shah retweeted
ChatGPT has now a big problem.
Researchers at Oxford and Cambridge exposed a massive threat to large language models.”
They call it “model collapse."
Internet ecosystem is rapidly changing, and generative AI will soon contribute much of the text found online. This forces us to consider what happens to future iterations like gpt-n when they are trained on data scraped from the web that was already generated by an llm.
According to the research, indiscriminately using model-generated content in training causes "irreversible defects" in the resulting ai. the model loses the "tails of the original content distribution." in other words, it begins to forget the creative, fringe, and unique nuances of actual human writing, collapsing into a repetitive echo chamber.
This isn't just a chatgpt issue.. the researchers built theoretical intuition showing this collapse is ubiquitous across learned generative models, occurring in large language models as well as in variational autoencoders and gaussian mixture models.
Tech companies rely on scraping the internet for large-scale data to build smarter models. However, the paper warns that if we want to sustain the benefits of training on web data, model collapse must be taken seriously.
Ultimate takeaway? data collected from genuine human interactions is going to become increasingly valuable in a web filled with ai content.
Shenyang, China. Heavy rains caused water to stagnate and flood all ground flood units. Cars got submerged. But nobody is complaining on Weibo or Wechat. Social media is full of "How this is good, as repairs and cleaning will help the economy." Very good perception management.
If you are in India, and have to share one fact to your friends and family this summer, use this:
In the last many decades, the whole world has been heating up. Annual mean temperature in every country has been rising 1-1.5 degrees a decade.
But India has been an anomaly. It has heated up the least at 0.5-1 degrees a decade. This is a scientific fact acknowledged by climatologists worldwide.
India has also increased its green cover, tree cover in cities, and forest cover in the last decade or so. This is not India's own data alone.
This is data by UN in their Global Forest Resources Assessment – GFRA 2025 report based on satellite imagery, which can't lie.
India has also increased ecological conservation efforts in recent decades as its wealth grew. It has improved its populations of endangered animals like Asiatic lion, Rhinos, Tigers and many birds and animals.
This is much better than many other countries - even developed ones (where they will shoot a wolf, shark, or bear if they increase in population, get close to human habitation, and attack a human).
Anyway, there will always be programmed Indians who will come to discount India's tremendous achievements in Paris climate goals and in protecting its environment even as a highly populous developing country.
We can't do much trying to convince these DS programmed and blackpilled bots. Just watch the comments below. They will come to tell you how numbers are fake, how they saw it has become bad, or change the goal post to another topic to berate India.
shah retweeted
What if the “Cry It Out” sleep training (aka extinction-based sleep training) has contributed to mental health issues in young people?
In some ways, it’s the most insane thing to do to a child (and is based on incredibly poor science). For centuries, families co-slept without issues, but in modern times, it has become increasingly taboo… why?
How can repeated emotional non-response to a baby be healthy? What does it do to their stress calibration, attachment expectations, and self-regulation? How does it play out in their long term relationships and social connections?
I’ve read the studies and they are poorly designed and weakly supported. Yet, we have an entire generation of parents that blindly follow this insane protocol without reviewing the data themselves.
To be fair, the data supporting co-sleeping is weak as well, but it has centuries of precedent so I feel much more comfortable supporting that than a new approach that was largely instituted since the 1920s.
For some context, in the 20th century, behaviorist John Watson (1928), interested in making psychology a hard science, took up the crusade against affection as president of the American Psychological Association. He applied the paradigm of behaviorism to childrearing, warning about the dangers of “too much mother love”. The 20th century was the time when “science" was assumed to know better than mothers, grandmothers, and families about how to raise a child. Too much kindness to a baby would result in a whiney, dependent, failed human being.
A government pamphlet from the time recommended that "mothering meant holding the baby quietly, in tranquility-inducing positions" and that "the mother should stop immediately if her arms feel tired" because "the baby is never to inconvenience the adult." A baby older than six months "should be taught to sit silently in the crib; otherwise, he might need to be constantly watched and entertained by the mother, a serious waste of time."
The truth is the opposite. We now know that ignoring a child raising cortisol levels and hurts trust and attachment. Yet, every young parent I know today has been brainwashed to let their child cry in silence. It’s truly wild.
shah retweeted
⚡️Copper screaming to all time highs at the same moment China’s industrial profits are collapsing tells you the global economy has split into two realities.
