The astounding growth of the hair-transplant industry in Turkey is not just a medical tourism success story; it’s also a tale of “hacked” medical equipment and algorithmic craftsmanship.

From a biological and evolutionary perspective, human hair is often viewed as an unremarkable mass of keratin that still plays some important functions—protecting our scalps from the sun’s harmful ultraviolet rays and regulating our body temperatures—but, for the most part, is no longer essential to our survival.

Yet, since ancient times, our subconscious perceptions of whether another person is healthy, young, or fertile have been based on visual cues such as skin radiance, the integrity of teeth, and hair density. Deep within our perceptions, hair has become one of the most powerful representations of our identity and self-confidence. It’s key to social communications and perceptions.

Today, the global hair-transplant and restoration industry, which has evolved around this deep psychological and evolutionary need, has grown into a massive, multibillion-dollar industry. Various research firms have estimated the total size of the global hair-transplant market as sitting somewhere between $7.33 billion and $11.61 billion in 2024. And those figures don’t include the underground economy. According to Ministry of Health data, 1.39 million people visited Turkey for medical treatments in 2025. The revenue generated from medical tourism is $3 billion in 2025 (roughly the same as in 2024). While there is no data about how many of these individuals came for hair transplants specifically, it is estimated that one-third of them visited for aesthetic treatments.

The role that hair transplantation plays in promoting Turkey is also noteworthy. For example, Turkish Airlines is occasionally referred to as “Turkish Hair Lines” or simply “Turkish Hair,” a nod to how significant hair transplants are when it comes to tourism to the country. (Similarly, Istanbul Airport has been jokingly referred to as “Istanbul Hairport.”)

It’s possible to see current examples of this in virtually every aspect of popular culture. Last March, a social media user shared a post titled “There won’t be a single bald Spaniard left in the world,” accompanied by an image of the famous soccer player Andrés Iniesta with long hair. It was in response to Spanish prime minister Pedro Sánchez’s stance against the war in Iran, a position that Turkey supports. The post went viral and made headlines on Spanish news channels. Similarly, American basketball star Shaquille O’Neal’s joke in Turkcell’s 5G ads—“I’m here for a hair transplant” while wearing a long curly wig and footage from Turkey’s seven regions—is likely to be talked about for a long time.

Turkey’s global success in hair transplantation and the dominant position the country has achieved are issues too complex to be explained solely by affordable labor, low costs, and favorable exchange rates. Instead it is the result of a bold and at times chaotic yet highly innovative evolution. This includes everything from the adaptation of motors designed for dental devices and sapphire blades used in eye surgery to Anatolia’s ancient craft culture and the master-apprentice relationship transferred to microsurgical techniques.

Makeup for the Modern Man

The development of the institutional infrastructure needed to meet this massive demand in Turkey dates back to the late 1990s. At a time when Turkey’s most famous figures were traveling to Europe for cosmetic surgeries, Dr. Mustafa Tuncer, who attended the Medica trade show in Düsseldorf in 1999, adopted a radical new vision. Tuncer laid the foundation for the Esteworld plastic and aesthetic surgery clinics when he announced, “If Turkey’s celebrities are going to Europe for cosmetic surgery, I will build the best hospital, hire the best doctors, and bring Europeans to Turkey.” Thus, Health Tourism 1.0 began, characterized by fully equipped institutions that combined plastic surgery and hair transplantation under one roof while raising standards to the highest level.

As medical director of the Esteworld Health Group and a member of the second generation of his family to share this vision, Dr. Burak Tuncer says that at the heart of this innovative evolution lies a philosophy with psychological and medical depth—one that does not view the matter merely as a cosmetic procedure. “Hair is a tissue that cannot be replaced or cloned,” he says, adding, “If roots are damaged during the hair-transplant process—whether while being extracted or implanted—we permanently lose that unique tissue. That is why we treat every single strand of hair with the same value and care as we would a kidney or a heart.”

London-based AI lab Inherent emerged from stealth on Wednesday with a $50 million seed round co-led by Index Ventures and Radical Ventures. Nvidia’s venture arm NVentures also participated, alongside Ex/Ante, Metaplanet, Macroscopic Ventures, and Mythos Ventures. It is among Europe’s largest AI stealth-to-launch rounds in 2026.

The founding team comes from DeepMind, Microsoft, and Reka AI. Tantum Collins and Edward Hughes previously collaborated on cooperative AI research at DeepMind. Louis Kirsch, another co-founder, also worked at DeepMind. Kaloyan Aleksiev came from Reka AI and Microsoft.

