Idea Generator is about deep value. It gives you access to simple, deep value ideas structured in a risk-managed Portfolio.
Portugal
Joined June 2012
- Tweets37.4K
- Following508
- Followers3.7K
- Likes258
The next blowup is upon us - there are widespread security breaches all across the world, using AI. All the way up to official government websites. It's everywhere.
A funny thing about cooling data centers in space is that if the radiators need to be pointed all in the same direction, they'll provide thrust.
Paulo Santos retweeted
In 1917 a French microbiologist found a virus that kills only bacteria. The West forgot about it. A lab in Tbilisi didn't, even after its founder was executed in Stalin's purges.
This month, a startup bet AI can bring it back.
The microbiologist was Félix d'Hérelle. He called the viruses bacteriophages, "bacteria eaters." Each one hunts a specific strain and leaves your own cells alone.
In 1919 he used them on a 12-year-old boy dying of dysentery in Paris. The boy recovered.
A young Georgian scientist, George Eliava, met d'Hérelle at the Pasteur Institute and brought the idea home. Together they built a phage institute in Tbilisi.
In 1937 Eliava was arrested and shot.
Then penicillin arrived. Antibiotics were cheap, broad, and easy to mass-produce, and the West dropped phages almost completely.
Georgia didn't. Eliava's institute kept going for decades. Today, patients with infections no antibiotic can touch still fly to Tbilisi as a last resort.
Meanwhile, resistance caught up. In 2019 alone, drug-resistant bacteria directly killed 1.27 million people.
The catch with phages was always speed. You have to find the right virus for one patient's exact strain, and that can take months of searching.
On 20 October a California startup, Radical Numerics, said it had used its AI model to design phages for a patient with drug-resistant E. coli. They were given under a single-patient FDA approval after killing the bacteria in the lab. It's one case, and the outcome hasn't been shared.
But the bottleneck that made the West give up on phages, finding the right one fast enough, is exactly the kind of problem AI is good at.
We may have spent 80 years with one answer to antibiotic resistance sitting in a lab in Georgia.
Which abandoned idea would you bet AI revives next?
Shortage pricing. Historically it doesn't last, but maybe this time is different.
Elon Musk says there's no AI demand issue and explains why SpaceX rents its compute on short-term contracts: AI compute is worth more per watt every year, renewals are coming in at higher prices, and the market is a “cheese shop” with no chips left
“Our belief is that the value of compute per watt will increase over time, not decrease. And the reason for that, I think it should be obvious to people really, is that as you improve the AI software, as the models get better and better, they’re able to, you get more useful intelligence per watt. And the value of that compute is proportionate to the useful intelligence per watt. And that’s why you see things like even old computers like H100s renting for more than they rented when they were new.”
“And I think that’s likely to continue. I think we’re going to face a shortage of superintelligence compute. And that shortage is going to remain for a long time. So, I think the smart move is to do short-term contracts and to essentially place the bet on useful intelligence per watt increasing rapidly, which obviously is done, so that’s what’s happened so far.”
“And also if you look at, say, how are our short-term contracts renewing? Are they renewing at higher dollars per watt or lower? And we see them renewing at higher dollars per watt, not lower dollars per watt. We are also frankly hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute. So we could rent out our compute several times over if we said yes to everyone. That’s just the situation here.”
“That’s why I posted that sort of video meme of the Monty Python cheese shop sketch where the guy goes into a cheese shop to get some cheese and they actually have no cheese. And so he goes like, what about this kind of cheese? What about that kind of cheese? And they really just don’t have any cheese.”
“So we’ve actually tried to offload some of our inference compute to other cloud providers so that we can use more compute for training. And we get the cheese shop sketch. You know, they’ll say, like, well, we’ll ask them about what about some, you know, GB300s? Nope, don’t have any of those. What about, you know, some GB200s? Nope, don’t have any of those. What about some TPUs? Nope. What about some Trainiums? Nope. What about some, okay, how about some H100s? And like, well, we might have some overseas somewhere and then they’ll check and come back and say, nope, we don’t have that either. So it’s like ridiculous.”
“Yeah. So I don’t think we’re going to have a demand issue here. And like, it’s not impossible. I’m not saying this will happen. It’s not impossible that we will have a six-month payback period for compute next year.”
Some kind of communist craziness going around in Germany.
