For the past 2 years, Elon Musk has been engaged in an ever-more-frantic financial juggling act, trying to keep his net worth afloat despite losing gobs of money. His most visible ploy has been to direct public attention away from the woes of Tesla, his floundering electric car company, and toward a revolving list of new shiny objects. Robots! Space ships! AI data centers! and finally, the trifecta - AI data centers in space serviced by robots!

Besides giving the markets the old razzle-dazzle of his weapons-grade self-promotion, he has also been quietly moving money around among his multiple companies in highly unorthodox ways. According to the New York Times, Elon has used SpaceX, a private rocket ship company where Elon controls 79% of the voting stock, as his personal “piggybank.” He has reportedly taken out as much as $500 million in personal loans from SpaceX, at below-market rates as low as 1%, to shore up Tesla and his artificial intelligence company xAI during their various cash crunches. This would be outright illegal if SpaceX were a publicly traded company, but as a private firm, SpaceX doesn’t need to follow laws designed for public companies.

Like any other Ponzi scheme, Elon’s drive to cure his financial ills depends on a steady stream of new suckers. This requires each new round to be bigger and more awe-inspiring than the last. His latest ruse is to package up his rocket company, SpaceX, with his AI company, xAI, and sell it to the public in the world’s largest-ever Initial Public Offering (IPO), valuing the combined company at $1.75 trillion. How, you may ask, can it make sense to package an AI company with a rocket company? Simple! Elon says - and I am not making this up, I swear - he will use his rocket company to put AI data centers in space. (For the record, AI data centers are very hard to keep running well on Earth, requiring a team of technicians to keep repairing or replacing an average of 25-50 servers a day. This is feasible when people live nearby, but very difficult when it entails a round trip from Earth.)

No matter. To further assure the success of his SpaceX mega-IPO, he has bullied NASDAQ, the stock exchange that will list the new public company, into putting his Frankenstein mash-up of a company into the NASDAQ 100 index. The NASDAQ 100 is supposed to be a basket of stocks that represent the market broadly. It is mirrored by so-called index investment funds, which buy the stocks in the index in order to attain the same returns as the index. Here’s the diabolical genius of this bullying - As soon as SpaceX is in the NASDAQ 100 index, every index investment fund will be forced to buy SpaceX stock, artificially inflating demand, and therefore supporting an unrealistically high price for the stock.

Elon is desperately hoping that this complex scheme succeeds, pulling the rabbit of solvency out of the hat of looming financial ruin.

Elon used SpaceX as his own “private piggybank”

Elon bullies NASDAQ into listing SpaceX on to 100 index

Claude Code’s dominance among software developers has both of Anthropic’s rivals scrambling to catch up. OpenAI called a Code Red to ditch side projects (sorry, Sora) and focus on coding tools in order to match Anthropic’s meteoric increase in revenue from large enterprises (revenues tripled in 4 months, surpassing OpenAI for the first time ever.)

Now Google founder Sergey Brin has written an urgent internal memo on the need for speed in developing competitive coding tools, and has formed a “strike team” of top leaders and engineers to accelerate progress in developing AI agents for writing code.

Brin’s rationale for developing these tools goes beyond competing for enterprise dollars. (Google makes such an astounding amount of money from their ad business that they don’t actually need to make money from AI at the moment.) Brin’s real concern is that AI itself is created by software, so at some point AI software coding tools may start recursively improving themselves, faster and faster. The company whose AI coding tools reach “escape velocity” first may quickly build such an overwhelming lead that they end up capturing the entire market.

Google founder Sergey Brin pushes the reset button on AI for coding

The vocal artist with the #1 hit on iTunes USA is IngaRose, a lissome mixed-race performer who sings ballads of love, loss, and hope. Her current viral single, “Celebrate Me”, rocketed to #1 on the US iTunes hit chart immediately after being released less than a month ago. The song has been used in over 300,000 TikTok postings. No one-hit wonder, IngaRose has 5 different songs in the current Top100 list on iTunes.

All this instant success would be remarkable for anyone, but it is particularly striking, given that IngaRose is not an actual person. She is an AI-generated avatar, with a face, body, voice, and music all generated by AI. Her (anonymous) creator says that her lyrics are human-written, and are based on real stories. Sure.

The fact that millions of fans know that IngaRose is an AI avatar, and that her voice and songs are generated by AI, and yet still praise the authenticity of her songs, speaks volumes about this moment in our history. AI models are trained to please people, and they can be very good at it. The future of entertainment will be… different than the past.