Collins has a policy background that most AI lab founders lack. He worked on AI policy at the Biden White House before co-founding Inherent. Matt Clifford, co-founder of Entrepreneurs First and the UK government’s former AI tsar, has joined as an adviser.

Inherent is building a platform called Faraday, named after the scientist. Its purpose is not to answer questions faster. It is to figure out which questions are worth asking in the first place.

“Most AI is built to answer questions. What it can’t do yet is figure out which questions are worth asking, the open-ended curiosity that produced penicillin, the microwave, the GPU,” said Danny Rimer, partner at Index Ventures. “That’s the gap Inherent is building into.”

Faraday pairs human researchers with AI agents that are designed to improve themselves iteratively on hard scientific problems. The company describes this as “AI-native science,” a paradigm it says will look and feel different from the scientific method as practised for the past 400 years.

Index Ventures framed the bet in those terms. “AI-native science will be messier, less legible, but capable of exceptional outcomes,” the firm wrote in a blog post announcing the investment. The conviction is that the most valuable application of frontier AI is not automating existing workflows but enabling discoveries that human researchers could not reach alone.

Inherent is structured as a public benefit corporation, a legal form that requires the company to consider its impact on society alongside shareholder returns. The structure is unusual for a venture-backed AI lab. It signals that the founders view governance as a competitive advantage rather than a constraint.

European AI startups are increasingly demonstrating that they can raise at scales previously reserved for Silicon Valley. Inherent’s $50 million seed sits alongside Peec AI’s $10 million ARR in six months, Lovable’s $100 million single-month revenue, and Mistral’s $300 million ARR. The gap between European and American AI funding is narrowing for companies building in categories where the technology is genuinely new.

Anthropic’s Glasswing project demonstrated that frontier AI can find vulnerabilities at a rate that outpaces human remediation. Inherent’s bet is that the same dynamic applies to scientific discovery: AI agents that can explore hypothesis spaces faster than human researchers can, while humans provide the judgment, taste, and ethical guardrails that agents cannot.

The team’s combination of DeepMind research credentials and White House policy experience gives it unusual positioning. It can credibly pitch to both the scientific establishment and the government institutions that fund basic research. Whether Faraday delivers on the promise of AI-native science will take years to evaluate. The $50 million buys the time to find out.

A Google security engineer was charged with insider trading after winning $1.2 million using confidential company data to place bets on the cryptocurrency-based Polymarket decentralized prediction market.

36-year-old Michele Spagnuolo, an Italian citizen residing in Switzerland and a Google employee since 2014, appeared on Wednesday in the Southern District of New York.

In parallel, the Commodity Futures Trading Commission (CFTC) filed a separate civil complaint the same day, seeking restitution, disgorgement, civil monetary penalties, and trading and registration bans.

According to the criminal complaint, throughout this scheme, Spagnuolo used his access to an internal software tool containing confidential “Year in Search” data (Google’s annual ranking of top trending search terms), which was marked with a “Google Confidential” banner in red text.

Beginning in October 2025, Spagnuolo allegedly used a Polymarket account under the alias “AlphaRaccoon” to bet on whether specific individuals would appear on Google’s top trending search lists. He also allegedly used confidential data from Google’s internal data tool and placed bets with near-perfect accuracy across approximately 25 unlikely outcomes, while risking roughly $2.75 million in total.

After Google publicly announced its Year in Search results on December 4, 2025, Spagnuolo’s AlphaRaccoon Polymarket account collected approximately $1.2 million in USDC.e winnings.

“From on or about December 4, 2025 through on or about December 10, 2025, when the Polymarket markets regarding Google’s Year in Search resolved, the software released approximately 3,914,362 million USDC.e to the AlphaRaccoon Polymarket account. On or about December 10, 2025, the AlphaRaccoon Polymarket account sent approximately 5.045 million USDC.e, to Wallet-0xAf6,” the complaint reads.

The FBI traced the AlphaRaccoon account to a payment processor account registered in Spagnuolo’s name and linked to an Italian government identification card. After online communities on Discord and X began speculating that AlphaRaccoon was a Google insider, the username was removed from the account, reverting it to an alphanumeric wallet address.

Prosecutors said that Spagnuolo subsequently moved the illegal proceeds through multiple cryptocurrency-swapping services, including one that removes wallet addresses from the blockchain.