Germany Explores Stealing Properties from Landlords in Berlin and Paying them Below Market Prices, Bonds Crater (this is the end of Berlin as we know it)
EUROPE IS DOOMED BEYOND BELIEF IF THIS PASSES
On Saturday Die Linke (German Left Party), Greens and SPD in Berlin presented the exploratory paper underpinning their coalition agreement.
It envisages socialising, under Art. 15 of the German constitution (GG), private landlords' Berlin stock above 10k units per company.
On documented counts that is ~152k units: Vonovia 128k, Heimstaden 10k, Adler 7.4k, Covivio 7k.
Compensation is capped at €16bn, about €105k per unit or 58% of book on a pro rata basis, and even the cap is conditional on fiscal sustainability, i.e. it can fall if financing conditions worsen.
The paper commits only to a 2027 financing concept; whether that means cash or 100-year bond compensation is left open. The bill reaches the Berlin parliament in mid-2028 and takes effect for transfers one year after promulgation, so mid-2029 at the earliest.
- Adler exposure: Adler has got 7.4k units above the 10k unit cap, ie its yielding portfolio is exposed by
~43%. At the cap Adler would receive €105k per unit on a flat split or closer to €115k pro rata to book, i.e. €777m to €860m against €1.49bn of book - a 42% to 48% discount to book. The implied GRI yield/cap rate at the cap is ~7.0%.
- Heimstaden exposure: Heimstaden has more than 20k units in Berlin so ~10k units above the 10k allowance are exposed. That's 6% of group total. By value the exposed share will be a tad higher given the German book has higher values per unit vs group.
Beyond Art. 15, the exploratory paper explicitly names Adler's Steglitzer Kreisel as its example of a "speculative construction ruin" (Bauspekulations-Ruine): it will sanction inactivity on such sites, expects completion "often only after insolvency proceedings", and will use "transfer of ownership" (Eigentumsübertragung).
Moreover, a rent regulator (Mietenaufsicht) starts in 2027, handling rent overcharging (§ 5 WiStG, rents more than 20% above the local reference rent) and running a rent and housing register from 2028, with a new rent-usury (Mietwucher) unit at the public prosecutor's office alongside it.
gruene.berlin/fileadmin/user…
Meanwhile, a reminder that $TSLA still ships cars from China to Europe with tariffs up to 66% lower than those the European brands pay to do the same. 😁
Paulo Santos retweeted
테슬라 로드스터 ‘날아다니는’ 데모, 결국 사람이 탈 수 없는 쇼였다
테슬라가 9년 만에 보여준 로드스터 스턴트는 진짜 인상적이었다. 램프를 타고 올라가 공중에서 뒤집히고, 스러스터로 자세를 잡아서 착륙하는 장면은 확실히 영화 같았다.
그런데 한 가지가 계속 걸렸다.
사람이 타면 안 되는 환경이라는 점이다.
이벤트 현장에서 관람객들에게 귀마개를 나눠줬다. 스러스터 소음이 청력 손상 위험이 있을 정도로 크기 때문이다. 그래서 데모 자체도 운전자 없이 원격으로 진행됐다. 고압 가스를 고속으로 분사하는 방식이라 진동도 상당할 수밖에 없다. 엘론 머스크도 “안전 최적화 차량이 아니다”고 직접 말했다.
탱크 용량도 제한적이라 짧은 버스트만 가능하고, 일반 도로에서 쓰기엔 규제·소음·진동 문제가 크다. 결국 이번 데모는 ‘차가 이렇게까지 할 수 있다’는 기술 과시이자 마케팅에 가깝다.
2017년에 예약금 받고 지금까지 제대로 된 차를 내놓지 않은 상태에서, 사람이 편하게 탈 수 없는 쇼를 먼저 보여준 셈이다.
물론 극소수 트랙 애호가나 브랜드 팬에게는 의미 있을 수 있다. 하지만 대부분의 사람에게는 실용성이 거의 없는 제품이라는 결론이 자연스럽게 나온다.
차가 날아다니는 것보다, 예약자들이 기다린 차를 언제 받게 될지가 더 중요한 문제 아닐까.
