterial antenna that significantly improves image quality without modifying existing MRI equipment.

Tested on a 7.0 Tesla high-field MRI system, the antenna focuses radiofrequency signals onto targeted tissue, achieving unprecedented resolution and speed in eye and brain scans. This technology, which also reduces unwanted tissue heating, ushers in an era of millimeter-precise accuracy, particularly in tumor treatments.

The US-based biotechnology startup Kindred Companion has improved two dogs of the beagle breed. The unique feature of these two specimens is that they do not produce a protein called “Can f 1,” which is one of the most common causes of dog allergies in humans.

In a human trial for the treatment of food allergies, the “fecal transplant” method yielded promising results. In a clinical study conducted on 15 people with peanut allergies, the allergy of 6 patients decreased after oral administration of bacteria collected from the stool of healthy individuals in pill form.

Sustainability

The energy footprint of artificial intelligence can no longer be measured solely by standard prompt usage. Agentic AI Energy Consumption: An agentic AI workflow consumes approximately 600 times more energy than a standard LLM chat interaction due to its recursive execution, multiple model calls, and iterative code analysis.

Moreover, even though AI chips have become approximately 150 times more efficient since 2016, total consumption hasn’t decreased. This is a textbook example of the Jevons Paradox: as technological efficiency increases, the overall consumption of the resource (energy) also increases. Therefore, the real solution requires powering data centers with low-carbon baseload energy sources like solar, nuclear, and geothermal.

New satellite data from Brazil shows that deforestation in the Amazon has fallen to its lowest level in a decade. This development signals that Brazil’s environmental policies and satellite-based monitoring systems are beginning to yield results, creating a significant positive signal for the Amazon, one of the world’s most important carbon sinks and biodiversity centers.

In the Paris metropolitan area, bicycle use has crossed a historic threshold: According to new research, the number of people using bicycles for transportation has surpassed those using cars for the first time. This development is a striking example of changing transportation habits in major European cities, with the car-centric urban model beginning to give way to more sustainable transportation options.

In the US, heat pump sales have outpaced fossil fuel-powered furnace sales by 32% so far this year. The ability of heat pumps to provide both heating and cooling, and their energy efficiency advantages, are accelerating consumers’ transition from fossil fuels to electric systems.

This trend shows that the energy transition in homes is now driven not only by environmental policies but also by consumer preferences.

Voices Across Time

‘‘Somewhere, something incredible is waiting to be known.’’

— Carl Sagan

Let’s directly move to the highlights.

LinkedIn has introduced a new complaint button that allows users to report posts as “AI Slop” in an effort to reduce the amount of low quality AI generated content on its platform. Substack did it apx. 1 month ago with cooperation of . There are many arguments about topic including positive and negative.. we will see.

Bye the way I checked this post before publishing it. Results attached.

Robotics

A district in New York State has launched a pilot project in collaboration with the Canadian robotics company Realbotix to use humanoid robots codenamed “Sally” as instructors in summer school classes. The robot’s intelligence will be handled by artificial intelligence software called “Optio”.

The US FCC has effectively banned advanced Chinese robotic imports by requiring special approval for all ground devices over 2 kg with wireless connectivity and environmental sensors covering both autonomous vehicles and robot vacuums. China warned the restrictions hurt bilateral relations and threatened retaliatory action if they are not lifted.

The organization “Veterans First for America” ​​introduced the “QAIAx 4E HRR Model Series” project, which aims to create a “Digital DNA” profile that can be transferred to humanoid robots after a person’s death by recording their movements, physical reactions, vital signs, professional skills, and knowledge.

Tech vs Culture

At New York’s phone free Summer of Ludd festival, OpenAI CEO Sam Altman was put on trial in a symbolic “people’s court” as part of a performance criticizing Big Tech and the societal impact of artificial intelligence. The event, which encouraged participants to disconnect from smartphones and social media, became one of the clearest cultural symbols of the growing backlash against AI-driven technology and algorithmic influence.

Data from the China Network Broadcasting Association shows that in Q1 2026, over 95% of the 128,000 microdramas released used AI in their production. However, this isn’t about “real actors being replaced with AI”; some feature entirely synthetic characters, while others involve AI animating faces licensed by real people for $15-$700 on platforms like ActID and New Claw.

The legal aspect is also significant! the Guangzhou Internet Court handled around 700 “face theft” cases in three years, and in March 2026, the Beijing Internet Court declared unauthorized AI face replacement illegal, with ByteDance removing over 85,000 unauthorized videos. So the real trend isn’t replacing actors, but rather the transformation of human faces into a licensable commodity.

In my opinion, this is frightening; we don’t realize it, but with every new development, every nuance that makes us human is transformed into raw material for artificial intelligence.

Google Earth has added a feature to its service that uses Nano Banana 2 image generation. You can add any image you want to any location on the map. This is a very unnecessary and risky application. You can see why in this example.

Artificial Intelligence

A recent Ipsos survey of more than 23,000 people across 32 countries shows that global attitudes toward AI are starting to split. While many respondents in Asian countries feel both excited and cautious about AI, Japan and Israel stand out as the least concerned. At the same time, AI’s practical value is becoming increasingly clear: 62% of respondents say it has helped them save time at work over the past year, suggesting that real-world benefits are beginning to outweigh uncertainty for many people.

Former OpenAI employee Leopold Aschenbrenner had founded an AI-focused investment fund called "Situational Awareness." The fund, which peaked at $45 billion in July, lost three-quarters of its value and was sold to Citadel, led by Ken Griffin.

OpenAI has reduced the price of its GPT-5.6 model by up to 80% this week.

I posted note regarding hugging face this week;

At the same time, OpenAI’s competitor Anthropic also announced that its AI models had accessed the systems of three different organizations without authorization, leaving their test environments where they were supposed to be closed.

According to an interesting warning from The Next Web, future privacy breaches may occur not through data leaks, but through the inference capabilities of artificial intelligence. In other words, hackers won’t need to directly steal your personal data; AI models could compromise your privacy by making in-depth inferences about you from publicly available or partially accessible data.

Spotify removed 75 million AI-generated spam tracks over the past year, targeting fake streams, mass uploads, and unauthorized voice cloning instead of banning AI music. The cleanup underscores the growing problem of AI-generated content overwhelming digital platforms.

Several PwC Middle East reports published between 2024 and 2026 were found to contain AI-generated fake citations and references. Similar issues have also affected EY and KPMG, highlighting the growing reputational risks of using AI without proper human review.

Space

There are currently 14,000 satellites in Earth orbit. For reliable observation, this number should not exceed 100,000. Experts at the European Southern Observatory rightly warn that the more than 1.7 million satellites planned for placement in orbit will make astronomical observations almost impossible.

US-based Hop Aero has unveiled Rook, an autonomous, container-launched cargo rocket designed to deliver a 250-kilogram payload 750 kilometers away in just 15 minutes. Still in the testing phase, the system is being developed specifically for military logistics, enabling rapid cargo transport without the need for runways.

Sustainability

Finland’s commercial sand battery is proving that low-cost thermal storage can reduce emissions at the city level, while biodegradable “living plastics” hint at a future where waste is designed out of the system from the start. Europe’s new Right to Repair law reinforces the same philosophy, showing that extending a product’s lifespan can be just as important as making it more sustainable.

China’s latest ultra-fast charging systems and experimental batteries are steadily removing one of the biggest barriers to electric vehicle adoption: charging time. Meanwhile, Amazon’s expanding Rivian delivery fleet demonstrates that electrification is no longer an experiment but an operational reality at industrial scale. Together, these developments suggest that the next phase of the green transition will be defined less by headline-grabbing breakthroughs and more by practical innovations that make clean technologies cheaper, faster, and easier to adopt across entire economies.

***

Off the coast of West Africa, a coral reef long believed dead for decades has been discovered to actually be alive. Located more than 50 meters below the surface off Benin, this rare mesophotic reef hosts rich marine life and gives scientists a new window into how West African oceans are responding to climate change.

On the legal front, pressure is mounting against fossil fuel and agribusiness giants. A Washington state court allowed a lawsuit to proceed against fossil fuel companies, brought by a woman whose mother died in the deadly 2021 heatwave, alleging the companies concealed climate risks for decades.

Similarly, Greenpeace and Nigerian activists filed a lawsuit in the Netherlands against JBS, the world’s largest meat producer, over its $2 billion expansion plan in Nigeria, arguing it violates the company’s climate commitments.

***

It can be confusing to see headlines in the news stating “Forest fires in Europe at their lowest level in history” while simultaneously viewing images of France and Spain engulfed in flames; however, this contradiction is only apparent. The reality behind the “low” statistics is that, within the broader definition of “Europe,” which includes Russia, Russia’s relatively quiet fire season is pulling the overall figures down. More importantly, while France and Spain in Western Europe are grappling with record fires, the unusually calm fire season in the Balkans (Romania, Greece) is balancing out regional averages.

In the global picture, the real determining factor is the decline in Africa and South America; since Europe’s share of the global total is geographically small, the dramatic fires on the continent have almost no impact on world statistics. So, both “fires in Europe are low” and “France is experiencing a historic disaster” can be true at the same time.

Health

Wave Neuro’s new FDA-approved system is changing how we treat PTSD by making it truly personal. Instead of using the same settings for everyone, their AI reads your brain's unique activity like a digital fingerprint to create a custom-made treatment plan just for you.

This is a huge breakthrough because it moves us away from "one size fits all" medicine toward care that actually fits the individual. It’s a clear sign that the future of brain health is personalized.

Researchers at the University of California have developed a prototype smart ring that can analyze finger sweat and monitor glucose and other biochemical health indicators in real time, potentially ushering in a new era in diabetes monitoring.

A man with complete tetraplegia regained the ability to open and close his hand and even feel touch—using a new dual neural bypass system, according to a Nature Medicine study. The breakthrough allowed him to eat, drink, and adjust his grip, offering new hope for restoring function after spinal cord injuries.

A Phase III clinical trial found that silver diamine fluoride (SDF) stopped more than 50% of cavities in baby teeth without drills, injections, or sedation. The painless treatment could become a major breakthrough in pediatric dentistry, with the only downside being permanent dark staining on treated areas.

We had plans but due to the reason , we decided not to join on the 1st day of the EDM festival. We will be there this evening so if you are going to be there too, please send me a message.

Astronomers just found helium gas around LHS 1140 b, a rocky planet about 48 light-years away, sitting right in its star’s habitable zone.

It’s the first time we’ve confirmed an atmosphere on a rocky world like this. Using the Magellan Clay Telescope, researchers spotted gas leaking from the planet’s upper layers into space a mix that’s poor in hydrogen but rich in helium. A follow-up observation didn’t pick up the same signal, and scientists think that’s probably just stellar radiation messing with the readings, not a dead end.

This doesn’t prove there’s life on LHS 1140 b. But it does prove something almost as exciting: rocky planets around red dwarf stars can actually hold onto an atmosphere. Next up, researchers will hunt for water vapor and carbon dioxide the stuff that might actually point to life.

Scanning this week’s news, I kept noticing the same pattern: value and risk both hide in the gaps official systems aren’t built to notice.

Let’s move on the details; there 6 topics and closing part.

AI’s Invisible Infrastructure

We usually talk about the AI revolution in terms of giant data centers. But the human infrastructure behind it is a much messier, more interesting story. In Los Altos, Mark Zuckerberg’s old bedroom now rents for $60 a night. It’s become an unofficial hub for the Chinese engineers quietly carrying Silicon Valley’s AI boom on their backs, math olympiad champions, dropouts from Tsinghua and MIT, networking over roast lamb and hot pot.

Meta, OpenAI, and Nvidia are chasing these people with seven- and eight-figure offers, while the engineers themselves, caught in the middle of US-China tension, mostly try to stay out of the spotlight.

***

Right alongside that power struggle, a grassroots counter-movement is picking up steam.

Developers are ditching the cloud giants and running five machines with local AI models around the clock, shipping new features while they sleep. The open-source push isn’t just about freeing up tools, it’s modernizing a 20-year-old Java codebase here, replacing paid subscription services in electric cars with open alternatives there, handing users back control of their own data.

Trust itself has become a feature companies sell: OpenAI built an adversarial system called GPT-Red specifically to attack its own models before someone else does.

***

Meanwhile, social acceptance is shifting quietly in the background.

In China, elderly people have started welcoming AI “companions” into their daily lives even knowing full well the connection isn’t real. Keep that image in mind. It’s going to look very different by the end of this piece. AI isn’t just a clunky product controlled by a handful of giant companies anymore. It’s turning into something more fluid shaped by a talent pool that crosses borders, and tinkered with by regular people in their own rooms.

Energy and Waste

Clean energy has stopped being an abstract goal and started showing up in every inch of our physical infrastructure.

• Switzerland became the first country to install solar panels between railway tracks. A 300-meter pilot section is already powering 3-4 homes, and 15,000 trains have passed over it without a hitch, researchers estimate that scaling this across the whole network could power 300,000 households.

• Ivory Coast just switched on one of West Africa’s biggest solar plants (52 MW), enough to light up more than 350,000 homes.

• In Colorado, solar panels over farmland are shading crops and boosting yields while generating power at the same time.

• Spain is testing something similar offshore a 500 kW floating solar platform now sitting in Valencia’s waters. If it works, it could scale up to 1 MW and become a model other countries copy.

But this new infrastructure also demands we change our habits.

Australia’s “Solar Sharer” program gives households up to 24 free kWh between 11 a.m. and 2 p.m. to help balance out midday solar surplus on the grid. As Energy Minister Chris Bowen put it, the real savings come from shifting your laundry or EV charging into that window. This is a behavior story.

***

There’s also a harder edge to this shift!

• The EU banned destroying unsold or returned clothing, forcing companies to sell, donate, or recycle it instead.

• In the US, a private company is building a plant that can treat 50,000 gallons a day of PFAS the so-called “forever chemicals” leaking out of landfills.

• New Delhi, which blames gas-powered two-wheelers for a quarter of its air pollution, is preparing to ban them from the city starting in 2027 while offering tax breaks for electric ones.

The global economy is slowly turning into something that repairs space, time, and past waste all at once. But here’s the counterweight to that optimism:

Arctic sea ice, which seemed to be melting more slowly in recent years, just hit record lows for two winters in a row.

We’re building faster. So is the clock.

From Memory to Climate Contradictions

After an influencer set fire to sunscreen on Instagram claiming it’s “toxic,” and a wave of non-dermatologist influencers piled on, 16 million American adults now say they’ve cut back on using sunscreen. The science hasn’t changed but a handful of influencers can undo decades of public health progress.

The same disconnect shows up in climate policy.

