Mabi

Entrepreneur,
AI enthusiast
#foreverlearning

[The Kobeissi Letter @KobeissiLetter]

The relationship between stocks and bonds has completely flipped:

The 90-day correlation between the 10Y Treasury Yield and the S&P 500 is down to -0.48, its most negative reading since 1999.

This means that rising Treasury yields have recently been associated with weaker stock market performance, while falling yields have supported equities.

The current reading is even more negative than the 2022 bear market low of -0.42.

For context, before the 2020 pandemic, the correlation was positive for over a decade, with Treasury yields and equities often rising together as higher yields reflected stronger economic growth.

Currently, the negative correlation suggests investors are viewing higher yields less as a sign of economic strength and more as a result of inflation uncertainty and fiscal concerns.

All eyes are on the bond market.

[Leopold Stock Tracker @LeopoldATracker]

THE CITADEL GAME THAT CAUSED LEOPOLD ASCHENBRENNER’S FORCED SALE

> Citadel sees a young man in his early twenties managing a fund worth over $45 billion, nearly twice the size of Bill Ackman’s fund, which has been in operation for 23 years.

> The fund, in just 2 years, has grown from a modest $225 million to over $40 billion, with just 8 employees and no background in asset management. The fastest growing fund in history.

> Citadel sees a weakness in the portfolio ⟶ leverage.

> Citadel knows the thesis is right: AI infrastructure is where the real money is in AI. No matter who wins in the model race, the infrastructure wins.

> Now, this young man is holding billions in these assets.

> Then July 10 came after the $SKHY trade, followed by a major selloff across all AI stocks, with most down more than 30% in 2 weeks between July 10–20.

> Leopold Aschenbrenner’s short positions starts moving up against his direction, causing more losses (he was short on software stocks), with positions worth billions of dollars.

> July 24th, Leopold Aschenbrenner sends a letter to investors calling the selloff a buying opportunity and asks for more money, with that window to be open on August 1st, the same day of his wedding. With good news that the fund is up 439% through June YTD.

Perfect opportunity:

> July 27, Citadel Securities raised a false alarm, saying the Fed is gonna do a surprise rate hike.

> July 28, retail, out of fear, the market sells off after Citadel’s report & Leopold Aschenbrenner tries to raise & borrow more money to avoid a margin call for bleeding options positions. Both long and short positions bleeding.

> July 29, banks’ margin calls began. News people spread fear, the selling continues. Perfect opportunity.

> July 30, Jane Street and other market makers bid, but Citadel’s price is better. Citadel buys the majority of his portfolio for a massive discount.

> 3rd August, Citadel Securities is up more than $3 billion in just 2 days after buying Leopold Aschenbrenner’s Situational Awareness public book.

> Same Monday, 3rd August, Citadel Securities says, “the bull market drivers are intact.”

No AI bubble news, no market top, but buying the majority of the Leopold Aschenbrenner portfolio is the AI bottom and the start of the bull market.

Sad for those who didn’t buy the bottom. Because Citadel, as a market maker, won’t sell those positions anytime soon.

The game is rigged against retail and anyone that offers retail help.

[Porter Stansberry @porterstansb]