One reality is demand destruction, margin compression, overcapacity, and deflation inside legacy manufacturing systems. China lives there right now. Overbuilt factories. Weak domestic demand. Export dependency. Falling profits.
The other reality is physical inevitability. Power, compute, grids, electrification, defense, data centers. This reality does not care about consumer confidence. It does not care about GDP optics. It does not slow politely. It demands materials or it fails.
Copper sits at the intersection of those two worlds.
That is why this move feels confusing if you are still using old mental models.
Copper is not rising because growth is booming.
Copper is rising because systems cannot function without it.
This is the deepest truth most people are missing:
We are pricing maintenance of complexity.
Modern civilization has crossed a threshold where keeping things running requires more energy, more wiring, more redundancy, more throughput than before. AI accelerates this brutally. Electrification accelerates it further. Grid hardening accelerates it again.
Once complexity crosses that line, materials that sit at the base layer stop behaving cyclically.
They behave existentially.
Copper is one of those materials now.
That is why copper can make highs while profits fall, while housing freezes, while trade volumes wobble, while credit tightens. Those systems can contract. The electrical substrate cannot.
This is also why you are seeing the same pattern across silver, copper, energy, uranium, and grid infrastructure.
They are failure prevention inputs.
And here is the part nobody wants to say out loud.
The world built a digital future on an analog foundation that cannot scale smoothly.
You cannot software your way out of copper shortages.
You cannot financialize your way out of grid physics.
You cannot narrative your way around ore grades and permitting timelines.
Price is simply the first signal that the foundation is straining.
So when copper hits all time highs it means this:
The system is quietly admitting that the future is heavier, hotter, more expensive, and more constrained than anyone planned for.
That is the real signal.
And it does not reverse until the architecture itself changes.
shah retweeted
"One of the most helpful things that anybody can learn is to give up trying to catch the last eighth – or the first.
These two are the most expensive eighths in the world"
Jesse Livermore
shah retweeted
Books recomended by @monsterstocks1
shah retweeted
How I Value Biotech Companies:
Many investors don't even try to value biotech as it is just way to complicated. Some even believe it is impossible. All analysts will use the standard Discounted Cash Flow model which is very subjective. My system for valuing biotech is something I developed over the years. I built this system to be an easy way to get a general ballpark valuation for a company so I can tell if it is overvalued or undervalued. I built it from many years of watching how company's stocks respond to clinical data success or failure. The value a company loses or gains on a clinical data read out informs us to what the value was attributed to that program in the value of the stock.
For this to work, I have to be able to assume two variables being true. The first is the valuation takes into account the potential of the program. A program that could do $1 billion in sales would be worth more than one that could do $250 million in sales. The second assumption is the valuation takes in consideration the level of validation of the program. A phase 3 program is more validated than one in phase 1. It would have a bigger impact on valuation than a program in phase 1. If both these assumptions are true, then we can use the move in a company's stock price based on a clinical failure or success to inform us of the value.
Then my model becomes a sum of the parts model. I take each drug in the pipeline and figure out what it could earn in revenues at peak sales. Then I adjust that value based on the level of validation of that drug. That gives me the value for that single drug. Then I add up all the value for all the programs and add in the cash. If I have a drug worth $500 million value, a second worth $250 million, and the company has $500 million cash, my valuation would be $500 million + $250 million + $500 million = $1.25 billion value.
This is a very basic concept to do valuation. The reality is it can be complex to come up with the value of programs. You need variables like Number of Patients, Drug Price and Competition. This will give us an estimate of peak sales and market share. The key values here are estimating peak sales and adjusting those sales based on market share for competition and then the level of clinical validation. You can often get estimates for sales from analysts or even the company. The harder way is more accurate but takes more time. If you follow your companies, they will often define for you the size of the patient population in their presentations. Then all you need to do is find prices of comparable drugs, and you can calculate peak sales. The only change would be to adjust for potential market share for competition. The level of validation for the drug would simply be the level of clinical success it has achieved. I take the peak sales and adjust them for the phase of development using .1x for preclinical, 1x for phase 1 data, 2x for phase 2 data, 3x for phase 3 data and 5x sales for a mature biotech valuation.