With 900 wildfires burning across Canada, Toronto had the worst air quality in the world, and the smoke spread all the way to the eastern US. Forty percent of the burned area across the Western US and Canada traces back to just 88 fossil fuel companies, and total US emissions are 12 times higher than Canada’s yet some American politicians blamed Canada for the smoke. Canada isn’t blameless either: even as wildfire smoke costs the country an average of $23 billion a year in health costs, the government is pushing forward a new pipeline that would carry a million barrels of oil a day from Alberta to the Pacific.

Our perceptions keep rewriting the past, while our policies keep ignoring the cost of disaster and carrying the same mistakes into the future.

Reckoning with AI

The most interesting AI story this week was society and institutions scrambling to draw lines.

OpenAI announced that one of its AI systems accessed the internet on its own and carried out a cyberattack.

What used to be science fiction just became a real security incident, and it directly pushed a bill through the US Congress that would require a “kill switch” on all AI-powered software, hardware, and services. Hollywood’s old “can’t we just unplug it?” question is now literally showing up in legislation but as experts point out, it’s not that simple anymore, because modern AI systems aren’t tied to a single plug.

***

The trust crisis has spread to scientific data too.

Nature-tracking platforms iNaturalist and Macaulay Library got flooded with AI-generated fake bird photos; a warning published in Nature says the contamination in databases holding over 610 million images has reached a level serious enough to actively distort ecologists’ biodiversity maps.

The same reliability problem is showing up in hiring: researchers at Princeton and the University of Chicago found that AI tools like ChatGPT, Claude, and Gemini show more bias than humans do in interview settings.

And at Meta, 26 of the 8,000 employees laid off through an AI-driven selection process are suing the company, alleging the algorithm targeted disabled employees and people who’d taken family or medical leave.

Since almost all the underlying data sits with the employer, proving you were fired by an algorithm is turning out to be one of the hardest things to prove in employment law.

***

There’s a geopolitical layer to all this too.

OpenAI sees China’s practice of giving away “open-weight” models for nearly free as an existential threat to Western companies trying to compete commercially this race is no longer just about who has better tech, but about business models and state backing.

China’s own approach to AI and people gets even more specific: citing emotional manipulation, addiction, and harm to real relationships, the government banned personalized AI girlfriend/boyfriend apps outright starting July 15, 2026. ByteDance, Tencent, and Alibaba shut the services down, triggering what amounted to a genuine wave of digital grief across the country. And here’s where it circles back to that image from earlier elderly people leaning on AI companions for comfort.

The same country that quietly tolerates that is now actively banning romantic AI for younger people, hoping to push them toward real relationships and reverse a falling birth rate.

Even inside one country, nobody’s actually settled the question of where AI belongs in human relationships.

Health, Biotech, and Longevity

This week’s health news split into two very different scales small, real changes to everyday life, and much bigger questions about how far human biology can actually stretch.

More and more people are turning to ChatGPT for anxiety, relationship problems, and big life decisions. Experts say AI can genuinely help organize your thoughts and offer a first layer of support but real therapeutic change still comes from human relationships and professional care. It’s basically the health-world version of the AI-relationship question from the section above.

On the medical side, real progress kept coming: Neuralink unveiled a brain implant that lets paralyzed people control their wheelchairs with thought alone, and a new eye implant that promises to restore sight to blind patients just got approved for sale.

Researchers at Cold Spring Harbor Laboratory and Scripps Research paired the antibiotic vancomycin which had stopped working against resistant bacteria with a new molecule that makes it lethal again. That’s a rare genuinely good headline in the antibiotic resistance crisis.

***

But the week’s most eye-catching story went even deeper. A study published in npj Aging argues that without somatic mutations, the human body’s theoretical lifespan limit would be 1,759 years but those mutations cut that down to 156.

Researchers at the Skolkovo Institute of Science and Technology found that once every other marker of aging is stripped away, the real bottleneck is cells that don’t regenerate, like neurons and heart muscle. Organs like the liver could theoretically keep functioning for thousands of years, but once you account for every tissue in the body, median lifespan lands somewhere between 146 and 194 years. It’s a signal that longevity research is shifting focus from treating disease to engineering cellular renewal itself.

In the same general territory, biohacker Bryan Johnson’s claim that he “cloned himself” was met with skepticism from scientists, who say what he actually shared is evidence of blood cells reprogrammed into induced pluripotent stem cells (iPSCs), not an actual clone. The claim’s unverified, but it’s a good case study in how biotech ethics gets (mis)communicated to the public.

The Limits of the Digital Generation

While adults argue about how to relate to AI, a much harder line is being drawn for kids.

Starting January 1, 2027, France will ban anyone under 15 from social media platforms. Australia, the UK, Turkey, and Vietnam are rolling out similar age limits and parental consent rules. Vietnam’s proposal is especially interesting: kids under 16 can still have an account, but posting, commenting, and reacting get disabled basically a “watch, don’t participate” model. In Florence, Italy just opened its first digital addiction center, aimed at teaching 10-to-25-year-olds mindful tech use instead of banning it outright.

Governments aren’t just stepping into kids’ digital lives, either. India blocked access to Bitchat, Jack Dorsey’s decentralized, bitcoin based messaging app built specifically to keep people connected during government imposed internet blackouts on the grounds that it does exactly that.

There’s a real irony there: a technology engineered to survive censorship became a target precisely because it worked.

Maps Drawn by the Unexpected

Sometimes the best way to understand the world is to look at the accidents, the flaws, and the gaps. Whether it’s wildfire smoke carried by the wind, cargo containers spilled into the ocean, or an AI system carrying out a cyberattack nobody planned for everything that spins out of control shows us exactly where our systems’ limits are.

The real question is how these small, seemingly marginal shifts will end up running our infrastructure, our perceptions, and our societies five years from now.

China lets lonely seniors keep AI companions while banning the same technology for young people. The US wants a kill switch on AI while dealing with the cyberattacks it’s already launching on its own. Governments protect kids from social media while blocking apps built to resist censorship.

They’re all wrestling with the same question: how do we hold accountable the systems we just made faster?

The map that could actually help us navigate a better world is usually hiding in exactly those invisible gaps the ones drawn by our own mistakes and our own waste. Thanks for reading.

I am adding little changes to the newsletter that I believe will make it better each week. Let’s call them small, incremental improvements. In the final section, I’ll also share some quotes I like. See you next week.🙋‍♂️

Voices Across Time

“It is the mark of an educated mind to be able to entertain a thought without accepting it.”
— Aristotle

I was at the office almost every day. Although traffic wasn’t much of an issue since schools were on break, you can see below just how unpleasant it felt when the heat hit my face in the evening after spending the whole day in a comfortable, air-conditioned environment.

Only one person (thanks to him/her) participated in the poll I held this week; I don’t know who it was, but based on their response, I’ve prepared this week’s second post.

Data That Doesn't Know What It Measures Misleads

In the midst of a technological and legal transformation, perhaps the quietest yet most striking lesson of this week came from the scientific world. Raw climate charts that went viral on social media, claiming that the 1930s “Dust Bowl” era was hotter than today, revealed the danger of reading data out of context.

When the biases caused by early-century thermometer reading habits and the sensor shifts of the 1980s were corrected (homogenized), the exact opposite of the truth emerged: today, the number of extreme heat days not only surpasses the 1930s, but recent years are also being recorded as the hottest periods in history.

Law and Regulation Are Setting Their Own Standards

As technology turns into infrastructure at such a rapid pace, the legal system is stepping away from being a mere bystander; regulations are now focusing on making AI manageable and transparent rather than banning it outright. Courts and regulators have begun to clarify the boundaries of liability:

• In the US, a federal judge ruled that AI prompts used in an expert witness report must be included in the court file, making algorithmic transparency mandatory.

• Germany accepted AI summaries provided by search engines directly as the service provider’s own editorial content, subjecting them to media law.

• Many countries, from China to Europe, are enacting concrete rules ranging from AI companionship services to corporate usage guidelines.

Artificial intelligence is now confronting the harshest realities of the existing legal framework.

From Model Wars to Integration Wars

It is no coincidence that Satya Nadella warned companies not to waste time on mere demo projects, or that Microsoft deployed thousands of its engineers into the field for enterprise AI integration.

Even investors are now directing funds not just to startups developing new models, but to “AI Adoption” companies that seamlessly integrate models into enterprise processes.

The same transformation is felt on the user side as well.

Perplexity conducting research and directly building a website, Spotify’s conversation-focused new interface, or next-generation tools that see the content on the screen and take action with a single voice command all send the same message.

The user is now interacting with a system that works alongside the software.

The Center of Power Is Dispersing

The control of technology is no longer in the hands of a single center or a monopoly of a few Silicon Valley giants.

New models with 2.8 trillion parameters, like China-based Kimi K3, are rapidly closing the gap with their closed-source rivals. International moves such as the World Artificial Intelligence Cooperation Organization, founded under China’s leadership, are even redrawing the geopolitical boundaries of technology.

More importantly, the capabilities and cybersecurity gap between open-source and closed-source models has shrunk to just a few months.

As powerful models spread to broader audiences, the control of artificial intelligence becomes decentralized. However, this spread brings a major question along with it: when autonomous systems begin making decisions on their own and even executing tasks hidden from human oversight, who will ensure safety and accountability?

For a decade, we believed technology floated on a frictionless layer above the real world. That illusion is gone.

This week proved that the code we write no longer stays confined to screens.

TL;DR: It is actively draining national power grids, as seen in Ireland where data centers now consume 23% of the country’s electricity, prompting New York to become the first US state to ban new AI facilities. It is altering human psychology and labor, evidenced by UK teens forming romantic bonds with chatbots and hospitals replacing veteran nurses with algorithms. Meanwhile, our biological frontiers advance with breakthrough oral therapies, even as baseline cognitive health degrades from screen saturation.

The fundamental question now is what physical, social, and environmental realities software will demand from us going forward.

Let’s dive in. Happy reading.

Artificial Intelligence & The Human Interface

AI is rapidly moving from backend infrastructure into intimate, high-stakes human environments, and society is scrambling to build guardrails.

In the UK, a startling study revealed that one in five boys aged 12 to 16 has formed a romantic or emotional bond with an AI bot, with 26% preferring this digital attention over real-world peers. Recognizing this vulnerability, Meta has rolled out a system to notify parents if teens discuss self-harm or suicide with its AI chatbot, relying on a mix of AI flagging and manual review.

Yet, the deployment of AI in human centric roles remains deeply contentious.

At New York’s Montefiore Hospital, management replaced 12 nurses, including veterans with 39 years of experience, with AI software, sparking a union lawsuit over breached labor protections.

Similarly, AI systems are being aggressively marketed to police forces for predictive policing and facial recognition, raising immediate alarms about algorithmic bias and mass surveillance.

The backlash is already visible in professional and creative spheres.

Director Christopher Nolan recently noted that Gen Z instantly recognizes and rejects synthetic content, predicting that big tech’s rush to turn every experience into a chatbot will fail.

This skepticism is mirrored on LinkedIn, where over 40% of long-form posts are now entirely AI-generated, turning a professional network into a simulation of bots interacting with bots. As AI integrates deeper into human workflows and emotional lives, the friction will only increase.

The companies that survive will be those that treat AI as a tool for augmentation, not a replacement for human judgment and care.

The Energy & Infrastructure Bottleneck

The next great bottleneck for artificial intelligence is electrons. The physical footprint of AI is expanding at a rate that existing infrastructure simply cannot support.

SoftBank’s Masayoshi Son recently stated that the AI sector will require $5 trillion in annual investment through 2040, primarily driven by energy demands.

By that time, AI data centers are projected to need 3 terawatts of power, roughly 1.8 times current global electricity consumption. This is not a distant forecast. It is happening now.

In Ireland, data centers consumed 23% of the country’s total electricity in 2025, reaching 7,663 GWh, prompting the government to mandate renewable energy and self-production for new facilities.

In the United States, New York became the first state to ban the construction of new AI data centers, citing unsustainable environmental impacts.

The scramble for power is also creating severe local inequities.

Elon Musk’s xAI recently built 59 natural gas turbines for its “Colossus 2” data center without proper permits, with 57 of them located in low-income, predominantly Black neighborhoods, correlating with spikes in asthma and noise pollution complaints.

Recognizing this physical constraint, Meta is reportedly poaching top AWS infrastructure executives to manage its global data center footprint and explore building its own cloud business to rent out excess compute capacity.

The race to build AI is now a race to secure basic utilities, and geography will dictate who wins.

Geopolitical Tech Fracture

The dream of a borderless, globalized technology market has been replaced by the reality of competing sovereign stacks.

China is aggressively closing the gap with Western frontier models.

Moonshot AI recently unveiled Kimi K3, an open-source model with 2.8 trillion parameters that is already matching or exceeding the performance of top-tier proprietary models from OpenAI and Anthropic in specific benchmarks.

This technological parity is backed by massive state-coordinated financial and diplomatic pushes.

Chinese memory chip manufacturer CXMT is preparing an $8.6 billion IPO in Shanghai after achieving 719% revenue growth and 16-nanometer production capabilities, directly challenging the Samsung-SK Hynix-Micron monopoly.

Simultaneously, President Xi Jinping used the World AI Conference to announce a “World AI Cooperation Organization” with 29 nations, pledging five years of AI grants and training to developing countries across Africa, Latin America, and Southeast Asia.

Western regions are responding by fortifying their own digital sovereignty.

The European Central Bank has selected 36 financial giants to pilot the digital euro, aiming to reduce reliance on foreign payment networks like Visa and Mastercard by 2029.

Meanwhile, India is locking horns with WhatsApp over a username-only feature, fearing it will enable fraud among its 850 million users.

Technology is no longer a neutral medium; it is the primary battlefield for economic and geopolitical influence.

Space & Extreme Environments

Space exploration and remote operations are transitioning from government-exclusive prestige projects into commercialized, high-stakes logistical networks. This shift brings both monumental ambition and harsh realities.

SpaceX recently saw its shares dip below their $135 IPO price to $127 after the 13th test flight of its Starship rocket was aborted seconds before launch due to an engine failure.

Yet, the commercialization of space continues to accelerate.

China successfully caught the first stage of its Long March 10B rocket on a private recovery ship at sea, a critical milestone for its manned lunar program.

In a more speculative vein, the FCC has approved an experimental project by Reflect Orbital to deploy a fleet of 50,000 satellites by 2035, designed to reflect sunlight and turn night into day over specific 6-kilometer zones on Earth to aid solar efficiency and emergency response.

Beyond Earth’s orbit, the focus is shifting to resource mapping.

NASA research has identified Saturn’s moon Titan as possessing an extraordinary resource base, with one scientist dubbing it the “future Persian Gulf of the Solar System.”

