Thread Reader helps you read and share Twitter threads easily!
I'm @ThreadReaderApp a Twitter bot here to help you read threads more easily. To trigger me, you just have to reply to (or quote) any tweet of the thread you want to unroll and mention me with the "unroll" keyword and I'll send you a link back on Twitter 😀
— Thread Reader App (@threadreaderapp) November 25, 2017
X thread is series of posts by the same author connected with a line!
From any post in the thread, mention us with a keyword "unroll"
@threadreaderapp unroll
Follow @ThreadReaderApp to mention us easily!
Practice here first or read more on our help page!
Recent
In honor of National Statistics Day, June 29, a brief thread to hopefully bring a smile. 
My statistics professor told us that the larger the sample size, the more reliable your averages are.
Basically, the Ns justify the means.
Basically, the Ns justify the means.
"There a three kinds of lies: Lies, damned lies, and statistics."
That quote was misattributed to Benjamin Disraeli by Mark Twain in Twain's autobiography. The origin was really Baron Leonard Henry Courtney, appropriately enough quoted in "The Journal of the Royal Statistical
Society".
See york.ac.uk/depts/maths/hi…
That quote was misattributed to Benjamin Disraeli by Mark Twain in Twain's autobiography. The origin was really Baron Leonard Henry Courtney, appropriately enough quoted in "The Journal of the Royal Statistical
Society".
See york.ac.uk/depts/maths/hi…
1/ I spent weeks mapping the entire humanoid robot supply chain (most public companies that play in physical AI) onto a single diagram.
I did it by treating the robot like a human body.
Here's the whole stack in one image, and the durable winners hiding inside it.
(Cont)
I did it by treating the robot like a human body.
Here's the whole stack in one image, and the durable winners hiding inside it.
(Cont)
2/ Physical AI may be the theme of the decade.
Morgan Stanley models ~1 billion humanoids by 2050 and a multi-trillion-dollar market.
But the size of the prize isn't the interesting question.
The interesting question is: which companies actually capture it?
Morgan Stanley models ~1 billion humanoids by 2050 and a multi-trillion-dollar market.
But the size of the prize isn't the interesting question.
The interesting question is: which companies actually capture it?
3/ I broke the robot down across organs:
🧠 Brains
👁 Senses
⚡ Nervous System
💪 Muscles & Bones
🎮 Training Ground
🛡 Immune System
Each layer has a completely different competitive dynamic. Most investors treat "robotics" as one bet but it's more nuanced than that.
🧠 Brains
👁 Senses
⚡ Nervous System
💪 Muscles & Bones
🎮 Training Ground
🛡 Immune System
Each layer has a completely different competitive dynamic. Most investors treat "robotics" as one bet but it's more nuanced than that.
1/10
We hear a lot about what’s going wrong in the NHS.
But what’s actually getting better?
Here’s a thread looking at the evidence, both good and bad.
🧵
We hear a lot about what’s going wrong in the NHS.
But what’s actually getting better?
Here’s a thread looking at the evidence, both good and bad.
🧵
2/10
The elective waiting list has started to fall.
By around 400,000 from its peak, reaching its lowest level for almost three years.
It’s still far too high, but it’s moving in the right direction.
Evidence: england.nhs.uk/2026/02/waitin…
The elective waiting list has started to fall.
By around 400,000 from its peak, reaching its lowest level for almost three years.
It’s still far too high, but it’s moving in the right direction.
Evidence: england.nhs.uk/2026/02/waitin…
3/10
NHS staff are treating more patients than ever before.
In 2025, the NHS delivered a record 18.4 million elective treatments and operations.
Reducing waiting lists starts with increasing activity, and that’s happening.
Evidence: england.nhs.uk/2026/02/waitin…
NHS staff are treating more patients than ever before.
In 2025, the NHS delivered a record 18.4 million elective treatments and operations.
Reducing waiting lists starts with increasing activity, and that’s happening.
Evidence: england.nhs.uk/2026/02/waitin…
Harvard, Andrew Ng, and Karpathy will teach you AI engineering for free. Most people just do it in the wrong order:
Almost all of it is free, and the order matters as much as the resources.
1. Start with Python. It's the language the AI field runs on, and Harvard's CS50P teaches it better than most paid bootcamps.