“Today’s charges reinforce a decades-old message: corporate insiders cannot use confidential business information to turn a profit in our markets,” said U.S. Attorney Jay Clayton. “As alleged, Spagnuolo violated the duties he owed to his employer and used Google’s confidential business information to make more than $1.2 million in trading profits on Polymarket.”

“Employees who are entrusted with confidential business information cannot misappropriate that information for personal financial gain,” added CFTC Director of Enforcement David I. Miller.

Spagnuolo now faces a maximum of 10 years in prison on a commodities fraud count and 20 years each on wire fraud and money laundering counts.

Automated pentesting tools deliver real value, but they were built to answer one question: can an attacker move through the network? They were not built to test whether your controls block threats, your detection rules fire, or your cloud configs hold.

This guide covers the 6 surfaces you actually need to validate.

Download Now

“I used to walk into a room and just go ‘ta-da!’ Now, I tiptoe.” This is a line I’ve directly quoted from episode 1 of Russell T Davies‘ blistering new Channel 4 drama Tip Toe, but it’s also the sentiment of any LGBTQIA+ person living in the UK over the last few years.

As Western politics continues to shift to the right, being out and proud is something that’s starting to have a brand-new set of consequences — and instead of shying away from it, Davies presents us with an alarming vision of the future.

South of Tucson, in a stretch of Arizona desert that looks like any other patch of scrub and sun, a plain concrete entrance leads straight down into one of the most complete remnants of the Cold War. This is Complex 571-7, the single surviving Titan II missile site preserved exactly as it stood when the last crews walked out in 1987. Everything else from the original fifty-four sites was destroyed or buried. Here the underground command center and the missile itself remain untouched.

The descent begins at ground level and lowers approximately thirty five feet through a reinforced gateway. We’re talking about gigantic blast doors, each weighing several tons, that can seal the facility in an instant. They were designed to endure a nuclear strike anyplace nearby, including the shock, heat, and fallout. As you pass them, the air changes, as if you’ve stepped into another world. The temperature remains constant throughout the year, and the stillness is so terrible that it feels like it weighs ten times more than a typical basement.

LEGO Architecture New York City – The Big Apple Building Set for Adults, Ages 18+ – Decor for Bedroom…

• THE CITY THAT NEVER SLEEPS – Construct the ultimate souvenir with the LEGO Architecture New York City – The Big Apple (21066) building set for…
• CREATE 6 ICONIC LANDMARKS – Builders can create six NYC architectural icons in minimalist white, including the Brooklyn Bridge, Empire State…
• DISCOVER HIDDEN DETAILS – As they construct their city skyline, builders can discover details that are uniquely New York, such as yellow taxi cabs…

This self-contained three-story steel and concrete rig serves as the command center. Four individuals lived and worked here when there were 24-hour alerts. The dwelling quarters were located on the ground floor. There is a row of bunks for when the crew takes a break between shifts. Next to that is a small kitchen outfitted with metal cupboards and basic appliances similar to those found in an average 1960s American home. A little table and a couch chair complete the setup. Meals were prepared upstairs and then carried down because two guys had to stay on the lower level to keep an eye on things. It was simple and unpretentious, but it made sense for folks who could have been stuck down there for days.

Moving down, the second level housed the operations hub. Consoles and rows of equipment filled the entire area. The officers would monitor missile status, communications, and security from above. Everything was analog, based on what was available at the time, with sequencing technology that employed punch cards and mechanical timers. Every button and indicator has a reason for being present. The area was constructed so that the crew could go through all of the launch procedures without ever needing to surface.

Next, there’s a long tunnel that connects the command center to the missile silo. Even walking it now takes a few minutes. Back then, it must have been a long journey because the tunnel carried power, data, and staff between the two structures. At the further end, it opens into the silo. The Titan II missile, one hundred and three feet tall, remains perched in its launch position. I’m talking huge, since the surrounding concrete walls rise over one hundred forty feet. There are access platforms and equipment distributed around the area. That one missile carried a warhead that was far more devastating than the ones launched on Japan in 1945.

Engineers had designed the entire complex with independent systems so that it could function even if it were cut off from the rest of the world. The complex was self-sufficient due to a diesel generator that provided power throughout, as well as air filtration systems and water storage that could keep the crew going during extended periods of isolation… and some seriously necessary upgrades to protect sensitive equipment, such as shock-absorbing mounts to prevent damage. Then there was the strong reinforced concrete and various blast doors, which formed layers of security to keep the personnel and vital equipment secure from any attacks. The fundamental mission was simple: maintain the crew alive and able to take commands and carry them out regardless of what was going on above.