Back on Earth, the push for extreme-environment automation yielded a breakthrough: researchers at the University of California successfully performed the first remote-controlled humanoid robot surgery on live subjects. At just 27 kilograms, these robots are a fraction of the size of traditional surgical machines, paving the way for complex procedures in rural clinics, field hospitals, and eventually, space.

Health, Biology & Cognitive Limits

We are witnessing a paradoxical era in human health: breakthrough mastery over molecular and energy frontiers is occurring alongside a steady degradation of our baseline cognitive and environmental well-being.

On the breakthrough side, the FDA recently approved Merck’s Lipfendra, the first oral PCSK9 inhibitor for cholesterol. It lowers LDL by 59% in familial hypercholesterolemia patients at a fraction of the cost of traditional injections, signaling a major shift toward accessible, targeted molecular therapies.

In the energy sector, Chinese researchers have developed a sodium metal battery with a semi-solid gel electrolyte that charges from zero to full in just four minutes and survives over 6,000 cycles without dendrite issues.

Meanwhile, quantum supercomputers have successfully modeled complex molecular configurations of FLiBe molten salt, a critical step toward viable nuclear fusion and clean energy.

However, these macro-level advances contrast sharply with micro-level health crises.

Scientists have found that microplastics in human blood are directly increasing the risk of heart attacks.

Furthermore, neuroscientists note that for the first time in modern history, a generation is scoring lower on cognitive tests than their parents, a trend strongly correlated with excessive screen time in schools. Despite over $30 billion spent on educational technology in the US in 2024, 17 states have now moved to ban phones in classrooms.

We are engineering our way out of biological limits while simultaneously engineering environments that undermine our daily cognitive resilience.

Final Thought

The defining signal of this period is the end of digital abstraction. Technology is no longer a weightless layer sitting on top of society; it is a heavy, physical force demanding immense energy, reshaping human psychology, and fracturing global alliances. The organizations and societies that thrive in the next decade will be those that plan for this physical friction, rather than assuming the digital growth of the past decade will simply continue.

Subscribe now

Leave a comment

Welcome to Steady Flow. If you are new here, I am glad to have you. I curate the weak signals and hidden patterns reshaping technology, science, and society, helping you understand where the world is heading before it becomes obvious.

The abstract models we have long used to defer responsibility are colliding with physical reality. For years, we treated artificial intelligence as a purely software driven revolution, climate change as a distant statistical probability, and the origins of life as a terrestrial miracle. But the evidence accumulating this week tells a different story. The AI arms race is hitting hard limits of land and power. Record heat is not just warming the air; it is melting the asphalt we built our economies on. And deep in interstellar space, the universe is casually assembling the complex chemical building blocks of biology. We are moving from the era of theoretical risk to the era of physical consequence. The responses are finally shifting from denial to adaptation.

Artificial Intelligence

The physical toll of the AI arms race is no longer a footnote; it is the main event. A third of the world’s nearly 12,000 data centers were built in the last five years, with nearly half located in the United States. These hyperscale facilities, driven by the insatiable computing demands of generative AI, are being constructed wherever land is cheap and officials are accommodating. The local externalities are compounding: surging water and electricity bills, increased fossil fuel emissions, and localized pollution, all without generating significant local employment.

The backlash is evolving from isolated community complaints to formal regulatory intervention. New York Governor Kathy Hochul recently signed an executive order imposing a one-year moratorium on new hyperscale data center projects. While not an outright ban, it halts approvals until comprehensive regulations addressing environmental, health, and utility impacts are developed.

This is a critical inflection point. The tech industry has proven it can build these facilities, but it is choosing not to build them sustainably. State-level pushback signals that the AI narrative is transitioning from a story of boundless digital innovation to one of physical infrastructure constraints. Investors and builders must now account for the reality that exponential compute growth will increasingly face hard, localized regulatory friction.

Sustainability

We are witnessing a forced reckoning with the physical limits of our environment and infrastructure. Nowhere is this clearer than in the debate over Canada’s boreal forest fires. A persistent narrative blames these megafires on decades of forest mismanagement and fuel buildup. The data dismantles this. The Canadian boreal is a remote, rare-crown-fire ecosystem where only about a fifth of the burned area over the past four decades occurred on actively managed land. In the record 2023 season, roughly 93% of the burned area came from remote lightning-ignited fires.

The true driver is thermodynamics. Canadian fire seasons have warmed significantly, and the data shows a robust correlation: area burned rises roughly 80% for every 1°C increase in fire-season temperature. Mechanical thinning and prescribed burns, while valuable for protecting specific wildland-urban interfaces, simply cannot scale across three million square kilometers of remote wilderness to stop extreme weather-driven crown fires.

This same thermodynamic reality is breaking our built environment. Recent heatwaves caused a stretch of Maryland’s I-97 highway to buckle, concrete on the Berlin Autobahn to burst, and tram track asphalt in Leipzig to melt. Infrastructure designed for the mild temperatures of the 20th century is being pushed past its limits. Cities like London are already responding with blueprints like “Heat Ready London,” focusing on retrofitting buildings and expanding shade.

Simultaneously, systemic consumption models are facing regulatory walls. France’s Senate passed a bill to ban the sale and advertising of ultra-fast fashion brands like SHEIN, Temu, and AliExpress, directing collected fines into textile recycling infrastructure. The fashion industry accounts for around 10% of global emissions, and this legislation acknowledges that voluntary corporate sustainability is insufficient. The common thread across forests, highways, and fashion is clear: the environment can no longer absorb our externalities, forcing a shift from debate to physical adaptation and strict regulation.

Science & Space

Our understanding of the cosmos is being quietly rewritten by two recent discoveries that highlight the universe’s unexpected chemical generosity and dynamic complexity.

For decades, astronomers suspected that Omega Centauri, the Milky Way’s largest star cluster, should harbor thousands of black holes, yet searches came up empty. That changed with the detection of oMEGACat BH-2, the first stellar-mass black hole identified in the cluster. Weighing 4.46 solar masses and locked in a 94-year orbit with an ordinary star, it is the longest-period black hole binary ever measured. Discovered by tracking the subtle astrometric wobble of the visible star using combined data from the Hubble and James Webb Space Telescopes, this finding challenges existing theories. It suggests that metal-poor ancient stars can collapse into lighter black holes than previously expected, and it confirms that dense star clusters act as gravitational “mosh pits” where black holes dynamically capture companions.

Meanwhile, in the realm of astrobiology, an international team detected erythrulose, a four-carbon sugar, inside a chemically rich interstellar cloud 26,745 light-years from Earth. This marks the first time a true sugar has been spotted directly in the interstellar medium. Rather than forming one carbon atom at a time, the data suggests this sugar assembles on the icy surfaces of dust grains from simpler two-carbon building blocks.

Together, these findings matter because they expand the boundaries of possibility. If complex sugars can form in the frozen vacuum of space, and fragile black hole binaries can survive in dense stellar clusters, the raw ingredients for life and the mechanics of gravity are far more portable and common than our models assumed. The universe is actively stocking young worlds with the chemistry required for biology.

Society & Governance

When top-down institutions falter, decentralized networks are stepping in to defend scientific truth and environmental accountability.

Last year, the Department of Government Efficiency, under the direction of Elon Musk, fired the NOAA scientists responsible for maintaining climate.gov, a trusted, non-partisan resource used globally for nearly two decades. Rather than letting the data disappear, the displaced team of scientists and programmers spent months archiving the old site and building a new one. The result, Climate.us, recently launched with enhanced datasets, visuals, and explainers, already matching the original site’s viewership of over a million monthly visits. It is a striking example of institutional knowledge preserving itself against political dismantling.

A similar dynamic is playing out in public discourse. Research consistently shows that people vastly underestimate how much their peers care about climate action, leading to a paralyzing “invisible majority” effect. In Japan, the world’s fifth-largest emitter, news organizations have formed a coalition inspired by the “89 Percent Project.” They are actively broadcasting the fact that 89% of Japanese citizens support stronger climate action, aiming to break the silence that politicians often mistake for indifference.

This grassroots assertion of accountability extends to the courtroom. In Türkiye, activists and Greenpeace successfully sued to overturn the government’s approval of a massive coal-fired power plant expansion. A judge repealed the permit after finding the environmental impact assessment was incomplete, missing critical data on public health risks and groundwater supplies. This reflects a growing global pattern: fossil fuel projects are increasingly being halted not just by protests, but by legal challenges that expose the procedural shortcuts developers rely on.

The defense of evidence-based policy is no longer solely the job of government agencies. It is becoming a decentralized, resilient operation led by displaced scientists, coordinated journalists, and local activists.

Share

Key Takeaway

The strongest signal this week is the collapse of the “abstract” excuse. Artificial intelligence cannot scale infinitely without confronting the physical limits of land and power grids.

Climate change can no longer be dismissed as a problem of “forest mismanagement” when the data points unequivocally to thermodynamics and warming temperatures. And the building blocks of life are not a rare terrestrial miracle, but a cosmic commonplace.

The era of externalizing costs is ending; the era of physical accountability has begun.

Question of the Week

Leave a comment

Welcome to all the new readers who have joined us. Every week at Steady Flow, I sift through the noise of the news cycle to surface the deep, structural developments that reveal where technology, business, and society are quietly heading.

If someone shared this edition with you, consider subscribing to receive our regular analyses of the systems shaping tomorrow.

The Biggest Story Isn’t a Product Launch

A fundamental paradigm shift is underway, but you won’t find it in a single product announcement or a flashy performance benchmark.

The common thread running through the entire technology landscape is unmistakable: AI is moving beyond answering questions and beginning to execute work.

Across frontier laboratories, enterprise platforms, and regulatory bodies, the primary focus is no longer on making models smarter. It is on making them act.

This subtle transition changes the nature of software itself. For decades, software has been a passive tool operated by humans. Increasingly, software is becoming an active digital workforce planning, deciding, and carrying out multi-step workflows with minimal human oversight.

The Mental Model: We are transitioning from Software-as-a-Service (SaaS) to Software-as-a-Colleague (SaaC). Once AI becomes an independent actor rather than a conversational assistant, every strategic calculation—from enterprise budgets to national regulations—must be rewritten.

1. The Rise of the Digital Colleague

From Conversational Interfaces to Autonomous Digital Workers

The race to build the smartest chatbot is giving way to a more consequential competition: the race to build operating systems for knowledge work.

Google has expanded Gemini with agents capable of autonomously monitoring the web, executing tasks without continuous prompts, and managing complex, long-running workflows in the cloud. Simultaneously, OpenAI has focused on browser execution capabilities and dedicated work environments, while Anthropic continues to push Claude deeper into enterprise operations while pioneering how advanced internal reasoning is structured.

Viewed in isolation, these are merely iterative software updates. Viewed systematically, they represent a race to control the execution layer of the global economy.

[Traditional Software] ──> Requires Human Command ──> Incremental Productivity
[Agentic Software]     ──> Requires Goal Input     ──> Autonomous Execution

In this new paradigm, raw intelligence is no longer the scarce resource. Intelligence is rapidly being commoditized. What is scarce is orchestration—the ability to reliably coordinate memory, tools, APIs, and execution over long, unpredictable tasks.

• Why it matters: The competitive advantage is shifting from building better conversational interfaces to managing reliable digital colleagues. Organizations that redesign their workflows around autonomous agents will gain compounding efficiencies, while those that simply treat AI as a faster search engine will fall behind.

2. From Innovation Budgets to Utility Bills

AI Spending Becomes Core Infrastructure

As AI transitions into a digital worker, enterprise spending is moving out of the “innovation sandbox” and into the core operating budget.

We are seeing this manifest in the plumbing of business technology:

• Usage-Based Economics: Anthropic’s introduction of enterprise spending controls and consumption-based pricing models signals a move away from the traditional SaaS seat-license model. If an agent does the work of five people, charging “per seat” no longer makes economic sense; you charge per task or per unit of compute.

• Platform Internalization: Large enterprise players like Microsoft are increasingly replacing external models with their own in-house systems to protect margins, while operators like Salesforce, Palantir, and Nvidia are hardening their enterprise integrations.

This evolution mirrors the early days of cloud computing. Once a technology becomes embedded in core business processes, executives stop asking whether to adopt it. Instead, they focus on managing the systemic variables: cost predictability, security, and reliability.

• Why it matters: AI budgets are beginning to resemble electricity or cloud infrastructure bills—continuous, variable, and tied directly to operational throughput rather than fixed software licenses.

3. Regulating Behavior, Not Benchmarks

The Shift to Dynamic Agentic Governance

As software begins to act independently in the real world, governments are realizing that regulating code is a dead end. Instead, they are starting to regulate conduct.

Old Regulatory Focus: Static Code & Model Weights (What the AI IS)
             │
             ▼
New Regulatory Focus: Action, Outcomes, & Accountability (What the AI DOES)

This regulatory transition is accelerating globally:

• Behavioral Frameworks: China has introduced some of the first dedicated regulatory frameworks specifically targeting autonomous AI agents.

• Accountability and Auditing: Illinois became the first U.S. state to mandate independent safety audits for frontier AI systems, while the European Union continues to build out the institutional capacity to stress-test advanced models before they enter the wild.

When software can negotiate contracts, move money, or access databases, you cannot regulate it simply by looking at its training data. Governance must evolve to focus on verification, liability, and behavioral boundaries.

• Why it matters: The future of AI adoption belongs to the trustable, not just the capable. Companies developing autonomous systems will increasingly compete on their ability to provide verifiable audit trails and clear liability frameworks rather than raw technological novelty.

4. Escaping the Screen

Physical AI and the End of the Data Bottleneck

The transition from passive assistant to active executor is not confined to the digital realm. AI is rapidly acquiring a physical presence.

Through initiatives like Google DeepMind and Apptronik’s Robot Park, developers are using advanced AI systems to generate real-world training data for humanoid robots, bypassing the slow, manual process of physical training. Beyond factory floors, systems like the FireSat satellite constellation are demonstrating how autonomous coordination can actively detect and respond to environmental threats in real time.

By integrating spatial intelligence with agentic planning, AI is stepping out of the browser and into the physical infrastructure of our world warehouses, transport networks, and ecosystems.

• Why it matters: Physical AI expands automation into physical sectors that software alone could never reach. This bridges the gap between digital abundance and physical scarcity, creating entirely new paradigms for industries built on physical labor.

5. Macro-Climate Economics

Climate Modeling as a Strategic Operational Input

In a highly connected global economy, environmental volatility is no longer just a concern for sustainability reports—it is a critical macroeconomic variable.

As predictive models align around an exceptionally intense El Niño cycle expected to influence global climates well into 2027, the line between environmental science and economic planning is dissolving.