갑자기 상업적으로 완전 실패한 사이버트럭이 생각나는 건 나만일까?? 아니 그보다 더한…
Paulo Santos retweeted
Imagine following nearly 13 million real people for more than 10 years just to see what cholesterol level helps them live the longest. That is exactly what Korean scientists did in a giant 2019 study. They tracked 12.8 million adults who had check-ups between 2001 and 2004, and they kept watching until 2013. More than 694,000 people died during that long stretch.
The results were clear and surprising. Cholesterol and death formed a U-shape. Very low cholesterol was linked to higher death rates. So was very high cholesterol. The sweet spot for the longest life was 210 to 249 mg/dL for most adults. That is higher than many doctors call “ideal.” Younger people had a slightly lower best range, but still not the super-low numbers often pushed today.
Because the study lasted so many years, reverse causality is less likely to explain the low-cholesterol danger. Reverse causality means a hidden illness might lower your cholesterol right before it kills you. With more than a decade of follow-up, early sickness cannot easily account for the extra deaths seen in people who started with low cholesterol. The pattern held across ages and for both men and women.
The huge number of people—almost the entire adult population of a country—makes the findings hard to ignore. Smaller studies can be wrong by chance. This one is massive. It shows that moderately higher total cholesterol, not the lowest possible number, went with the best chance of living longer in this large group.
In short, the biggest study of its kind found that cholesterol levels in the low-to-mid 200s were tied to the longest lives, and the long tracking time makes it tougher to blame the higher death rates at low levels on reverse causality alone.
In general, it's impressively obvious that this was an optimized prototype to put on a demo.
That such can then be mated with a 2028 date which obviously won't happen, and that the same scheme can be run over and over, is the most impressive part of it all.
Paulo Santos retweeted
Cool stunt. Great hype marketing.
Time for me to be that guy everyone hates - It's a scam and this will never ship.
The external compressor and cooling system to charge the tank once, likely cost more than the car. A multi-stage compressor and heat echanger setup that would fit in a car, would take several days to the better part of a week of constant running for a single charge.
250kg of 10,000psi air in a composite tank, with all the potential failure points of complex plumbing and valves in a human occupied confined space... If your sphincter didn't just make a diamond reading that sentence, you're not comprehending it.
Go ask grok about the physics involved just in compressing, cooling, storing and controlling a compressed air system with these specs, and what's possible with typical industrial equipment. This is just a cool gimmick to pump fans and stock.
Id bet money, the car in the demo didn't have a full size battery in it. Wouldn't be surprised if someone isn't currently doing analysis on the video to calculate the vehicles weight.
Sorry.
Also, $TSLA is already fully aware this car won't be for sale in 2028 - not even the most conventional version.
Why? Because the car shown is a lightweight, unrepresentative, prototype only made for demo purposes. Look at the car from below while it flies.
Hence, in late 2026 the car doesn't even exist as more than a demo shell. It won't be for sale in 2028 even excluding the flashy thrusters.
The flying $TSLA roadster will also not be sold, just like the previous Roadster.
People seem to have a problem distinguishing between what's possible on a demo, and what can be sold because it works reliably and safely enough. Flying cars, upsidedown cars, etc, none of that is new.
I have actually found out what China's social credit system is, and how it's misrepresented.
The restrictions often ascribed to China's "social credit system" are exactly those applied to people under 1 of 2 court-mandated measures:
* "High consumption restrictions" and
* "unreliable/dishonest judgment debtor"
These restrict a person's ability to enjoy what are deemed luxuries, including air travel, some high-end fast train travel, star-rated hotel stays, buying new real estate, etc.
A person sees these measures applied to him if he fails in paying a credit a court judged he was liable for, and also if he has the means to pay but doesn't (the list is longer, this is the gist).
So there ARE a lot of people in China going around with these restrictions, but they don't mean anything regarding a supposed social credit system - just these measures are distorted into making it seem like such a thing exists.
Of note, in 2026 $TSLA is exporting nearly 50% of its Shanghai factory output.
China’s Ministry of Finance and State Taxation Administration have jointly announced that the VAT export rebate rate for new energy vehicles (NEVs) will be cut from 13% to 9%, effective April 1, 2027.
In simple terms, this will raise the cost of exporting vehicles and squeeze automakers’ gross margins.
The move is also likely to push major Chinese automakers to accelerate overseas factory construction and local assembly, reducing their reliance on exports of fully built vehicles from China.