Life in the underground bunker followed a very tight sched. Crews took turns staying at their positions for alerts, with a couple of people riveted to the consoles while the others rested, ate, or performed necessary repairs. The living quarters, kitchen, bunks, and tiny communal rooms were really merely tools to help the workers prepare to launch at a moment’s notice, much like the launch equipment a whole level below.

The majority of the installations were demolished in accordance with arms-control agreements once the Titan II program ended. However, veterans of the 390th Strategic Missile Wing refused to let the final remaining site be demolished. So they teamed up with the Arizona Aerospace Foundation to keep Complex 571-7 intact. Today, it serves as the Titan Missile Museum. Visitors are taken on guided tours that follow the same route as the original crews: down to the command center, down the tunnel, and alongside the real missile. They even get to participate in a simulated launch sequence in the control room to get a sense of what those operations sounded like.

Here’s what we read and liked this week.

As the COOs from both Uber and Microsoft recently learned, encouraging company engineers to use AI aggressively can lead to hefty usage bills, perhaps even offsetting all the gains from laying off employees. 

The AI bills at Netflix may not be so eye-popping thanks to company senior engineer Tejas Chopra, who has created software to prune agent instructions, as measured in tokens, before they hit the LLM. 

Chopra has estimated that as much as 90% of tokens are redundant to the giant thinking machine of your choice. 

Although not an official Netflix project, several teams there already use Project Headroom, and a number of external projects rely on it as well.

In a talk at the Open Source Summit last week, Chopra said that Headroom has saved an estimated $700,000 for its users, who collectively now have 200 billion tokens to spend elsewhere. 

Not bad for an open source application that’s been out only since January. Headroom, currently at a still-raw v0.22, has gathered 2,000 stars on GitHub and has been forked over 120 times.  

“A lot of our users are people who have been really burned by token costs, more than anything else,” Chopra said in his presentation. 

Lossless context compression

A $287 bill from Claude Sonnet first brought Chopra’s attention to the idea of token economization.

The bill was typical home project stuff: a bit of debugging, some refactoring, MCP tools querying a database. At the time, Claude Sonnet’s token-based pricing seemed pretty generous: $3 for every million input tokens, or $6/million if you went over the 200,000 token limit for your context window. Still, that $287 added up quickly. 

Upon deeper inspection, Chopra found a lot of this data was highly redundant to the LLM. By and large, his own hand-crafted instructions were not the culprit. Rather it was all the boilerplate and machine metadata that came along for the ride: Needlessly-verbose JSON schemas, nested templates within API responses, identical database columns. 

“This isn’t prose. This isn’t creative writing. This is compressible data masquerading as text,” Chopra wrote in a blog post introducing his software.  In 2025, a group of researchers found that reading user input accounted for about 76% of all token consumption. 

The model providers have their own tools to save tokens. But to date, the settings on these tools are somewhat oblique to end users. By default, Claude has a prefix cache setting of just five minutes. After five minutes of inactivity, the entire context window needs to be refreshed, even if the LLM needs the exact same data. Another setting is exposed in the API documentation: a one-hour time to live (TTL). But there is a catch. “You pay two times the cost for your writes to get 90% savings for your reads,” Chopra told the audience. It’s up to you to find the sweet spot.

There are also a number of new commercial token barbers popping up, such as YCombinator-funded Token Company, which offers token compression as a service. On the open source side there is RTK (Rust Token Killer), which trims to the output of verbose commands, such as calls to a repository. Another open source project, LeanCTX, is a variant of RTK. 

All these tools are useful, Chopra admitted, but he designed Headroom to keep the operations confined to the developer’s workflow. And it had something none of the apps and services could offer: reversible compression.

Headroom’s job is to compress all the source material that is fed into the user’s context window – not only the conversation history, but also logs, tool outputs, files, chunks of documentation that the RAG found useful – before it arrives at the LLM. 

The context window is the set space for each user session. The latest frontier models are rapidly expanding their context windows upwards towards two million tokens, which holds both input and output.  

Such generosity is a mixed blessing, as Pope Leo might point out. As a unit of measurement, a single token is more or less equivalent to a human word. For pay-as-you go plans, the more you feed the context window, the more you’ll pay. 

Gobbling tokens like Pac-Man

Running on Python and Node, Headroom runs as a proxy (port 8787) on the engineer’s computer. The user wraps their LLM at the command line interface (i.e. “headroom wrap codex”) and it then parses the input.