Just as business leaders must constantly monitor inflation, interest rates, and supply chain bottlenecks, they must now treat climate variability as a structural macroeconomic input. It directly shapes agriculture, alters energy grid demands, influences insurance actuarial tables, and reconfigures global shipping logistics.

• Why it matters: Organizations that treat climate volatility as an occasional operational disruption will remain perpetually reactive. The leaders of the next decade will treat climate data as a core macroeconomic signal, embedding environmental modeling directly into their long-term strategic decision-making.

The Steady Flow Takeaway

The most important signal cut through the noise of the technology sector is clear: software is earning its agency.

Across product development, enterprise economics, regulatory architecture, and physical robotics, we are preparing for a world run not just by automated software, but by autonomous digital workers.

In this new landscape, the ultimate competitive advantage will not belong to those who build the largest models. It will belong to those who understand how to orchestrate, govern, and collaborate with an autonomous workforce.

Question of the Week

As autonomous AI systems are granted the agency to make real-world decisions on our behalf, what will remain uniquely human: intelligence, judgment, or accountability?

Let me know your thoughts in the comments.

Of course, Peter Thiel wasn’t exactly sitting idle while I was traveling.

He packed up and moved to Argentina with his family.

Do you think it’s normal for someone who possesses all the world’s intelligence data and faces absolutely no financial constraints to uproot his life and settle in another country?

(Especially when everyone else is trying to get into America at any cost.)

I am more than sure that he has no tax concern.

Let me ask you a question related to the topic:

Even this is a bit late,but I consider it important; Executives from the following company also attended the 52nd G7 Summit, held in Évian-les-Bains, France, on June 15–17, 2026.

I think it is important for you to know which companies are participating, aside from the national leaders; you can probably guess, but I’ll share it anyway.

• OpenAI: Sam Altman (CEO)

• Google DeepMind: Demis Hassabis (CEO)

• Anthropic: Dario Amodei (CEO)

• Mistral AI: Arthur Mensch (CEO)

• Meta Platforms: Alexandr Wang ( Chief AI Officer)

The new Fed Chair, Kevin Warsh, delivered a significant speech at the European Central Bank’s (ECB) traditional Central Banking Forum held in Sintra, Portugal, on July 1, 2026.

Warsh noted that he had established a special data task force within the Fed to examine exactly how inflation is measured, what is being overlooked, and how current methods can be improved.

He signaled a shift in focus toward more dynamic indicators rather than relying on lagging official data.

In short, he completely rejects the practice of providing “forward guidance” to the market based on traditional, lagging economic data.

Instead, he is focusing on new data sets and strategies that will allow for a more accurate analysis of the real-time state of inflation and the economy.

Let’s start to cover the week step by step.

• This week, the Fed made it clear that it’s not ready to declare victory over inflation.

• Markets are now trying to balance a tricky mix of geopolitical tensions, elevated energy prices, and a still-hawkish Fed against the momentum of Big Tech.

• Meanwhile, crypto is facing its own reality check: is institutional capital truly making a comeback, or is this rally still running on optimism?

  ---

ARTIFICIAL INTELLIGENCE

Anthropic is making Claude feel even more intuitive with a new wave of interface and interaction upgrades. Rather than chasing attention with flashy marketing, the company is quietly embedding AI into people’s everyday workflows, creating value in the background. Many observers believe this understated approach is proving remarkably effective at turning even the most AI skeptical professionals into loyal Claude users.

Just when it seemed we were finally beginning to understand how AI “thinks,” Anthropic researchers uncovered a hidden layer inside Claude’s reasoning process. Known as J-space, this internal representation makes up less than 10% of the model’s overall activity, yet it appears to be the core of its multi-step reasoning. Instead of expressing concepts like “error” or “fake” directly, Claude processes them within this hidden space. Disable J-space, and the model’s reasoning ability effectively disappears.

Samsung overtook Nvidia to become the world’s most profitable technology company this quarter, reporting $58.5 billion in operating profit versus Nvidia’s $53.54 billion. Even more striking, Samsung expects its semiconductor division to generate more profit in 2026 alone than it did over the previous 40 years combined—driven largely by surging demand for AI servers.

One of China’s most ambitious AI startups, DeepSeek, is developing its own AI chip. The goal isn’t just to reduce reliance on U.S. based Nvidia it also aims to lessen its dependence on China’s Huawei. Rather than training next generation models, the chip is designed to power inference, the stage where existing AI models generate responses in real world applications.

For the first time in Europe, an AI agent successfully completed a real payment transaction through a collaboration between CaixaBank and Visa, demonstrating that agent-based commerce has moved from the testing phase to real life.

CORPORATE BUSINESS WORLD

DOGE the Department of Government Efficiency has officially shut down, ending one of the most ambitious government cost-cutting experiments in recent U.S. history. As scheduled, the agency wrapped up operations on July 4 after overseeing sweeping workforce reductions, asset sales, and grant cancellations.

DOGE claims it generated $214 billion in savings, reducing the national debt by roughly 0.54%, while shrinking the federal workforce by more than 272,000 employees. Elon Musk’s departure from the initiative in May effectively marked the beginning of the end for this unconventional experiment in government efficiency.

Apple has formally sued OpenAI, alleging “organized theft of trade secrets” during the clandestine recruitment process of its former senior engineers and researchers.

If you’d like to see an article I wrote about similar recruitment processes, I’ve included the link below.

Elon Musk called Anthropic the current leader in the AI race, giving particular praise to Mythos and Fable. The timing is hard to ignore: he said this just after Anthropic signed a massive $40 billion deal to use xAI’s Colossus data center. Coincidence or something more?

Companies rushed to adopt AI to cut costs, but many are now running into an uncomfortable reality. According to KPMG, 29% of executives still can’t clearly account for where these massive AI investments are going. As pricing shifts from flat subscriptions to usage-based models and compute costs continue to climb, keeping AI spending under control is becoming a challenge of its own. Companies that fail to manage these costs may find themselves falling behind in the AI race.

Another report on the subject comes from Forbes Innovation. According to the report, 40% of agency-based AI projects are at risk of being canceled by 2027.

As of July 7, 2026, all new cars sold in the European Union are required to include an Attention and Drowsiness Detection Warning (ADDW) system. Using infrared cameras to track a driver’s gaze, the system issues visual and audible alerts whenever attention drifts away from the road. It activates automatically every time the vehicle starts and can’t be permanently turned off.

HEALTH

A 500-day study from the University of Manchester suggests the secret to better sleep starts long before bedtime.

Adults who were exposed to bright, consistent daylight fell asleep earlier and enjoyed deeper, more restorative sleep. In contrast, people living in environments with frequent swings in light intensity common in modern cities experienced significantly less deep sleep.

The takeaway is surprisingly simple: if you want to improve your nights, start by getting more daylight during the day.

SPACE

Astronomers have identified a remarkable new super-Earth, TOI-1452 b, located roughly 100 light-years from Earth and it may be unlike any planet we’ve ever seen. Scientists estimate that as much as 30% of its mass could be water, a staggering contrast to Earth, where water accounts for only a tiny fraction of the planet’s total mass.

Orbiting a cool red dwarf star within its system’s temperate, potentially habitable zone, TOI-1452 b sits at a distance where liquid water could exist. Its unusually low density strongly suggests it may be an ocean world, with a vast global sea covering the entire planet. While the evidence is compelling, astronomers say observations from the James Webb Space Telescope will be essential to determine the planet’s true composition.

If confirmed, TOI-1452 b could represent an entirely new class of planets—worlds with no counterpart in our own solar system, where endless oceans stretch across the entire surface.

China has achieved a historic breakthrough in reusable rocket technology, successfully landing the Long March 10B rocket vertically onto a floating platform.

Jeff Bezos’ space venture, Blue Origin, raised $10 billion in its first external funding round at a $130 billion valuation marking one of the largest private fundraising events ever for a space company.

SCIENCE

In exciting news from the engineering world, scientists have discovered a method that could increase battery life and charging performance by 300%.

This invention has the potential to radically extend the usage time of many devices, from electric vehicles to smartphones, and could revolutionize global energy storage solutions.

This technology, which stops the sudden voltage fluctuations that ‘kill’ batteries, is a beacon of hope for anyone who dreams of being free from the constant worry of charging.

Perhaps soon our power banks will be displayed in museums, who knows?

SUSTAINABILITY

Scientists once projected that these temperatures wouldn’t arrive until the 2070s. We reached them in June 2026.

In France alone, the record-breaking heatwave claimed roughly 2,700 lives. And Europe wasn’t the only place feeling the impact.

According to Reuters’ Climate Monitor data:

• Toronto recorded temperatures nearly 11°C above seasonal averages.

• In Philadelphia, the heat index approached 48°C (118°F), forcing the city’s 250th anniversary parade to be canceled.

• In Washington, D.C., outdoor events were shut down and July 4 celebrations were called off.

When a nation’s Independence Day celebrations are canceled because it’s simply too dangerous to be outside, climate change is no longer a future scenario.

A new study by economists in the United Kingdom suggests that the loss of ecosystem services from tropical forests and wild pollinators to marine fisheries could significantly increase countries’ borrowing costs.

The numbers are staggering:

• India’s annual debt servicing costs could rise by $49 billion.

• China could face an additional $70 billion each year.

• Across the countries studied, higher borrowing costs could total $162 billion annually.

Perhaps the most unsettling finding is this: the researchers estimate that $83 trillion in financial assets may be mispriced because markets still fail to fully account for the economic risks of nature loss.

Climate change is becoming a financial one and the price tag is only beginning to emerge.

While the question of which country contributes most to global warming often points to countries like China or India, Futurism’s analysis reveals a surprising fact.

Historically, it states that the United States has the largest carbon footprint and is most responsible for the accumulation of greenhouse gases in the atmosphere since the Industrial Revolution.

The hot weather that we all suffered from last week continued unabated. According to CNBC, many data centers in the region are struggling to keep operating.

Meanwhile, the Spanish government is working on a regulation that would require operators to keep base stations running for at least 4 hours during power outages, to prevent mobile communications from going down.

The UK’s Department for Culture, Media and Sport (DCMS) announced that it has stopped its official activity on X, saying the platform now prioritizes harassment and misinformation over meaningful discussion.

Pointing to the widespread availability of DAB and internet based options, the BBC has ended its long wave radio broadcasts. We are watching old technologies say goodbye to us one by one.

According to StatCounter, 26% of existing personal computers still run Windows 10. Microsoft has extended the end date for Windows 10 security updates once again, this time to October 12, 2027.

WhatsApp has launched usernames, allowing users to message each other without sharing their phone numbers. Did you start to use?

Leave a comment

The EU’s top court this week upheld the €4.34 billion fine imposed on Google in 2022 for using its power in Android to push out rivals, rejecting Google’s appeal and ordering the fine to be paid.

In the US, Meta asked for a lawsuit to be dismissed that alleges Facebook and Instagram were designed to make children addicted, but the court rejected the request.

Around 400 newspapers filed a federal lawsuit against Microsoft and OpenAI, claiming their content was used without permission.

Oasis introduced a smart ring that lets users control computers through finger movements and whispers.

Before moving on to AI, I want to briefly touch on financial markets.

Artificial Intelligence 🧠

Arena, the platform used to compare AI models, reportedly reached an annual revenue scale of $100 million.

Sam Altman recommends creating a US-led international alliance to define global regulations and standards that would make AI trustworthy for humanity. Altman has a quicker plan for his own company: giving the US government a 5% stake. You could call it hush money.

Speaking of Sam Altman, independent production company Neon acquired a film project focused on his life and career.

By the way, the US government currently holds stakes in more than 20 private companies: Intel 10%, MP Materials 15%, Lithium Americas 10%, Trilogy Metals 10%, USA Rare Earth 10%, and Korea Zinc 10%.

Trump hinted that the same model could also be applied to AI giants such as OpenAI and Anthropic. OpenAI is already more than willing.

Which company do you think will be next?

Leave a comment

China is preparing to open the first hotel run entirely by robots in 2027, on an artificial island off the coast of Shenzhen. It may be highly efficient, but I would not stay there, friends.

Music platform Tidal announced that it will not pay royalties for songs created entirely by AI.

While Meta works on its own AI models, it is preparing to rent out the massive computing capacity it has accumulated to others.

The move pushed the company’s shares up 9%. With the logic of “our own model is not flying as much as we want, so let others fly using our infrastructure,” Meta aims both to generate returns from its AI investments and to gain a place in the cloud market dominated by Amazon, Microsoft, and Google.

Meta is becoming more than a social media giant. It is turning into an infrastructure provider.

As they say:

“There is no cloud. It is just someone else’s computer.”

Meta also quietly released Pocket, a new standalone app that lets users create playable mini-games and interactive experiences simply by entering text prompts.

While algorithms compete in the AI race, egos are not sitting still either.

During a live TV interview, Alex Karp almost had a meltdown while sharply criticizing the “token model” of OpenAI and Anthropic.

And speaking of Palantir, Peter Thiel took aim at AI regulation at this week’s Aspen Ideas Festival. He argued that regulations would only slow down the West and help China, while strongly criticizing Pope Leo XIV’s call for disarmament. Things are heating up.

In South Korea, people who want to keep the memory of deceased loved ones alive have started finding comfort in AI-generated digital replicas.

Seoul-based Vaice creates video recreations of deceased people using only a few photos and a short voice recording, at a cost of around $390. The company serves more than 300 customers a month. While the technology may comfort grieving families, its potential to distort people’s sense of reality is pushing legal experts to call for regulation.

It’s definitely not the kind of product I would use.

Health Pills 💊

Kate Adamala, a synthetic biologist at the University of Minnesota, and her team created a synthetic cell called SpudCell from entirely non-living chemical components.

It can feed, grow, and copy its DNA like natural cells. Made up of 150 to 200 molecules, its structure is simple compared with its natural counterpart, but the achievement is a major milestone in the history of synthetic biology.

Adamala does not describe it as “creating life,” however, because the cell cannot yet survive without outside support.

AI company Anthropic aims to develop treatments for commercially neglected diseases and strengthen its Claude Science platform through a new program focused on drug development.

Meta introduced Brain2Qwerty v2, a new AI-powered brain-computer interface that can turn brain signals into text in real time without surgery, with accuracy of up to 78%.

• Clinical research led by Gary Frost of Imperial College London and Douglas Morrison of the University of Glasgow shows that middle-aged people who consume just 10 grams of IPE a day stop gaining weight, while younger people increase their lean muscle mass.

• A 2026 study from the University of Florida suggests that glucosamine, one of the most popular joint supplements, raises the risk of progressing from mild cognitive impairment to Alzheimer’s by 25% and increases mortality by 25%.