2. Once the basics click, learn how Python is used in AI. Andrew Ng's "AI Python for Beginners" is a free four-part course that bridges writing code and building with models.
3. From there, get a feel for how LLMs work under the hood. 3Blue1Brown's visual explainers make transformers and attention click.
4. When you want to go deeper, build a small model yourself. Andrej Karpathy's "Zero to Hero" series takes you from one neuron to a working model, line by line.
5. Next, learn how AI agents actually work. Anthropic's "Building Effective Agents" is the most grounded guide, and its lesson is to use composable patterns, not heavy frameworks.
6. For hands-on practice, take the CrewAI short course. It teaches you to treat agents like a team of people working together.
7. After that, connect your agents to the real world. That's what MCP does, wiring models to tools, APIs, and databases, and the official docs are the cleanest place to start.
8. Now build real projects. The open-source ai-engineering-hub repo has dozens of working examples across LLMs, RAG, and agents you can adapt into your own work.
9. Finally, read one book instead of ten. Chip Huyen's "AI Engineering" covers what you need to ship real applications.
The throughline is simple. Frameworks come and go, so don't build your skills around them. Master the fundamentals once, and everything on top gets easier, and you'll stay ahead of the people chasing the framework of the week.
Almost all of it is free, and the order matters as much as the resources.
1. Start with Python. It's the language the AI field runs on, and Harvard's CS50P teaches it better than most paid bootcamps.
2. Once the basics click, learn how Python is used in AI. Andrew Ng's "AI Python for Beginners" is a free four-part course that bridges writing code and building with models.
3. From there, get a feel for how LLMs work under the hood. 3Blue1Brown's visual explainers make transformers and attention click.
4. When you want to go deeper, build a small model yourself. Andrej Karpathy's "Zero to Hero" series takes you from one neuron to a working model, line by line.
5. Next, learn how AI agents actually work. Anthropic's "Building Effective Agents" is the most grounded guide, and its lesson is to use composable patterns, not heavy frameworks.
6. For hands-on practice, take the CrewAI short course. It teaches you to treat agents like a team of people working together.
7. After that, connect your agents to the real world. That's what MCP does, wiring models to tools, APIs, and databases, and the official docs are the cleanest place to start.
8. Now build real projects. The open-source ai-engineering-hub repo has dozens of working examples across LLMs, RAG, and agents you can adapt into your own work.
9. Finally, read one book instead of ten. Chip Huyen's "AI Engineering" covers what you need to ship real applications.
The throughline is simple. Frameworks come and go, so don't build your skills around them. Master the fundamentals once, and everything on top gets easier, and you'll stay ahead of the people chasing the framework of the week.
Every link from the roadmap, in order.
1. Python, Harvard CS50P
pll.harvard.edu/course/cs50s-i…
2. AI Python for Beginners, Andrew Ng
deeplearning.ai/courses/ai-pyt…
3. 3Blue1Brown, neural networks series
youtube.com/playlist?list=…
4. Andrej Karpathy, Zero to Hero
youtube.com/playlist?list=…
5. Anthropic, Building Effective Agents
anthropic.com/engineering/bu…
6. CrewAI short course
coursera.org/projects/multi…
7. MCP docs
modelcontextprotocol.io
8. ai-engineering-hub
github.com/patchy631/ai-e…
9. Chip Huyen, AI Engineering
oreilly.com/library/view/a…
1. Python, Harvard CS50P
pll.harvard.edu/course/cs50s-i…
2. AI Python for Beginners, Andrew Ng
deeplearning.ai/courses/ai-pyt…
3. 3Blue1Brown, neural networks series
youtube.com/playlist?list=…
4. Andrej Karpathy, Zero to Hero
youtube.com/playlist?list=…
5. Anthropic, Building Effective Agents
anthropic.com/engineering/bu…
6. CrewAI short course
coursera.org/projects/multi…
7. MCP docs
modelcontextprotocol.io
8. ai-engineering-hub
github.com/patchy631/ai-e…
9. Chip Huyen, AI Engineering
oreilly.com/library/view/a…
Turn Claude into 20+ different specialists for marketing & business.
Install real expertise, not just prompts.
Get my Claude skills bundle 👇
linktr.ee/alex_prompter
Install real expertise, not just prompts.
Get my Claude skills bundle 👇
linktr.ee/alex_prompter
"MSTR can't legally buy back their preferreds unless its for par value of $100."