While Headroom does compress a bit of programming code and human instruction, it is best at chopping server logs (90% of which can be jettisoned), MCP tool outputs (70% redundant JSON), Database outputs (it’s all one schema), and file trees (much repeated metadata).

Headroom’s first step is a process called CacheAligner which looks only for information that has been changed within input that’s already been entered, and ships only the new info, eliminating the need to replace an entire body of mostly unchanged text in KV Cache, the cache where the AI provider stores the user’s context window.  

“If your system prompt contains a date field or contains some UUID that changes per session, you are effectively getting a cache miss every single time,” he told the audience. “That will blow up your costs.”

Then, a router process infers the type of content and sends it to one of a number of compressors. An Abstract Syntax Tree (AST) compressor squishes programming code. JSON and Document Object Model (DOM) compressors snip unneeded JSON and Web boilerplate, respectively.

Headroom also has some “squashers” that look at text or JSON input and decide which bits are actually relevant, based on statistical analysis. These tools learn in a feedback loop if they are over- or under-compressing, based on how often the model has to call back into the original uncompressed prompt.  

The final process, called Compress Cache and Retrieve (CCR), offers that ability for the LLM to look at the original unsquashed data. It puts markers to where the data has been compressed, so if the LLM wishes to get the original context, it can call a Headroom MCP to retrieve the needed material from the user’s machine. The original context is stored on Redis or SQLite.  

There is still work to be done to this software stack, Chopra admitted, particularly on testing accuracy. It should be an easy task because the CCR stores the original prompts. More compressors can also be built for other specific types of data, such as financial data.  

Audio, image, and video will also have to be tackled (one user has already forked the project for video parsing). A related project, which Chopra says will be open source soon, is Headlight. Headlight will keep track of the origin of each token, which could be especially handy for ensuring the accuracy of multi-model work.  

A token saved is a token earned

Minding your tokens does not only save money, it can improve results, research suggests. 

Agents send more context than the model can possibly use, which, in addition to emptying the user’s coffers, can actually make the LLM dumber. 

Like the rest of us, LLMs get confused when presented with too much information. A group of Stanford University boffins found that LLMs tend to pay more attention to the beginning and the end of the context window, and tend to disregard the middle bits. 

Likewise, a set of researchers from data integrator Chroma deduced that, across 18 LLMs, “performance grows increasingly unreliable as input length grows.”

“Context rot,” they called this phenomenon. 

Trimming prompts can also improve latency. In his presentation, Chopra relayed how one of Headroom’s users forked the software for a voice-activated application. With voice, even silence can generate tokens. The user expects a response from the app within 200 milliseconds for the service to sound natural, so the company is using Headroom to help shrink that latency window down as much as possible. 

Headroom also offers some good news for those worrying about data centers heating the world into a fiery inferno with their energy usage. Fewer tokens means a smaller context window, which means less energy use – at least until Jevon’s Paradox kicks in and people find even more power-hungry ways to render their animated cat movies. ®

The YouTube-to-prestige-horror pipeline is looking very strong this weekend.

Taking the number one spot at the box office is “Backrooms,” a feature film expansion of Kane Parsons’ series of YouTube videos featuring eerie found footage of a mysterious office space (drawn from a 4chan thread) that defies physics.

Directed by Parsons, “Backrooms” made $38 million on Friday, and is expected to bring in a total of $80 million to $90 million at the domestic box office over this weekend alone. For indie studio A24, that’s its biggest opening by far — the previous record was held by “Civil War,” which made $25.7 in its first weekend of release.

The number two film, “Obsession,” is pulling off something that’s arguably even more impressive. True, it made a mere $8 million on Friday, with an estimated weekend haul of $28.5 million — but the movie (about a romantic wish gone nightmarishly wrong) already made more money in its second weekend than its first, and now its third weekend is set to grow another 19 percent.

For context, most wide release films normally fall between 50 to 70 percent in their second weekend; last year’s “Sinners” was considered an extraordinary word-of-mouth success because it fell less than 5 percent. Outside of Christmas releases (which have more staying power, thanks to the holidays), growing from weekend to weekend is unheard of — according to the Hollywood Reporter, “Obsession” is the first film since 1982 to grow on both its second and third weekends.

And like “Backrooms,” “Obsession” is a horror movie directed by filmmaker who first made his name on YouTube — Curry Barker, who released the hourlong found footage horror film “Milk & Serial” on YouTube in 2024. Barker has already shot his next film and is set to direct a new remake of “The Texas Chainsaw Massacre.