Once you have the value for each drug in the pipeline, you just add them up and add in the cash on the balance sheet. This will give you a total valuation for the company. You can compare the market cap you calculated to the current market cap to see if it is too expensive or too cheap. You can even adjust your valuation based on the outstanding shares and get a stock price for that valuation. Every time a company reports they will issue a 10Q or 10K. The very first page will list the number of outstanding shares at the time of reporting. You can divide your valuation by the shares outstanding and get a price for the valuation of the company.
This is the system I developed to value my biotech companies. I have been using it for at least 10 years now. It has proven to be very accurate as long as I keep my expectations realistic.
Example Using Blueprint Medicines
Blueprint has three drugs in development with Ayvakit, Elenestinib, and BLU-808. We will do the sum of the parts model for the company.
First comes Ayvakit which the company has guided to $2 billion in peak sales by 2030. I am a conservative investor so I marked it down to about $1.5 billion. This drug is already early commercial with good sales growth. That means I would give it the full 3x multiplier for a phase 3 program, but not quite the full 5x multiplier for a fully mature drug. As you can see in my chart below, I got to $4.5 billion in value for Ayvakit.
Then comes Elenestinib. The company believes their entire Systemic Mastocytosis franchise could reach $4 billion. This means they think Elenestinib could reach another $2 billion in sales beyond Ayvakit. I think I would be far more conservative than they are this early in dvelopment until we see the full phase 3 data set. I gave them $1 billion in peak sales for this program in my spreadsheet below. The data for this program has been phase 2 so far. That would set it at a multiplier of 2x. That brings the value for this program at $2 billion.
The last program is BLU-808. This is being developed in Wild Type KIT for a group of mast cell disorders. It could be really big depending on how many of the disorders they test it in work out. For now, I set it at $1 billion which is a low. I can always upgrade that number if it works in multiple indications as the data develops. Since it only has phase 1 data, I will set the multiplier to 1x. That makes this program worth $1 billion in value.
All together, we have $4.5 billion + $2 billion + $1 billion as our sum of the parts for the pipeline. That is a total value of $7.5 billion in value. If we add back in the current cash on the balance sheet, we get an additional $850 million in value. That brings us to a grand total sum of all the parts of $8.35 billion. The outstanding shares as of the end of the 3rdquarter are around 63.6 million. That would come out to a per share price of $131.28.
shah retweeted
The business community is giddy with excitement about the @realDonaldTrump administration. I am hearing this from everyone, including from people who didn’t vote for Trump.
Business confidence is a self-fulfilling prophecy. Business leaders are becoming more confident about the country and the economy. This means they will be making more investments in our future which will drive the economy and the stock market, reducing the cost of capital and bolstering confidence further, catalyzing more investment and more growth in a self-reinforcing, virtuous cycle.
Merger and acquisition activity is about to explode as there are an enormous number of deals that have been deferred pending a more favorable regulatory environment for transactions. M&A will drive efficiency, greater profitability and growth. It will also enable the return of capital to investors who will seek to redeploy their profits and proceeds in new investment opportunities.
The DOGE and deregulation will drive government efficiency and make America a vastly better and lower-risk place to do business efficiently and effectively.
I am also hearing about non-US companies that are desperate to immediately create a presence on U.S. soil. They are frightened to be locked out of the most important economy in the world. They will build factories and make new investments here to avoid the risk of tariffs and because they don’t have a better place to invest capital.
China’s economy is in trouble. Europe’s is a mess. The U.S. has now become by far the best country for investment.
Growth is about to explode.
shah retweeted
If there is one person on Earth who stands as the embodiment of unbreakable tenacity, unshakeable grit, relentless determination and the courage to stay true to his beliefs, it is Donald Trump. Fascinating to see America’s democracy empower its people and uphold the nation's founding principles. Congratulations to the 47th POTUS-elect @realDonaldTrump.
shah retweeted
“Why are you bullish all the time?”
Because if you’re bearish more than 10% of the time you’re an idiot.
shah retweeted
Turn your laptop into a $10,000-a-month money machine with just:
• An internet connection
• A laptop
• Basic Google research skills
Here's how to do it:
shah retweeted
MISSION SUCCESS for our 53RD Electron launch! Welcome to orbit, @KineisIoT. We’ll see you back on the pad soon!
shah retweeted
Robert F. Kennedy Jr. Address to the Nation teapawt.net/i/broadcasts/1yoKMygbV…