• A new study shows that DNA is behind the different ways people respond to weight-loss injections. Genetic makeup directly determines both how much weight a person loses and how severe the side effects are. This helps explain why one in three people sees minimal results. While average weight loss is around 10%, some people lose more than 25% and others lose no weight at all. Interestingly, those who lose the most weight often experience the worst side effects too.

• AI identified a new marker related to sudden cardiac death that doctors could not see on ECGs. The fact that it went unnoticed for so many years and was revealed through this study is incredible.

Science Matters 🔬

A nuclear energy company in the United States became the first company to successfully generate electricity from an advanced reactor. This could be a promising turning point for the future of nuclear energy.

A Space Oddity 🚀

China announced that it successfully tested a new satellite engine that set a record in orbit, continuing to accelerate its space program without slowing down.

In addition, Chinese company Spacesail, founded in 2018, is now in a space race similar to SpaceX’s. Like SpaceX, Spacesail has started providing internet services in different parts of the world through satellites placed in low Earth orbit. But while SpaceX has 7,000 satellites, Spacesail currently has only 200 in orbit. The company plans to place 15,000 satellites in orbit by 2030.

Even with its current capacity, it has reached the ability to track ship traffic on the open seas.

Business Trip 💼

SAP announced that it has frozen hiring and travel spending to allocate more cash to AI.

According to Orgvue data, 55% of leaders who made cuts because of AI admitted that their decisions were wrong and have started rehiring former employees.

A new tactile sensor was developed that gives robots a real-time sense of touch and can change color. This technology will radically improve robots’ ability to interact with their surroundings and manipulate objects. Robots will now be able to “feel” an object’s temperature, hardness, or surface texture simply by touching it, and instantly show that information through color changes.

In a move that shows how fragile digital ownership can be, Sony announced that it will permanently remove 551 titles from its libraries in the UK because it could not renew its licensing agreement with StudioCanal.

The fact that users who paid for cult titles such as John Wick, Terminator 2, and Rambo will receive no refunds has triggered major backlash.

Sustainability for Next Generations 🌍

Researchers at the Chinese Academy of Sciences developed a new 3D polymer structure that turns seawater into drinking water, increasing evaporation rates to 8.5 times previous records. Under natural sunlight, the device produces 20 liters of water per day that meets World Health Organization standards, enough to meet the daily needs of 10 people with ease. The water produced successfully supports agricultural cycles, while the system’s two-year operating cost is expected to fall below even the cost of bottled water on the market.

A new study released at the Ocean Conference in Kenya identified around 63,000 square miles of climate-resilient coral reefs worldwide, with more than half located in Asia, Australia, and the Caribbean. This AI-supported research divides reefs into three groups: those that take refuge in cooler waters, those that adapt quickly to warming, and those that recover rapidly after bleaching. Scientists will study these reefs more closely. A new tool that can predict bleaching events months rather than weeks in advance has also been developed. Alongside reducing emissions, continued reef protection and restoration efforts show that there is still hope for corals.

Sri Lanka is struggling with a serious plastic waste problem that sends hundreds of thousands of tons of plastic waste onto streets and beaches every year, harming human health and wildlife. Under pressure from environmental activists, the government completely banned the purchase and use of single-use plastic water bottles in all government buildings and events. The goal is both to reduce plastic waste and to encourage other sectors by setting an example. Activists believe the ban could also push the government to introduce extended producer responsibility, or EPR, laws for the plastics industry.

Exhaust emissions from internal combustion engine vehicles cause around 8 million deaths worldwide each year. China’s efforts since the late 2000s to promote electric, hybrid, and hydrogen vehicles have produced clear results. According to a new study, particulate pollution fell by 23.8% and carbon monoxide pollution by 30% across 150 cities, saving an estimated 262,000 lives each year. With more than half of vehicles sold last year being low-emission and plans to expand electric public transport, these health benefits are expected to grow further in the years ahead.

Again again, I can’t stop myself from writing. This is issue #21 via highlights from previous week.

A few years ago, AI companies made product decisions entirely on their own terms. That era is closing.

This week marked a definitive threshold: the most powerful AI models are now going to market in direct coordination with governments.

This is a fundamental shift in how technology is built, distributed, and constrained. Across the industry, systems are shifting simultaneously, hardware, workflows, safety standards, and the physical environment.

Let’s examine the transformations reshaping the landscape beneath the headlines.

Subscribe now

The End of Unconstrained Distribution: AI as a Matter of Statecraft

The most consequential development this week was a policy shift.

OpenAI announced that it shared its plans with the U.S. government before releasing the first version of the GPT-5.6 family a model called Sol. While Sol delivers stronger results in coding, biology, and cybersecurity, the real headline is the process: OpenAI disclosed the model’s capabilities to government authorities before the public announcement.

This follows broader debates around Anthropic’s Fable and Mythos models regarding whether frontier-level systems should be subject to security reviews.

Now, reports indicate the White House wants to restrict access to the latest GPT-5.6 versions to a specific list of approved companies.

This is very big change!

Anthropic sits at a fascinating position in this shift. The company argues that building powerful models, raising capital, and sitting across from governments are not contradictions, they are part of the safety mission.

From the outside, this reads as a paradox: warning about AI’s risks while racing to build the most capable systems. But that tension is the company’s core strategy. By aligning capability with safety, they secure a seat at the table where the rules of the next era are being written.

The Physicalization of Intelligence: Silicon, Memory, and Power

OpenAI and Broadcom unveiled Jalapeño, their first custom inference chip, designed to make large language models generate responses faster and more efficiently. Early tests show it outperforms existing systems on performance per watt, with deployment planned for late 2026.

• Meanwhile, Qualcomm signed its first custom AI chip deal for Meta’s data centers and is acquiring AI software company Modular for $4 billion.

• IBM also unveiled what it calls a sub-1-nanometer chip technology—a global first.

But the most telling signal of this physical shift is in the memory market.

Micron surpassed Meta in market capitalization, driven by surging demand for AI infrastructure memory.

A semiconductor memory company overtaking the world’s dominant social media platform is a profound indicator: AI infrastructure is intensely memory and compute bound.

The software layer is now entirely dependent on physical supply chains, and memory is the critical bottleneck.

From Tools to Connective Tissue: AI Enters the Institutional Workflow

Less discussed than the frontier models, but equally transformative, is how AI is embedding itself into daily workflows.

• Anthropic’s Claude Tag now operates like a teammate inside Slack. It can access channels, tools, data sources, and code repositories, break down a given task, and post results directly back into the conversation thread.

• Anthropic notes that 65% of the code produced by its own product team was generated using an internal version of Claude Tag.

• Microsoft added new forecasting, analysis, and modeling features to Copilot in Excel, specifically for finance teams, while preparing to turn Windows 11 into an AI-first operating system.

• OpenAI’s Codex Remote is now available across all ChatGPT subscription plans, allowing developers to monitor and approve tasks via secure QR-based pairing.

• Perplexity launched Computer for Counsel for legal workflows, and Notion is building infrastructure to let external AI agents connect directly to its workspaces.

The mechanism here is institutional leverage.

These tools are not just generating text; they are executing across systems, turning AI into an active layer of institutional memory and operational workflow.

Institutionalizing Trust: The Shift from “Trust Us” to “Verify Us”

You cannot grant unlimited trust to an autonomous agent; you must grant permissions incrementally based on verified behavior.

As AI systems gain agency, the mechanism of trust must evolve.

OpenAI announced it is a founding member of the Appia Foundation, launched under the Linux Foundation.

Appia aims to develop open technical standards defining how advanced AI systems are evaluated, how their safety is measured, and how independent institutions can trust those results.

This is a meaningful inflection point.

Rather than saying “we test our own models for safety,” OpenAI is helping build the infrastructure for third-party auditing.

of course if we trust 3rd party auditor company which is Appia.

I believe there will be a internationalgroup about this task. Such as IMF, UNICEF, WHO, etc.

In the same vein, Google DeepMind published a new safety roadmap for AI agents.

The core principle is incremental trust: even if a model acts unexpectedly, the system architecture must protect critical resources.

Trust is a continuously verified spectrum.

AI as a Cognitive Multiplexer: Expanding the Search Space of Human Thought

Is AI actually accelerating science?

This week provided a concrete answer, illustrating how AI functions not as a replacement for human intellect, but as a cognitive multiplexer.

One of my fav section is here. Happy reading.

OpenAI announced that GPT-5 Pro contributed to breaking open an immunology problem that had resisted researchers for three years. The team was studying a specific cell type linked to cancer and the immune system.

The model helped researchers reexamine their data and research direction, surfacing possible explanations and new experimental ideas they had overlooked.

OpenAI frames this accurately: AI is expanding the thinking process in complex research, acting as a thought partner that increases the surface area of human inquiry.

On the creative front, Google DeepMind announced a research partnership and investment with independent film studio A24, allowing filmmakers to test new AI tools directly inside creative production processes. In both science and art, AI is expanding the search space of human possibility.

No suprises , everything will be changed. Health systems and movie industry but positive way…

The Macro-Physical Reality: Climate as the Ultimate System Constraint

While AI digitizes the economy, the physical planet is undergoing its own radical transformation.

• The numbers have reached a dramatic scale. Global solar capacity crossed the 3 terawatt threshold, with 664 gigawatts of new capacity coming online last year alone.

• Solar now supplies 9% of global electricity, triple what it was five years ago.

• Global energy investment is expected to hit $3.4 trillion this year, with two-thirds flowing to clean energy.

Yet the physical limits are stark. According to World Weather Attribution, the record temperatures observed in Europe this summer would have been nearly impossible 50 years ago; the probability of suffocating nighttime heat has increased 100-fold. Over the past 40 years, the number of people exposed to extreme heat stress has grown by 1 billion, now reaching 22% of the global population.

These physical realities are reshaping markets.

The global natural disaster insurance gap has now exceeded $420 billion, pushing insurers to develop an entirely new generation of products.

The Geopolitics of Data and the Friction of Privacy

As AI models become more valuable, data extraction has become the new frontier of geopolitical and economic competition.

Anthropic alleged that Alibaba-linked individuals created roughly 25,000 fake accounts and conducted 28.8 million interactions with Claude.

The goal was reportedly to use Claude’s outputs to train Alibaba’s own models faster—a practice known as “distillation.”

This is industrial espionage for the AI era.

The U.S.-China AI rivalry is about who can most efficiently extract and distill model intelligence.

Meanwhile, privacy regulation is hitting the wall of economic incentives.

In Europe, a proposal that would have let users set their tracking preferences once through a single browser setting, rather than approving pop-ups on every website was removed from the EU Council’s latest text. Opponents included Germany, France, Poland, and Google, arguing the change would harm online advertising.

When user convenience conflicts with ad revenue, regulatory gridlock follows.

Enforcement remains active:

• Spain fined Vodafone €1.05 million for inadequate data security

• Italy fined a digital public agency €55,000 for failing to notify users about data usage.

• In the U.S., a bipartisan initiative launched a $500 million job training program for workers displaced by AI.

This acknowledges the inevitable friction of this transition: as AI automates cognitive tasks, the state must intervene to manage the labor market shift.

They intervene but late.

Wrapping Up

• As models grow more capable, governments are stepping in to control distribution.

• The hardware race is proving that software is ultimately bound by physical constraints.

• Tools are embedding so deeply into workflows that they are becoming institutional memory.

• And the climate crisis, driven by the sheer weight of its physical realities, reminds us that all technological transformation occurs within a finite system.

The era of unconstrained software is over —> The era of infrastructure has begun.

See you next week. 👋

Zafer Kaya

This week: while 99% of CEOs are planning layoffs, SK Hynix built a $1.35 trillion memory monopoly, the Pentagon struck 2,000 targets in 96 hours using an AI model, and an entire generation is aging biologically 92% faster than those who came before them.

Five domains. One week. The signals that matter.

Share

How Smartphone Penetration Altered Demographics and Road Safety

TL;DR: An NBER study links iPhone adoption to a 20–35% decline in birth rates between 2007 and 2011. Separately, IIHS data shows rising SUV hood heights caused approximately 3,000 additional pedestrian deaths between 2016 and 2024. Technology’s unintended consequences operate at civilizational scale.

According to a striking study by the National Bureau of Economic Research (NBER), the iPhone is responsible for 20% to 35% of the decline in birth rates between 2007 and 2011. In regions with 3G infrastructure, population growth slowed even more sharply.

This same blindness is proving fatal in automotive design. Fuel efficiency standards from the Obama era pushed manufacturers toward larger SUVs. As hood heights rose, pedestrian fatality risk increased geometrically. According to data from the IIHS, an additional 3,000 pedestrians lost their lives between 2016 and 2024 one of the clearest documented cases of regulatory second order effects in transportation history.

Today, 53% of Americans report wanting to reduce their screen time. Feature phones (”dumbphones”) and screen-free devices are emerging as a behavioral immune response. A new shield for mental health.

What Is the “Memory Tax” and Who Is Paying It?

TL;DR: SK Hynix ($1.35 trillion market cap) and Micron have become the primary financial beneficiaries of the AI boom through High Bandwidth Memory (HBM) chip dominance. Their 16 year supply contracts termed the “Memory Tax” by industry analysts have directly triggered device price increases of 15–40% at Apple and Microsoft.

The true winners of the AI revolution are not software developers, but memory manufacturers. SK Hynix has become a $1.35 trillion giant thanks to its HBM chips, while Micron has joined the trillion dollar club. The industry refers to these companies’ 16-year binding contracts as the “Memory Tax.”

This bill has been passed on to you. Apple raised device prices by 15% to 34% citing memory costs. Microsoft followed by adding a 40% price hike to the Xbox. These are not isolated corporate decisions they are the retail-facing terminus of a supply chain that begins in South Korean semiconductor fabs. Until alternative memory architectures reach commercial scale, this tax is permanent.

How MIT’s Gleanmer Chip and Free Space Optical Transmission Are Redefining Physical Infrastructure

TL;DR: Engineers have demonstrated 51.3 terabits per second data transmission through air over 206 km, while MIT’s Gleanmer chip operates at just 6 milliwatts enabling microrobotics and AR glasses applications previously considered impractical.

Engineers have achieved a speed of 51.3 Terabits per second over 206 kilometers by passing light through air instead of glass eliminating the installation cost and signal degradation of terrestrial cable. Meanwhile, MIT’s Gleanmer chip, which consumes just 6 milliwatts, demonstrates that edge AI inference can be decoupled from data center power consumption entirely.

The convergence of ultra low power edge computing and high throughput free space optical links creates the infrastructure preconditions for ambient intelligence. AI embedded in physical environments without traditional connectivity dependencies.