The two releases follow the surprise success of “Iron Lung,” a video game adaptation released earlier this year. Directed by Mark Fischbach — better known under his YouTube account name Markiplier — “Iron Lung” grossed nearly $41 million domestically.

In a New York Times article about the recent “YouTube-to-filmmaker boomlet,” Rutgers CInema general manager Mark DelVecchio noted that “lots of YouTubers have tried to make the leap to mainstream movies and come up short.” What sets Parsons, Barker, and Fischbach apart? DelVecchio said that despite their youth (Parsons is 20, Barker is 26), they all have “longevity.”

“At this point, some of them have been making videos for a very long time, and that’s how you develop a loyal audience that will follow you,” he added.

By the way, while I haven’t seen “Backrooms” yet (fingers crossed for tomorrow), I have seen “Obsession.” So I can confirm that it absolutely does not disappoint — I watched most of the second half with my fingers over my eyes, and I may even have screamed a few times.

Foundation Future Industries, a San Francisco startup founded in 2024, sent two of its Phantom MK-1 humanoid robots to Ukraine earlier this year. The company described it as the first known deployment of humanoid robots in a combat theatre. The tests, backed by the US government and conducted with Ukrainian officials, focused on logistics in hazardous areas.

CEO Sankaet Pathak told CNBC the MK-1 testing proved the robots can perform supply pickups that currently expose soldiers to danger. The robots carry approximately 44 pounds. They lack waterproofing and sufficient battery life for sustained deployment.

Foundation plans to send improved Phantom 2 units to Ukraine this year. Pathak says they will have “superhuman abilities” and double the payload capacity. The company is targeting front-line deployment with the US military within 12 to 18 months.

The political dimension is unavoidable. Eric Trump, the second son of the sitting president, recently joined Foundation as chief strategy adviser. The company has received $24 million in government research contracts across the Army, Navy, and Air Force for feasibility testing in inspection, logistics, and weapons handling.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

Democratic Senator Elizabeth Warren alleged the contracts were “corruption in plain sight.” A Foundation spokesperson told CNBC that Eric Trump had been an investor before becoming an adviser. The two parties share a vision of bringing manufacturing back to the US.

Pathak is best known for leading Synapse, a fintech platform that declared bankruptcy in 2024. Foundation also attracted scrutiny after suggesting it had close ties to General Motors, claims GM subsequently rejected. The company’s credibility is a live question.

The military argument for humanoid robots centres on urban combat environments. “Modern urban combat spaces, where there are stairwells, ladders, basements and narrow corridors, were created for human movement,” said Kateryna Bondar, senior fellow at CSIS. Humanoid systems could have advantages over tracked or quadruped robots in these scenarios.

The counterargument is cost and complexity. “Making robots look like humans is a complex and expensive engineering challenge,” said Melanie Sisson at the Brookings Foreign Policy program. “What Ukraine has taught us is the opposite, that we need the ability to adapt rapidly and manufacture quickly and cheaply.”

Ukraine’s war has already become the primary testing ground for AI and robotics in combat. Ground robots deliver supplies to front lines. Autonomous drones conduct precision strikes. The conflict is generating operational data that peacetime testing cannot replicate.

The European defence-tech sector is moving faster on autonomous strike systems. Berlin’s Stark is raising €300 million at a €2.5 billion valuation for kamikaze drones. Destinus manufactures 2,000 cruise missiles annually through a Rheinmetall joint venture. These companies build purpose-designed weapons. Foundation is trying to make a humanoid do the same job.

Pathak said some weaponised uses of the Phantom robots will retain human confirmation in the decision loop. In certain time-critical scenarios, the robots will need to make fully autonomous decisions. The ethical implications of autonomous lethal decision-making remain unresolved internationally.

Foundation’s ambitions are large. Pathak plans to scale production to thousands of units this year. The goal is to deliver “the best robots we can build” to the US military, “better than anything China has.” China has its own leading humanoid companies and has publicly funded military robotics initiatives, though the extent of its trials remains unclear.

The broader humanoid market is splitting into clear use cases. 1X ships home robots at $20,000. Colin Angle is building companion robots with bear cub ears. Foundation is building robots that carry supplies through artillery fire. The technology is the same. The applications could not be more different.

Toby Walsh, chief scientist at the University of New South Wales AI Institute, expects tracked, flying, and underwater robots to replace human forces before humanoids do. “It might be a science fiction trope to expect humanoid terminator-style robots,” he said. The age of AI robots in war is near. Whether they need to look human to fight is the question Foundation is spending $24 million in government contracts to answer.