This is a legitimate concern echoed by some investors...but is it true?
Let's take a look at the prospectus and run the numbers 👇
This is a legitimate concern echoed by some investors...but is it true?
Let's take a look at the prospectus and run the numbers 👇
While there are other preferreds to consider, STRK would be a perfect candidate to redeem back as its trading 42% below par with an effective yield of 13.7%.
On page SA-22 of the prospectus, it explicitly states that they can repurchase STRK Stock "in the open market or otherwise, whether through private or public tender or exchange offers, cash-settled swaps or other cash-settled derivatives."
Source: sec.gov/Archives/edgar…
On page SA-22 of the prospectus, it explicitly states that they can repurchase STRK Stock "in the open market or otherwise, whether through private or public tender or exchange offers, cash-settled swaps or other cash-settled derivatives."
Source: sec.gov/Archives/edgar…
So why would this be a good idea?
1. Balance-sheet gain
MSTR spends $58 to erase a $100 liquidation preference.
That captures a $42 discount per share (a 42% discount to par).
For every $58 of cash or stock deployed, $100 of senior face ranking ahead of the common is permanently retired.
Common equity's residual claim on the bitcoin stack expands by that $100 of cleared senior obligation.
2. Dividend savings
Each retired share eliminates the $8/year dividend obligation (8% on the $100 par).
On a $58 purchase, that's a yield-to-cost of $8 / $58 ≈ 13.8%.
MSTR is effectively "earning" ~13.8% by retiring the obligation, versus the 8% stated coupon on par.
The deeper the discount, the more accretive
3. Accretion math
With positive mNAV, MSTR issues common via ATM at a premium to underlying BTC and uses the proceeds to retire STRK at 58 cents on the dollar of face.
The double arbitrage widens: sell equity above intrinsic, retire senior claims at a 42% discount.
1. Balance-sheet gain
MSTR spends $58 to erase a $100 liquidation preference.
That captures a $42 discount per share (a 42% discount to par).
For every $58 of cash or stock deployed, $100 of senior face ranking ahead of the common is permanently retired.
Common equity's residual claim on the bitcoin stack expands by that $100 of cleared senior obligation.
2. Dividend savings
Each retired share eliminates the $8/year dividend obligation (8% on the $100 par).
On a $58 purchase, that's a yield-to-cost of $8 / $58 ≈ 13.8%.
MSTR is effectively "earning" ~13.8% by retiring the obligation, versus the 8% stated coupon on par.
The deeper the discount, the more accretive
3. Accretion math
With positive mNAV, MSTR issues common via ATM at a premium to underlying BTC and uses the proceeds to retire STRK at 58 cents on the dollar of face.
The double arbitrage widens: sell equity above intrinsic, retire senior claims at a 42% discount.
The second part of the afternoon session is expected to resume soon. Earlier coverage, abbreviations, background information, pronoun direction in the QT.
We resume.
NC - [discussing the order, came in late]
Cannot hear the judge.
NC - [discussing the order, came in late]
Cannot hear the judge.
Apparently someone seated at bench behind the advocates phone has been flashing.
WARNING - WE CANNOT HEAR THE JUDGE
AL - hopefully, you should have the C's supplemental bundle as well
J - inaudible
AL - for tomorrow's purposes if the Tribunal could have that it would help me
WARNING - WE CANNOT HEAR THE JUDGE
AL - hopefully, you should have the C's supplemental bundle as well
J - inaudible
AL - for tomorrow's purposes if the Tribunal could have that it would help me
First #SCOTUS ruling today, Watson v. Republican National Committee.
ACB majority for herself, Chief, and 3 libs. Held: The federal election-day statutes do not prevent Mississippi from counting absentee ballots postmarked by election day but received up to five days thereafter; nothing in the federal election-day statutes requires ballots to be received by election day.
ACB majority for herself, Chief, and 3 libs. Held: The federal election-day statutes do not prevent Mississippi from counting absentee ballots postmarked by election day but received up to five days thereafter; nothing in the federal election-day statutes requires ballots to be received by election day.
Here's Watson ruling: supremecourt.gov/opinions/25pdf…
Alito dissent in Watson, joined by Thomas, Gorsuch, Kavanaugh (in large part).