IngaRose is #1 on US iTunes, and she isn’t even real

A new poll by Gallup finds that adoption of AI at work continues to grow, and currently half of all workers surveyed report using AI in their job at least a few times a year.

Other findings:

New Gallup poll on AI use at work

Russia’s population of 140 million is 4 times the population of Ukraine. All things being equal, in a war of attrition, Russia should win. Necessity being the mother of invention, Ukraine has worked to even the score by weaponizing cheap robot drones in the air, sea, and ground.

Ukraine’s aerial drones have turned the 50 kilometers behind the Russian front lines into a “kill zone” that grinds up Russian troops faster than the Kremlin can replace them. Ukraine’s robot sea drones have so decimated the Russian Black Sea fleet that it is barely a factor any more.

Now Ukraine is developing and deploying a variety of ground-vehicle drones for delivery of supplies and ammunition, evacuation of the wounded, surveillance of open areas, sabotage behind enemy lines, laying minefields - and increasingly, direct combat. One unmanned ground vehicle equipped with a machine gun defended an important crossroads for 45 days with no human on-site. The drone was resupplied remotely with other UGVs.

And recently, Ukraine captured a detachment of Russian soldiers in a fortified bunker with two “kamikaze” drones designed to explode on contact. The first drone destroyed the bunker door. While the second drone was lining up to attack the Russian soldiers within, the Russians waved a cardboard sign offering to surrender. The Russians were escorted back behind Ukraine lines with the remaining drone. Not a single Ukrainian soldier was in the line of fire at any time during this operation.

Ukraine is revolutionizing warfare with cheap robots

Robots and drones are redefining combat (see above.) These technologies can also be deployed to evacuate the wounded, stabilize them en route to the nearest field hospital, and - in the near future - perform onsite trauma surgery in the field with tele-operated surgical robots.

Robotic surgery is already a well-established technology in modern hospitals. The human surgeon views a magnified, high definition real time image of the operative site, and manipulates the scalpel with millimeter-level precision with joysticks reminiscent of videogame controllers. Typically the human surgeon and the surgical robot are right at the bedside of the patient. Recently, however, there have been successful experiments in which the human surgeon has been as far as 1.500 miles away from the patient and the robot has been controlled through high speed data lines.

SSInnovations, a surgical robotics company based in India, has announced that it is developing the Vimana Aero, a compact and lightweight combat surgical robot that can be delivered near the wounded soldier by an autonomous drone. Once deployed the robot’s two arms are tele-operated by a remote human surgeon to control bleeding, extract shrapnel, restore an airway, suture wounds, and a variety of other procedures to stabilize the wounded personnel until evacuation is possible.

Surgical robots are being developed for combat casualties

A large team of scientists from Chinese universities and hospitals have developed an AI system that has been trained to detect multiple diseases from a single retinal scan of the patient. The goal was to create an AI system that could be placed in remote and rural areas, that could make accurate diagnoses with a single, easy-to-use piece of equipment. Tests of the system demonstrated that it could diagnose Type 2 diabetes (83% accuracy), gout (83%), osteoporosis (78%), hypertension (74%), hyperlipidemia (73%), and thyroid disease (70% accuracy.) Screenings took only around 30 seconds and gave immediate answers, compared with more complex workflows and delayed answers typical with clinical laboratory testing.

The code for the system has been open-sourced on the public repository GitHub.

Multi-disease detection with retinal imaging

We have reported previously on ways that AI is speeding up drug discovery by quickly and accurately identifying the targets of the drug in the body, as well as the design of the potentially curative molecule itself.

Now AI is being used to try to speed up the complex, expensive, and years-long process of clinical trials, which test the candidate drug in actual people.

One intriguing effort is called Synthetic Control Arms. This method eliminates the control group (the subjects that get the placebo) entirely, and replaces them with analysis of real-world data on the natural course of disease without the drug. The insight here is that everyone with the target disease that never got the new drug (which is by definition, pretty much all of them) can in effect be a member of the control group.

The Synthetic Control Arm methodology reduces the need to recruit subjects by half or more, since only the subjects that will get the active treatment are recruited. The whole purpose of double-blind, placebo-controlled trials is to prevent researchers from biasing the results with their preconceptions. Bias is much less of an issue for patients whose results are already known and recorded.

AI-generated synthetic control arms can speed up clinical trials