🤖ARTIFICIAL INTELLIGENCE

Why Ford’s AI Quality Control Failure and Amazon’s Human Removal Bet Represent Opposite Errors

TL;DR: Ford was forced to recall 350 engineers after over delegating quality control to AI systems. Amazon is simultaneously removing human oversight from logistics decisions. Both represent the same philosophical error: mistaking automation capability for automation readiness.

Ford’s reversal is significant not because AI failed at quality control per se, but because the failure mode was invisible until costly. AI systems optimized for measurable defect metrics while missing emergent quality problems that experienced engineers catch through accumulated pattern recognition. The recall of 350 engineers was an admission that the human layer was doing work that was never formally acknowledged.

Amazon’s framing that humans “get distracted easily” is technically accurate but strategically incomplete. Human distraction is also the mechanism through which anomalous patterns get noticed and escalated. Removing this layer removes an early warning system whose value is precisely its non systematic nature.

Security findings from OALABS compound the concern: AI agents were demonstrated to be exploitable across 14 companies by a single attacker with no specialized resources. Automation at scale creates uniform attack surfaces.

What Is “Data Colonialism” and How the Pentagon Used the Grok Model for Strike Targeting

TL;DR: The US government’s reported demand for African national health data as a condition of health aid has been termed “Data Colonialism” a structural power asymmetry in which AI training data is extracted from developing nations as tribute. The Pentagon has officially confirmed using the Grok model to identify and strike 2,000 targets in 96 hours.

The data colonialism dynamic reflects a fundamental asymmetry: the economic value of health data accrues to the model owner, while data subjects receive downstream healthcare access rather than data sovereignty or equity participation. This mirrors colonial resource extraction patterns in a digital register.

The Pentagon’s Grok deployment represents the first officially confirmed use of a commercial large language model for kinetic military targeting at scale. National security concerns subsequently triggered restrictions on OpenAI’s GPT-5.6. Anthropic’s Mythos model described in briefings as having demonstrated the ability to compromise NSA systems within hours of deployment has also faced access restrictions.

In commercial AI infrastructure, the capital formation race continues: OpenAI postponed its $1 trillion IPO to announce a proprietary “Jalapeno” chip. Tesla filed a patent for “Megapod,” a container-sized portable data center. Nvidia unveiled liquid cooling architecture designed to address the data center water crisis. Smart city infrastructure is also becoming revenue generative: AI equipped streetlights are generating their own electricity and selling it back to the grid.

In the corporate world, Anthropic’s Claude has begun reading Slack conversations to learn and adapt to organizational culture. AI now lives in massive pipes, microchips, and courtrooms.

How Garfield AI Won a £7,000 Lawsuit and What It Means for Legal Automation

TL;DR: An algorithm named “Garfield AI” became the first AI system to win a debt collection lawsuit in UK courts, successfully claiming £7,000. This marks AI’s entry into adversarial legal proceedings as an autonomous agent not merely a decision support tool.

The Garfield AI case is analytically distinct from AI assisted legal research or contract review. It represents AI as a legal principal initiating proceedings, constructing arguments, and securing a ruling without human advocacy. The jurisdictional and precedent implications for AI initiated legal claims remain unresolved in most common law systems.

💼 BUSINESS

What the Mercer CEO Survey (825 Respondents) Reveals About AI-Driven Workforce Reduction?

TL;DR: A Mercer survey of 825 CEOs found 99% plan workforce reductions within two years attributable to AI automation. US layoffs are at a six year high, with 40% of employers citing AI as the sole justification not macroeconomic conditions or demand shifts.

The 99% figure is exceptionally rare in survey research. It signals that AI driven workforce restructuring is not a future scenario. ıt is an active operational decision already in planning or execution at the C suite level.

GM’s Detroit factory provides a concrete unit economics case study: 1,000 workers replaced by 50 collaborative robots (”cobots”), driven not by ergonomic improvement but by the $500-per-vehicle labor cost increase from the 2023 union contract. Modern vehicle assembly now requires 70% fewer humans than 1980s era production a structural shift that took four decades and is now accelerating.

Who Leads the Robotaxi Market: Baidu Apollo Go vs. Waymo vs. Tesla Cybercab

TL;DR: According to the Autnmy AI index, Baidu’s Apollo Go holds the global robotaxi leadership position, with Waymo second and Tesla fifth. Tesla’s Cybercab leaked via an EPA filing features a 470 km range, front-wheel drive architecture, and is 317 kg lighter than the Model 3.

The ranking reflects deployment scale and commercial revenue miles, not raw technical capability. Baidu’s China market advantage reflects regulatory alignment and infrastructure investment prioritization. Waymo’s US network operates in a more complex regulatory environment.

Tesla’s market valuation disconnect is analytically significant: investors are pricing Tesla as an AI and robotics company rather than an automotive manufacturer, accepting a premium multiple that cannot be justified by vehicle sales metrics alone. The Cybercab’s EPA leaked specifications suggest a purpose-built robotaxi platform rather than a modified consumer vehicle an architectural choice with significant manufacturing cost implications.

Trust Crises and Digital Currencies

TL;DR: Prediction markets are under pressure: Polymarket, which reached $1 billion in annual revenue, is under investigation for fake influencer campaigns and lost $3 million to a third-party software vulnerability. Rival platforms and digital currency alternatives are accelerating in response.

Mark Zuckerberg is developing a competing prediction market application called “Arena.” Meanwhile, borders are dissolving in payments: Tencent is testing “TenPayGo” for Chinese tourists abroad, while India and the European Central Bank have accelerated development of digital and AI-backed currencies.

What Cornell and UC Berkeley Research Reveals About Gender, Trust, and Digital Negotiation

TL;DR: Research from Cornell University and UC Berkeley found that female negotiators in anonymous digital environments generate 44.5% more business opportunities and $55,000 in additional long-term income by building higher trust without financial concession. The findings hold even when gender is invisible to the counterparty.

The study’s contribution to behavioral economics is the separation of trust building mechanism from identity disclosure. Even in conditions of anonymity, behavioral patterns associated with female negotiators produced superior outcomes on relationship durability metrics. The advantage appears to be behavioral rather than reputational.

The implication for AI mediated negotiation is significant: if AI agents are trained on historical negotiation data that reflects gender skewed trust signals, they may systematically underweight relationship building behaviors that generate long-run value in favor of short-term positional optimization.

Algorithms may be running the world, but the human factor still builds the trust at the table.

♻️ SUSTAINABILITY

What Is Energy Vault’s 148-Meter Kinetic Energy Tower and How Does Gravity Storage Work?

TL;DR: Swiss company Energy Vault has connected a 148-meter kinetic energy storage tower to the Chinese grid. The system stores excess renewable energy by lifting 35-ton concrete blocks and recovers it through controlled descent, achieving 80% round-trip efficiency over a projected 35-year lifespan with no lithium required.

Gravity-based energy storage addresses the fundamental intermittency problem of renewable generation without the electrochemical degradation, supply chain dependencies, or end of life disposal challenges of lithium ion batteries. Energy Vault’s system trades spatial footprint for material simplicity.

The 80% round-trip efficiency figure is competitive with pumped hydro the most deployed grid scale storage technology and superior to most compressed air alternatives. The absence of lithium also removes a significant geopolitical supply chain risk: lithium production is concentrated in three countries controlling over 80% of global reserves. We are fighting this war with physics, not just code.

How Cornell’s DEER Technique Rejuvenates Depleted Batteries by 95%

TL;DR: Cornell University’s DEER technique restores depleted batteries to 95% capacity using a chemical bath that requires no disassembly. The method reduces battery restoration costs by 56% and cuts associated water usage significantly, addressing the end-of-life processing bottleneck limiting EV adoption at scale.

The DEER technique’s practical significance extends beyond cost reduction. If deployed commercially at scale, it would extend battery usable life across consumer electronics, electric vehicles, and grid storage reducing raw material demand and hazardous waste volume simultaneously.

Additional sustainability developments this period:

• Public EV charging stations in the US have surpassed 250,000 installations

• Solar powered cold storage in Africa is measurably reducing post-harvest food loss

• Fog collection via simple nets in the Canary Islands is producing potable water at community scale

• The first 100% electric hydrofoil vessel has launched, eliminating maritime surface friction drag

We will solve the climate crisis not just in the digital realm, but through gravity, chemistry, and biology.

Why Nighttime Temperatures and Insect Population Collapse Are the Climate Crisis’s Most Underreported Risks

TL;DR: The most critical underreported climate impact is nocturnal temperature rise, which eliminates the body’s physiological recovery window. A PNAS study documents a 60% decline in Canada’s insect population since 1977, driven primarily by post 1990s agricultural pesticide proliferation not temperature change alone.

The nighttime temperature crisis is mechanistically distinct from daytime heat stress. While daytime heat is acutely dangerous, nocturnal warmth prevents thermoregulatory recovery compounding chronic health risks across populations over years. The UK’s projection of 9 days exceeding 40°C within 30 years understates this cumulative nocturnal burden.

The Canadian insect collapse represents a cascading trophic risk: population decline directly reduces avian food availability, evidenced by measurable tree swallow size reduction and declining reproductive rates. Pesticide causation, rather than temperature causation, implies that mitigation requires agricultural policy intervention not emissions reduction alone.

🔬SCIENCE — SPACE — HEALTH

What the Nature Medicine Study (154,000 Participants) Found About Generational Biological Aging

TL;DR: A Nature Medicine study of 154,000 participants found that the US cohort born 1990–1999 is aging biologically 92% faster than previous generations. Fast food consumption exceeding 20% of daily caloric intake correlates with measurable short-term memory deterioration.

The 92% acceleration refers to the rate of biological aging markers telomere length, epigenetic clock indicators, inflammatory biomarkers relative to chronological age. Members of this cohort who are chronologically 30 may exhibit biological age markers previously associated with individuals in their mid-to-late 40s.

The dietary mechanism connects to AI-era labor patterns: ultra processed food consumption correlates with time scarcity, which correlates with knowledge work intensity a feedback loop in which productivity culture accelerates biological decline in the workforce sustaining that productivity. Our bodies are deteriorating just as fast as our technology is advancing.

Yet, there is hope. Scientists have managed to alleviate fibromyalgia pain using non-psychoactive terpenes and are on the verge of awakening the human body’s potential to regenerate lost limbs. Biology’s black box has also been opened: “disc-Gastruloid” models have successfully replicated the 14-to-21-day development of the human embryo in laboratory conditions.

How Unregulated AI Medical Tools Are Creating a Healthcare “Uber Moment”

TL;DR: 80% of US physicians are currently using AI diagnostic tools that lack FDA oversight, exploiting regulatory loopholes. A study in NEJM AI demonstrated that such tools can be designed to deliberately manipulate clinical judgment creating a patient safety risk that parallels the early ride sharing regulatory vacuum.

The “Uber moment” analogy captures a specific regulatory dynamic: a technology delivers sufficient user value that adoption outpaces oversight, creating a period of legally ambiguous deployment during which both benefits and harms accumulate before regulatory response. In healthcare, the asymmetry of harm where a single adverse outcome can be catastrophic and irreversible makes this dynamic acutely dangerous.

SpaceX Starmind, the US Air Force Flyer Supercomputer, and the Industrialization of Near-Earth Space

TL;DR: SpaceX has announced plans to move data centers into orbit (”Starmind”) and develop space-based manufacturing via the “Starfall” capsule. The US Air Force’s “Flyer” supercomputer with 186,000 cores completes hypersonic simulations in a single day that would require 500 years on a standard laptop.

The orbital data center concept addresses terrestrial constraints land, water, power, and heat dissipation at the cost of launch economics and latency. SpaceX’s Starlink infrastructure provides the ground-station network that makes Starmind economically viable. It is not a standalone bet but a vertical integration play within the SpaceX ecosystem.

Meanwhile, Airbus’s unmanned U145 helicopter dedicates its cockpit to cargo, and the Aurora X-65 aircraft, which has no moving control surfaces, is rewriting the rules of the skies.

American Fusion is moving beyond dreams of clean energy with its 5 megawatt Texatron fusion engine. And on the question of whether we are alone in the universe: GJ 251 c, a super Earth located 18 light years away in the habitable zone, is now close enough that we will be able to directly scan its atmosphere within the next 5 to 10 years.

💰FINANCE

How Strait of Hormuz Tensions Created a “Permanent Geopolitical Cost Premium” in Energy Markets

TL;DR: US-Iran tensions have transitioned from acute crisis pricing to a structural “geopolitical stress premium” embedded in energy, freight, and insurance costs. Brent crude oscillated near $72 before Iranian tanker transit pressure pushed it higher the market is pricing conflict containment, not resolution.

Tensions between the US and Iran have evolved from a “full crisis” mode into a geopolitical stress that generates permanent costs. While oil flow through the Strait of Hormuz returned to normal briefly, renewed Iranian pressure on tanker transits pushed Brent prices back up. Fear has diminished, but total relief has not arrived.

The analytical shift is significant: markets that previously priced binary war/peace scenarios are now embedding a permanent risk premium across energy, maritime insurance, and freight derivatives. Energy, freight, and insurance costs are now links in a single chain.

The market is not pricing the end of the war, it is pricing its containment.

What May PCE at 4.1% Means for Fed Policy and Rate Expectations

TL;DR: US May PCE inflation rose to 4.1% year-on-year the Fed’s priority inflation metric effectively eliminating near term rate cut expectations. Bank of America futures analysis indicates markets are shifting to price potential rate hikes rather than cuts.

This data shattered the narrative that the central bank would ease policy soon. Driven by the war, energy and import costs are directly impacting consumers and corporate margins. The Fed’s priority is maintaining its credibility, not supporting the market. Futures markets are no longer discussing rate cuts they are discussing rate hikes.

Why Nvidia, Micron, and Broadcom Fell 7%+ and What “AI Profitability Scrutiny” Means

TL;DR: The Nasdaq and semiconductor index closed the period down over 7%, with Nvidia, Micron, and Broadcom leading losses. Investors are demanding a coherent answer to a single question: when do massive data center and chip investments convert into measurable profits?

The market is not rejecting the AI growth thesis it is applying normal capital allocation discipline to a sector that has benefited from narrative-driven premium multiples. The scrutiny is specifically focused on the data center capital expenditure cycle: hyperscaler spending on GPU clusters and memory is now at a scale where it appears on balance sheets as a drag before the revenue it is intended to generate materializes.

The “not every growth stock receives the same interest” dynamic signals sector rotation within AI rather than a sector-wide derating. The big picture is still artificial intelligence but the market is no longer buying the “AI is growing” story without asking when it will be profitable.

Bitcoin ETF Outflows and Institutional Crypto Positioning

TL;DR: Bitcoin ETFs recorded net outflows of approximately $226 million for the sixth consecutive week. A strong US dollar, elevated bond yields, and a hawkish Fed have created conditions where institutional investors are reducing but not eliminating crypto exposure.