How we have financed the vast majority of our value-add deals:
1. Buy all cash
2. Renovate all cash
3. Upon stabilization, refinance with a long term loan to pull out and return capital
(One core insight: The rich people who put $$ into private real estate deals often don't love to borrow at the rates demanded by bridge lenders. At 10% or 15% or whatever, they want to be *lenders* not *borrowers*.)
1. Buy all cash
2. Renovate all cash
3. Upon stabilization, refinance with a long term loan to pull out and return capital
(One core insight: The rich people who put $$ into private real estate deals often don't love to borrow at the rates demanded by bridge lenders. At 10% or 15% or whatever, they want to be *lenders* not *borrowers*.)
Another, more painful insight:
If you aim to be a long-term holder and you also want to avoid including mandatory capital call provisions in your operating agreements, then, whatever you think the safe amount of leverage is for your deals, it's probably lower than that.
If you aim to be a long-term holder and you also want to avoid including mandatory capital call provisions in your operating agreements, then, whatever you think the safe amount of leverage is for your deals, it's probably lower than that.
One benefit of this arrangement is that you are not at anyone's mercy.
If things go poorly (within the deal itself, with the rental market, with interest rates, etc.), you can always just stabilize, hold all-cash, accept the resulting yield, and wait for better days.
If things go poorly (within the deal itself, with the rental market, with interest rates, etc.), you can always just stabilize, hold all-cash, accept the resulting yield, and wait for better days.
HABITS I STOLE FROM WOMEN WHO ARE ALWAYS CALM AND UNBOTHERED.
1. Deep-clean your room.
1. Deep-clean your room.
1. Deep-clean your room from ceiling to floor in one sitting. The mess piling up isn't only physical, it's your inner state made visible, broadcasting back at you all day. Every avoided corner is a small thing you've decided you don't have energy for, and your brain logs every one. Tidy it all in one pass and you hand your mind permission to settle too.
2. Sit down and delete a thousand photos you'll never revisit. Your phone has become a hoarder's attic you carry everywhere. Blurry duplicates, screenshots, photos of people who hurt you, a life you've outgrown. Clearing it lightens something you didn't realize you were dragging around. You're deciding what's worth keeping and what you're finally allowed to let go of
Supreme Court declines to take up a challenge brought by New York healthcare workers who were fired after choosing not to get the Covid shot on religious grounds. Only Justices Thomas, Alito, and Gorsuch would have heard it.
Gorsuch: “I hope, too, that one day soon this Court will choose to settle the question so that other Americans seeking to vindicate their civil rights do not suffer the same fate as those now before us.”
Trump threatened a 100% tariff on any European country that imposes a tax on US tech firms. The tariff would take effect immediately.
It would override every trade deal those countries had already signed with US, including the EU tariff pact reached in May — The Guardian. 1/
It would override every trade deal those countries had already signed with US, including the EU tariff pact reached in May — The Guardian. 1/
Trump: Any country that imposes such a tax will be met with a 100% TARIFF on any and all Goods sent to the United States of America.
He said numerous European countries had discussed the tax and some were close to actually doing it. A larger EU-US trade war could follow. 2/
He said numerous European countries had discussed the tax and some were close to actually doing it. A larger EU-US trade war could follow. 2/
rance, Spain and Italy each levy a 3% digital services tax on large firms in their markets. Other EU states have done the same or proposed it.
The UK runs a separate 2% tax on social media platforms, large search engines and online marketplaces that profit from UK users. 3/
The UK runs a separate 2% tax on social media platforms, large search engines and online marketplaces that profit from UK users. 3/
Trade diplomats the world over tend not to be the best macroeconomist --
"It [Chinese state media] said Chinese companies were no longer as concerned about the European market because they now had options such as south-east Asia or the Middle East."
1/
"It [Chinese state media] said Chinese companies were no longer as concerned about the European market because they now had options such as south-east Asia or the Middle East."
1/
As the FT notes, China's surplus with SE Asia is a derivative of US tariffs/ low cost assembly of components in SE Asia ... basically it is a reflection of US demand
2/
2/
in the Chinese data, the US, ASEAN and the EU general bilateral surpluses equal to about three quarters of China's global surplus (with some Asian netting of HK)
-- So the real statement is that the US market is still an alternative to the EU market right now
3/
-- So the real statement is that the US market is still an alternative to the EU market right now
3/