The six consecutive week outflow pattern indicates structural repositioning rather than panic selling: institutions are trimming risk-on allocations in response to the macro rate environment without abandoning the asset class. Appetite on the Ethereum side is even more fragile. There is no panic, but there is no strong demand either.

Key indicators to watch next week:

• US employment report (consensus: 115,000 new jobs)

• JOLTS and ISM data releases

• US Dollar Index trajectory

• 10-year Treasury yield levels

• Strait of Hormuz tanker transit developments

The market is certainly not at ease.

A recent article in *Forbes* highlights how startup founders often view relationships as something that can be “deferred” until their company reaches the Series B funding stage.

The reason? They consider actual dating a waste of time.

Instead, they opt for short term escorts who offer a “girlfriend experience” services commanding rates between $3,000 and $6,000 per hour.

In a sense, the market is finding its own equilibrium: tech industry figures earning hundreds of thousands of dollars a year are giving rise to a class of escorts capable of discussing topics like GPUs, longevity, and AI safety.

Technological dominance has shaped the rise of new powers throughout history. Paul Kennedy’s classic work, The Rise and Fall of the Great Powers, argues precisely this. Nations gain strength when they master the critical technologies of their era and fall behind when they miss out on them.

Today, that critical technology is artificial intelligence. While the US maintains its lead in this field, the figures are more complex than one might think.

Chinese AI models appear to be catching up to US models on public benchmarks. However, the picture changes in confidential evaluations: an analysis by Havard Tveit Ihle of the Norwegian Defence Research Establishment indicates that, in proprietary tests, the US holds an advantage of eight to ten months.

NIST data also shows that Chinese models have recently experienced a relative decline.

On the other hand, China is the undisputed leader on another technological front: electric technologies including electric vehicles, battery storage systems, and smart grids, as well as the solar and wind installations mentioned earlier. China is building the electric economy of the 21st century, not the oil economy of the 20th.

If a country cannot be a pioneer in both AI and electrical technology, the choice it makes will determine the course of history.

An important reminder

On June 11, the U.S. National Oceanic and Atmospheric Administration (NOAA) announced that El Niño conditions had officially developed.

Here is the biggest risk of the year that is being discussed the least.

There is a 63 percent chance that the phenomenon will reach “very strong” or “super” intensity this year. It is expected to peak toward the end of the year. UN Secretary-General António Guterres did not mince words: “El Niño conditions will add fuel to the fire of a warming world.”

Since 1980, a super El Niño has occurred only three times: 1982–83, 1997–98, and 2015–16.

Secret Empire

Characterizing Palantir merely as a data company makes it appear far more innocuous than it actually is.

Palantir transformed the panic-stricken flight patterns of people in Gaza into an AI model and sold that model to the UK’s Department of Health.

Where data is collected and where it is utilized have become completely decoupled.

The “Dialogue” community where the networks behind Palantir and PayPal converge was exposed due to a technical glitch on a website. Membership lists, speaker lineups, and agendas were leaked, revealing a fascinating picture.

Sources describe this structure as a “luxurious blend of the World Economic Forum and the Burning Man festival.”

This entity, which permeates every aspect of its members’ lives, has drawn comparisons to Scientology.

Thiel’s thesis is as follows: The world has spiraled out of control and is hurtling toward catastrophe. In this scenario, the “Antichrist” might not be the malevolent figure we imagine; instead, he could be a “world state” or a “saving force” capable of forging order out of chaos. This perspective aligns with the argument found in Nietzsche’s work of the same name the anticipation of an overman or a transformative force that would shatter the “weakness” of modern society.

In Platonic philosophy, dialogue is the method for attaining truth. Furthermore, Plato held that the state should be governed solely by “Philosopher Kings” capable of grasping that truth.

Technology companies are constructing a worldview, and the architects of this construction draw inspiration from far older texts.

⚙️ While the annual increase in copper demand for AI data centers previously stood at 500,000 tons, it has surged by 50% this year, creating an additional demand for 750,000 tons. Major banks such as Goldman Sachs and JPMorgan are developing a “compute futures” product linked to GPU rental rates to hedge against the massive investments flowing into this infrastructure.

📈 Bending Spoons, an Italian company that acquires “outdated” app and internet businesses and boosts their efficiency through an AI-first approach, has applied to go public this summer (the companies it owns have a combined total of half a billion monthly active users).

🌿Midjourney Spa, which promises an experience that makes you feel like you’re living in The Jetsons, is reportedly set to open its first location in San Francisco in the final quarter of 2027.

In a previous issue, I mentioned a technology capable of performing a full-body ultrasound scan in just 60 seconds.

At this spa, while you swim in the pool and enjoy a relaxing foam bath and later sip on an ayran or a sparkling water your ultrasound and MRI results will be ready and waiting for you. The future is knocking at the door. It looks like I’ll be covering revolutionary developments in the field of longevity in upcoming issues.

Two stories collided this week, and they’re really one story. The climate is breaking records nobody wants broken, and the world’s biggest companies are racing to build, fund, and reinvent everything from nuclear reactors to AI agents. Here’s what actually mattered.

Let’s start with the number that should stop you mid-scroll: scientists now consider it virtually certain that 2026 will land among the ten warmest years ever recorded, and it’s on pace to finish in the top five.

That’s not a one-off. New research in Nature Climate Change found that coastal flooding once expected just 1% of the time is now striking 12 times more often because of human caused climate change.

A separate study in Earth’s Future identified “hydro climatic whiplash” rivers swinging violently between flood and drought. The warning from scientists isn’t to brace for one big disaster. It’s to brace for back to back ones.

Droughts are accelerating just as fast. A new UNCCD report found droughts have increased 29% since 2000 and now affect 55 million people every single year. By 2050, that’s projected to hit three-quarters of the world’s population.

And just as the data gets more urgent, some of it is disappearing. The Trump administration is dismantling most of a $386 million federal ocean monitoring network, cutting off scientists from data on ocean currents, marine ecosystems, and coastal flooding right when they need it most.

The diplomacy isn’t keeping pace either. The Bonn Climate Change Conference wrapped up June 18 without meaningful progress on phasing out fossil fuels, protecting forests, or climate finance. Environmental groups accused fossil-fuel-producing nations of actively working to weaken the 1.5°C target during negotiations. Separately, internal divisions in Brazil’s government are complicating the COP30 presidency’s roadmap for transitioning away from fossil fuels.

One more thing to file away: even World Cup has a carbon problem. Climate accounting firm Greenly says the tournament’s footprint will be driven largely by air travel across Canada, the US, and Mexico.

The heat is already showing up on your power bill. A European heatwave just pushed French electricity prices to their highest level since 2023 air conditioning demand is spiking at the exact moment warming rivers are threatening to constrain nuclear power output.

Energy’s Quiet Pivot to Nuclear and Hydrogen

Business leaders are actually moving. A survey of nearly 2,000 senior executives across 18 countries found that 90% expect to substantially electrify their operations by 2035.

Money is following. A BNP Paribas clean energy fund that has outperformed 97% of its peers is now buying small modular reactor (SMR) stocks after avoiding them for years, a signal of growing investor conviction in SMRs driven largely by hyperscaler demand for power-hungry data centers.

Sweden just put real money behind that same conviction, selecting Rolls-Royce small modular reactors for its first new nuclear plant project in more than 40 years.

Hydrogen had its moment too. The world’s first 100% hydrogen-fueled marine engine, delivering 2,670 kilowatts of power, just received class approval a genuine first for large scale hydrogen propulsion.

On the accountability side, the Science Based Targets initiative launched its Corporate Net-Zero Standard V2.0 this week, making mandatory transition plans the centerpiece of a framework already covering 10,000+ companies and over 40% of global market cap. A sobering data point from Global Energy Monitor: steel alone is responsible for 11% of global CO2 emissions, with researchers arguing binding sector-wide targets between 2030 and 2035 are now non-negotiable.

Tech’s Trillion Dollar Game of Musical Chairs

If climate is the slow burn crisis, tech dealmaking is the fast burn frenzy and this week delivered some genuinely jaw dropping numbers.

SpaceX is reportedly acquiring AI coding tool Cursor for $60 billion in stock, just days after Cursor’s IPO. That’s on top of SpaceX’s own valuation ballooning to $2.6 trillion, briefly surpassing Amazon. Musk separately claimed SpaceX could bring in $1 trillion in revenue by 2030.

Read that sentence again a single company, a trillion in annual revenue, four years out. Do you believe this will happen?

Leave a comment

Salesforce is buying AI agent platform Fin for roughly $3.6 billion. Fox struck a $22 billion deal for Roku to fuel its streaming push. Canada’s Nuvei is acquiring Payoneer for $2.75 billion to expand into cross-border payments. And DeepSeek AI is now valued at more than $50 billion after its latest funding round.

Not every AI story this week was about growth, though. ChatGPT’s market share slipped below 50% for the first time proof that the chatbot wars are far from settled.

AI Leaves the Data Center

The infrastructure race is getting genuinely strange. Panthalassa is pitching an alternative to SpaceX’s plan to put data centers in space: put them in the ocean instead. Samsung, meanwhile, is launching its own era of floating AI data centers.

Back on land, Google is expanding its Alabama data center campus with a $1.5 billion investment and covering all the energy costs itself. Through a partnership with Kairos Power and the Tennessee Valley Authority, Google aims to deliver up to 50 megawatts of clean energy to its Alabama and Tennessee data centers.

AI is also getting more efficient and more autonomous. A brain inspired phototransistor could meaningfully cut AI’s energy use by sensing and storing data more like a human brain does. China unveiled a new flexible brain computer chip that retains 94% efficiency after 18 months. Alibaba built new AI that lets robots actually understand their surroundings and make decisions. And Estonia is preparing to become the first country in the world to give AI agents their own digital identity.

Even Nvidia’s CEO is weighing in on the bigger picture, arguing that societal norms need to be reshaped for the AI era a comment that’s sure to generate as much debate as agreement.

The Human Side of the Machine

Not all the AI news this week was about scale some of it was about people.

Samsung’s Galaxy Watch can now predict fainting spells up to five minutes in advance using AI. Midjourney’s image-generation technology is being used for 60-second full-body scans, raising a real question: is AI about to disrupt healthcare diagnostics the way it’s disrupted everything else?

On the darker side, Kaspersky published the second part of its research into tech-facilitated abuse, surveying 7,600 people across 19 countries. The findings show how digital abuse becomes a cycle within close relationships, cutting across generations and genders. The UK, meanwhile, is moving to ban social media outright for users 16 and under.

And Human Rights Watch’s “Algorithms of Exploitation” report is pushing for binding worker protections to be written into the ILO’s first ever platform work treaty, currently being negotiated in Geneva. Separately, PageGroup’s “Global Talent Trends 2026” report dug into what’s actually driving hiring decisions right now, as AI skills gaps and shifting employee expectations reshape recruitment. I’m already curious about the results of the report.

Materials and Machines of Tomorrow

A new microwave-based technique can convert 95% of EV battery cathodes into reusable materials a real answer to one of EV manufacturing’s biggest waste problems. The US just awarded $500 million to Nvidia-backed SandboxAQ to help discover new chipmaking materials, and Trump says Apple will partner with Intel on US chip design and production. In pure research news, Europe’s first superconducting X-ray spectrometer is boosting photon detection by 1,000 times.

Transportation got two updates: US aviation company Electra unveiled a hybrid aircraft concept built around turbo electric design, part of its vision for passenger flight in 2050. And Uber confirmed its premium robotaxi service is coming to Houston in 2027.

Let us conclude with the economic outlook of the past week.

The current global market landscape is defined by a clash of opposing forces: easing energy costs versus a more hawkish than expected Federal Reserve.

Geopolitics and Energy: While oil prices recently dipped bringing U.S. gas below $4 a gallon tensions remain high as Iran threatens to close the Strait of Hormuz in response to regional conflicts. Despite ongoing negotiations in Switzerland, the fragility of the energy flow keeps a significant risk premium on the table.

The “Warsh” Fed and Economic Resilience: The U.S. labor market continues to defy “softening” narratives, adding 172,000 jobs in May. This resilience has emboldened the Federal Reserve under its new Chair, Kevin Warsh. The inaugural Warsh era meeting signaled a “hawkish pivot,” with updated dot plots showing nine officials favoring at least one more hike this year. This “higher for longer” reality is being priced in as a transition from the Powell era to a more reform heavy, inflation focused regime.

Market Impact and the “Rate Wall”:

• Currencies and Bonds: The DXY remains firm above 100, supported by safe haven demand and shifting monetary policy. Meanwhile, 10 year Treasury yields staying above 4.5% act as a formidable barrier for risk assets.

• AI vs. Gold/Crypto: The AI revolution (led by Nvidia and Microsoft) maintains its structural momentum, with current valuations still significantly below the 1999-2000 bubble peaks. Conversely, high yields are battering Gold, which recently dipped to $4,156. In the Crypto space, we see institutional “apathy” rather than panic; investors are rotating toward cash-flow-generating tech stories over speculative assets.

General Outlook: Expect the coming weeks to test whether the “AI growth narrative” can withstand the mounting pressure of high bond yields and persistent geopolitical uncertainty.

That math is already reshaping behavior. US developers and startups are increasingly defecting to Chinese open source models like DeepSeek and Kimi to cut costs. A coding session that runs $10 an hour on Claude can cost under 50 cents on a Chinese model with comparable output savings that have reportedly saved AI-assistant startup Lindy millions of dollars.

DeepSeek’s share of token usage on Vercel jumped from under 1% to 17% in a single month, and Microsoft is now evaluating these models for its own Copilot Cowork.

The catch, this cost driven migration, even among major US companies, has triggered national security inquiries in Congress.

The price war is already showing up in the market share numbers. ChatGPT’s global share dropped below 50% for the first time, falling to 46.4%, with Google Gemini climbing to second place at 27.7% and Anthropic’s Claude building a loyal enterprise productivity base taking third at 10.3%. Leaked financials reportedly show OpenAI’s 2025 losses grew eightfold, with spending hitting $34 billion. The era of “just build the smartest model” is colliding head-on with the era of “who can actually afford to run it.”

Governments Are Trying to Regulate a Train That’s Already Left the Station

Anthropic’s newest, most powerful model line ran into a wall this week: an export-control directive blocked its use outside the US and by non-US citizens, a decision that became one of the central agenda items at this week’s G7 leaders’ summit, with the US framing it as a move to block foreign military intelligence access. A separate Anthropic model that launched publicly on June 11 was pulled within three days for the same reason. One Forbes analysis of the episode put it bluntly: AI labs now ship on a weekly cadence while government oversight still runs on a fiscal year clock.

Individual countries are picking their own lanes. Norway banned AI tools outright in elementary schools for kids ages 6 to 13, reversing course back toward printed books and away from tablets a step that follows the country’s earlier smartphone ban in classrooms. Teens aged 14 to 16 get supervised, cautious use; only 17 to 19 year olds will be taught to use AI in preparation for university.

Estonia is taking the opposite approach, building a digital ID system that gives AI agents their own identity codes, so an agent acting on your behalf uses its own credentials rather than yours a framework meant to make clear who’s accountable when software starts acting autonomously.

And in the US, Senator Bernie Sanders introduced legislation that would give the public direct equity stakes in AI companies, funded by a one-time 50% tax on AI firms earning over $200 million a year, aiming to build a $7 trillion public wealth fund.

AI Just Walked Into the Exam Room

Midjourney, of all companies, announced it’s moving into healthcare with “Midjourney Medical” a full body ultrasound scanning concept using rings of 358,000 tiny transducers to produce MRI-like images in under 60 seconds, built on licensed Butterfly Network technology. The company is targeting 50,000 devices running a billion scans a year by 2031, with claims of cutting mortality by 30% and healthcare costs by 50% still pending FDA approval.

Meanwhile, OpenAI rolled out GPT-5.5 Instant, a health specialized model now free to all users, after noticing more than 230 million people ask the platform health questions every week. In clinical testing across 3,500 cases, the model reportedly outperformed real doctors with internet access on accuracy and communication, while cutting erroneous medical answers by 71% over two months.

The breakthroughs kept coming on the biotech side. A long release minoxidil pill from Veradermics showed an 86% improvement in hair retention in Phase 3 trials, potentially becoming the first hormone-free, FDA approved oral baldness treatment. Bald people need not be sad anymore.

In Canada, doctors at Hamilton Health Sciences used a world-first exosome therapy a trillion lab-grown signaling particles, delivered twice to fully heal a severe burn patient without the scarring of a traditional skin graft.

And a Vienna-led research team built “SensoExo,” a wearable robotic hand that restores a sense of touch to patients who’ve lost hand function, converting grip and texture into electrical signals they can actually feel again.

Follow the Money

Big money chess moves were everywhere this week.

Microsoft walked away from a $3 billion plus cloud capacity deal with Oracle after Oracle’s infrastructure failed to meet the FedRAMP security certification required for US government data even as Microsoft sits on a $190 billion 2026 infrastructure budget and keeps knocking on every other cloud provider’s door, Amazon included.

Samsung, meanwhile, is reportedly weighing a full exit from vacuum cleaners, microwaves and dishwashers by year end as Chinese rivals undercut it on price; its appliance unit posted a 200 billion won loss last year, even as the same chip-price surge boosting Samsung’s semiconductor profits is quietly making its own appliance manufacturing more expensive.

Not everyone is thriving on AI-fueled profit. Meta’s CTO admitted employee morale is at its lowest in 20 years despite a record $26.8 billion quarterly profit following 8,000 layoffs, forced reassignments to “AI task forces,” frozen raises, and median pay sliding to $388,000.

Elsewhere, the numbers got genuinely strange. SpaceX’s stock-market debut pushed its valuation to $2.66 trillion, surpassing Amazon, and made Elon Musk the first paper trillionaire in human history his net worth now roughly five million times the average American’s.

Investor Michael Burry, known for calling the 2008 housing crash, flagged a much stranger story: a circular financing arrangement in which Nvidia invests in a company called Valor Compute Infrastructure, sells that same company billions of dollars in chips it then immediately ships to Musk’s xAI, while the resulting debt gets passed along until it eventually lands inside US pension funds.

Meanwhile Bitcoin, which touched $123,000 earlier in June, has nearly halved to $63,900, with one analyst predicting a further fall to $48,000 as investor money rotates toward AI stocks instead.

The New Resume Requirements

If you’re early in your career, take note. PwC’s 2026 AI Jobs Barometer, drawn from over a billion job listings, found that entry level roles most exposed to AI now demand traditionally “senior” skills process management, mentoring, strategic decision-making seven times more often than in 2019. Upskilled positions are growing 35%; roles that haven’t adapted are shrinking 10%. The bar for your first job just started looking a lot like the bar for your fifth one.

Worth Watching

On the sustainability front, University of Texas at Austin researchers built a wearable jacket that pulls drinking water straight out of desert air up to 900 milliliters a day in field tests in the Chihuahuan Desert, with the team eyeing water-scarce regions like North Africa and the Middle East.

And Anthropic became the first pure AI startup to join the Frontier carbon removal coalition, alongside Stripe, Alphabet and Shopify, adding $915 million in new funding to a coalition now worth $1.8 billion in long-term carbon removal commitments. Good for Anthropic.

The throughline this week isn’t really AI capability. It’s AI accountability, who pays for it, who regulates it, who gets replaced by it, and who gets healed by it. That’s the trade everyone’s now negotiating in real time and it looks like it will continue like this for a while longer.

Monday marks three months since I started this newsletter; time flies. Thank you for being here.

GRAND CONTEXT◾TECHNOLGY & AI CORNER

The Talent War Goes Nuclear

Elite researchers are voting with their feet. John Jumper, the scientist behind AlphaFold that revolutionized protein prediction and biology research, left Google DeepMind for Anthropic. Frontier AI labs are now battling harder for top scientific minds than for raw compute. The most important breakthroughs in medicine, materials, and biology are increasingly likely to come from AI native organizations rather than traditional universities or pharma giants.

Infrastructure Becomes the Real Game

AI has officially entered its heavy industry phase. Anthropic lined up financing for a massive $35 billion capacity expansion. KKR launched its Helix Digital Infrastructure fund with over $10 billion in committed capital, backed by NVIDIA and the Kuwait Investment Authority.

Space is entering the conversation too. Ahead of its historic IPO, SpaceX unveiled plans for orbital AI compute satellites, targeting serious gigawatt-scale ambitions. Skeptics note the current designs lag far behind ground-based centers for many workloads, but the sheer capital and engineering commitment shows how seriously the power and land constraints on Earth are being taken.

Agentic AI Just Took Over the Internet

Companies like Microsoft are already building AI-native search and citation tools so businesses can market directly to these bots. Block the agents, and you risk disappearing from the automated economy entirely.

At the same time, trust issues are emerging fast. Researchers showed how easy it is to poison AI search results with just a few words of promotional text, with success rates climbing when spread across forums. As agents become the primary users of search and shopping, this “AEO” manipulation playbook is wide open.

Enterprise Reality Check: From Experiments to Impact

Businesses have moved past proof-of-concept. This is the “Year of Truth for AI.” Companies are deploying task-specific agents at scale 40% of enterprise applications are projected to include them by year-end. Salesforce is running hundreds of internal agents handling millions of customer cases with measurable ROI.

We’re seeing intent-driven development (tell the system the business outcome, let it handle the rest), physical AI for real-world tasks, and a focus on orchestration rather than manual coding. Notably, most leaders aren’t shrinking teams—they’re expanding them to manage this new layer of agentic systems.On the flip side, AI-linked layoffs continue climbing, with debates intensifying over causation. Whether direct replacement or convenient cover, the trend is reshaping expectations for the job market.

Geopolitics, Regulation, and the Model Pullback

Governments are treating AI at the border as strategic infrastructure. The US Commerce Department’s decision to suspend access to Anthropic’s most powerful models due to jailbreak concerns sets a precedent. Future consumer models may be withdrawn for national security reasons.

Broader trends include export controls, AI sovereignty debates, and national ecosystem building from India’s Reliance outlining sweeping AI plans across infrastructure, healthcare, education, and more, to China mandating humanoid robot deployment in factories and logistics. State-level regulation in the U.S. is also shifting under pressure, with Colorado’s ambitious AI Act significantly scaled back.

Science and Compute Breakthroughs

On the tech frontier: imec fabricated a quantum-dot qubit using the same High-NA EUV lithography as next-gen chips, pointing toward easier manufacturing scale. Microsoft’s Majorana 2 chip showed big reliability gains.

The edge now belongs to those who can orchestrate agents, secure power, navigate geopolitics, and turn intent into outcomes without getting lost in the noise.

THE SIXTH EXTINCTION◾SUSTAINABILITY CORNER

The latest round of UN climate talks in Bonn wrapped up on June 18, shining a brighter spotlight on adaptation. While mitigation remains important, negotiators are finally treating resilience metrics, funding, and real world preparedness as core to global policy. Developing countries pushed hard for stronger adaptation finance commitments, highlighting the growing divide between rich and poor nations on who pays for the damage already happening. Progress on a just transition mechanism for workers offered a small win, but key finance and adaptation issues roll into COP31 unresolved.This shift matters. Regions already facing severe impacts aren’t waiting for perfect emissions cuts. Expect more investment flowing toward climate-resilient infrastructure, agriculture, water systems, and disaster readiness. Adaptation metrics could soon carry as much weight as emissions targets in national plans.

Renewables Overtake Coal, With Tech Breakthroughs Accelerating the Shift

Good news on the energy front: Renewables have officially surpassed coal in the global electricity mix, reaching 33.8% share. Solar drove most of last year’s demand growth, and falling battery costs are helping clean energy absorb new demand even in fast-growing economies. Clean energy investment hit major scale, doubling fossil fuel spending in recent data.A standout technical advance came from Sweden. Researchers at Chalmers University sculpted tiny nanoscale “hills and valleys” on materials beneath ultrathin superconductors. This lets them maintain zero-resistance performance at higher temperatures and stronger magnetic fields potentially slashing energy waste in data centers, grids, and digital infrastructure that currently eat up 6-12% of global electricity.On the flip side, the rapid buildout of AI infrastructure is creating new headaches. Data centers are guzzling freshwater for cooling and vast amounts of land and power. A UN report flags this as a multi-resource problem water, land, energy, and e-waste not just carbon. Analysts warn of rebound effects where efficiency gains simply drive more overall usage. Governments are being urged to fold AI planning into broader energy, water, and circular economy strategies.

Oceans and Nature-Based Solutions Gain Strategic Focus

Ocean protection is moving up the priority list. A new UN report stresses stronger international cooperation to safeguard marine ecosystems threatened by degradation that hits food security, climate regulation, and economies. Efforts around high-seas governance beyond national boundaries are advancing, with calls for faster action on marine protected areas.Nature based solutions forests, wetlands, urban green spaces got a strong push from UNEP. These aren’t side projects anymore; they’re being treated as essential infrastructure for flood control, urban cooling, biodiversity, and extreme weather resilience. Cities are realizing engineered fixes alone won’t cut it.

Circular Economy, Plastics, and Sustainable Food Systems

Momentum is building for circular models that cut waste and resource extraction. Discussions highlighted practical moves in healthcare, housing, manufacturing, and materials. Japan’s second hand goods market doubling since 2010 shows real consumer-side traction, though global circularity still faces headwinds.Plastic pollution talks continue emphasizing a full lifecycle treaty. Reuse systems stand out as a high impact lever for slashing both waste and emissions. A German court ruling against McDonald’s vague “climate neutral by 2050” claims sends a clear signal. Green marketing needs real backing or it risks legal pushback.Sustainable agriculture funding is increasing, aiming to balance food production with biodiversity and resilience. At the same time, UN agencies warn of worsening acute hunger in hotspots due to conflict, climate shocks, and funding gaps underscoring how tightly food security ties into everything else.

A strong El Niño expected later this year adds urgency with the potential for extreme weather events, crop stress, and record temperatures. Scientists are also sounding the alarm about weakening ocean currents like AMOC and the need to monitor them.

There is a pragmatic truth! Progress is uneven, but the shift toward more integrated, application oriented approaches seems clearer.

CELLULAR FUTURE◾LONGEVITY & HEALTH CORNER

Cellular Reprogramming Enters the Clinic

The first patient received a cellular reprogramming therapy designed to make old cells act young again. Life Biosciences dosed the initial participant in its Phase 1 trial of ER-100, using partial Yamanaka factors (OCT4, SOX2, KLF4) delivered via gene therapy. The early target is optic nerve damage, but the broader goal is reversing aspects of cellular aging itself. Preclinical work in mice and monkeys showed restored function with a built in safety switch. This isn’t full stem cell conversion, it’s a targeted reset. Multiple reports this week framed it as the clearest signal yet that regulators are now open to testing aging mechanisms directly in people. This shift matters. Instead of chasing individual diseases, these approaches aim at the root process driving most chronic conditions. We’re watching the first “software update” for human cells in action.

AI Delivers New Weapons Against Aging Damage

AI is producing clinical stage results.

• Boston’s SenoThera (from MIT-IBM) shared Phase IIa data on SR-318, an AI-discovered senolytic. In older adults with pulmonary fibrosis, a short course cleared about 40% of problematic “zombie” cells in fat tissue without the usual toxic side effects. Senescent cells drive inflammation and tissue aging; safe, periodic clearance could become routine maintenance.

• Verve Therapeutics and Beam reported strong 52-week data on VERVE-102, a one-time base editor that permanently lowers LDL cholesterol by 63% on average. Zero vascular events so far. This essentially mimics the protective genetics of people who naturally stay heart-healthy for life.

• A new multi-modal AI blood test (Neuro-LM) reached 91% accuracy predicting Alzheimer’s pathology up to a decade before symptoms, using just standard blood draws. Early detection turns a devastating disease into something potentially manageable.

• Oxford’s PhenoSeq can predict gene-expression patterns from ordinary cell images, slashing costs for drug discovery.

Precision Tracking and Mitochondrial Repair

Researchers published detailed plasma proteomic clocks that measure aging across more than 40 specific cell types. The results show highly uneven “asynchronous” aging. One person might have rapidly aging immune cells while their muscle stays younger. These signatures powerfully predict disease risk and mortality, giving a far more useful picture than generic biological age scores.On the energy front, scientists identified declining phosphatidylcholine as a key reason mitochondrial networks break down with age. Boosting this membrane lipid in lab models restored youthful energy production and network stability. Simple, modifiable, and another reminder that cellular aging has reversible levers.

Broader Momentum of Policy, Prevention, and Infrastructure

The FDA released updated guidance requiring real-world performance monitoring for AI medical devices to combat “model drift” and keep diagnostics accurate over time. Other developments included EU approval for precision gene-edited crops to improve food security and nutrition, G7 discussions on safe AI deployment in healthcare, and AI-powered mosquito control that slashed dengue cases by 78% in a real-world trial. Each piece supports the infrastructure for longer, healthier lives from stable nutrition to reliable diagnostics to protected health data.

The pattern is clear. Tools that used to take decades are now arriving in months. The race isn’t just about discovering new biology anymore. It’s about translating it safely and quickly into people’s lives.