What Everyone Missed In Leo’s Blow-Up��

Leopold Aschenbrenner lost $30 billion (~67%) in a month.
The consensus post-mortem, from the Wall Street Journal to the replies on X, is that a young man used 4-to-1 leverage on concentrated positions and got carried out. While that is true, it does not convey any useful information. Leverage is certainly the reason Leopold lost so much, so quickly. But it is not the reason he lost. Leverage is merely a magnifying glass. It doesn’t pass judgement.
The reason the reason his fund was doomed was because he’s wrong. And no one, anywhere, has explained why.
On the morning of Thursday, July 30, before the opening bell, Situational Awareness LP sold its entire public stock portfolio — the long side and the short side together, roughly $16 billion of it — to Citadel in a single block trade.
Millennium Management and Jane Street bid for the assets. Ken Griffin and Citadel won.
That night, Aschenbrenner wrote to his limited partners. Net performance for the month, unaudited: down 67%. Net performance for the year: still up 80%.
"We let you down this month," he wrote. "We came closer to permanent capital impairment than is acceptable to us."
Six days earlier, on July 24, he had written a different letter. That one reported a 439% net return for the first half of 2026, described the selloff in artificial intelligence stocks as one of the best buying opportunities since early 2025, and invited his investors to wire more money starting August 1. It closed with a postscript: "At times we call out opportunities that seem like a particularly good time to add funds, if you have been waiting for one."
Assets that stood near $45 billion at the start of July finished the month around $10 billion, and roughly half of what remains is a single illiquid private stake in Anthropic.
Leopold is 25 years old. He graduated from Columbia at 19, as valedictorian. He worked at the FTX Future Fund from February to November of 2022, then joined OpenAI's Superalignment team, then was fired in April 2024. Two months after the firing he published a 165-page essay called "Situational Awareness: The Decade Ahead," raised $225 million from Patrick and John Collison, Nat Friedman and Daniel Gross, and started a hedge fund. He had never managed money before.
Situational Awareness was constructed to express only two ideas.
The first conviction: the physical build-out of artificial intelligence — the chips, the memory, the power, the data centers, the neoclouds — was the trade of the decade. The fund's disclosed long positions read like an inventory of the second derivative of the AI boom. Bloom Energy Corporation (NYSE: BE), fuel cells for data centers. Sandisk Corporation (NASDAQ: SNDK) and Micron Technology, Inc. (NASDAQ: MU), memory. CoreWeave, Inc. (NASDAQ: CRWV) and Nebius Group N.V. (NASDAQ: NBIS), rented compute. IREN Limited, Core Scientific, Applied Digital, Riot Platforms, CleanSpark, Bitfarms, Bitdeer — bitcoin miners converting their substations into AI compute.
The second conviction: application software was going to be destroyed by A.I. Not disrupted. Obliterated.
Leo explained why on Dwarkesh Patel's podcast, in June 2024:
"I'm so bearish on the wrapper companies because they're betting on stagnation. They're betting that you have these intermediate models and it takes so much schlep to integrate them. I'm really bearish because we're just going to sonic boom you. We're going to get the unhobblings. We're going to get the drop-in remote worker. Your stuff is not going to matter."
That was the whole thesis. Buy the compute. Short the stuff that runs on the compute.
By CNBC's reporting, the short leg included Adobe Inc. (NASDAQ: ADBE). A 13F does not disclose short stock. It does not disclose swaps. We only know about Adobe because reporters were told… but you can look at the tape and, when you do, it’s clear that Leo was short software in a major way.
Between the June 30 close and the July 29 close — the last session before the block trade cleared his shorts — the two sides of his portfolio did this.
The longs:
· Sandisk: down 55.32%
· Nebius: down 46.33%
· Bloom Energy: down 45.90%
· CoreWeave: down 38.90%
· Micron: down 35.98%
· IREN: down 35.91%
The shorts, over the same 20 sessions:
· Workday, Inc. (NASDAQ: WDAY): up 37.24%
· Adobe: up 28.49%
· Intuit Inc. (NASDAQ: INTU): up 27.64%
· Salesforce, Inc. (NYSE: CRM): up 20.25%
· Veeva Systems Inc. (NYSE: VEEV): up 17.15%
Over that same window the Invesco QQQ Trust fell 10.14% and the SPDR S&P 500 ETF Trust fell 2.32%. Nvidia — the supposed epicenter of the AI trade — fell 5.04%, and finished the full month of July up 0.33%.
This was not an AI crash.
The S&P 500 stayed near its record throughout. This was a violent rotation out of the leveraged, capital-hungry, second-derivative end of the AI complex and into the profitable, cash-generating, asset-light end of it. Which is to say: the market rotated out of exactly what he owned and into exactly what he was short.
Then there is Microsoft.
Microsoft Corporation (NASDAQ: MSFT) closed at $390.54 on Wednesday, July 29. It closed at $451.10 on Thursday, July 30. That is a gain of 15.51% in a single session on 110.2 million shares, against a July average of 37.1 million. Yes, Microsoft reported its fiscal fourth quarter after the close on July 29. But the results were nothing out of the ordinary. Revenue came in at $90.007 billion against a $87.62 billion consensus. That is a 2.7% beat. Earnings were $4.74 per share against $4.21. It was a good quarter. Not a historic one. A 2.7% revenue beat does not add roughly $450 billion of market value to the most widely owned company on earth in six and a half hours. Something else was in that tape.
And the answer is extremely important. Leo blew up quickly because of leverage. But he failed because he is simply wrong.
Aschenbrenner's software thesis rests on a single premise: that a company selling enterprise software is selling the work the software performs. If a model can perform that work, the company is worth nothing.
That premise is what a very smart 25-year-old engineer believes. It is not what anyone who has ever run a business believes.
Nobody buys Microsoft because Microsoft writes the best code. They buy Microsoft because Microsoft is the rail everything else runs on. Active Directory is where your employee identities live. Excel is where your board deck's numbers come from. Teams is where the compliance-recorded conversation happened. Azure holds a FedRAMP High authorization and Department of Defense Impact Level 5 clearance, which means a defense contractor cannot casually swap it out for something cheaper without re-clearing the entire stack with the government.
Veeva runs the customer relationship management and regulatory document systems of the pharmaceutical industry. Nineteen of the top 20 biopharmaceutical companies use Veeva's regulatory information management platform. Those systems are validated under GxP — the good-practice quality regulations that govern anything touching a drug — and 21 CFR Part 11, the Food and Drug Administration's rule for electronic records and signatures. Every major release is formally qualified. When an FDA inspector arrives, the audit trail in that system is the company's defense.
You cannot replace that with a model that is very good at writing code. You would have to re-validate a decade of regulated records, in front of a regulator, on a system with no track record, to save a fee that rounds to nothing in terms of the cost of building a new drug.
How small a fee? Veeva's licensing runs somewhere between roughly $1,800 and $6,600 per sales representative per year. A fully loaded pharmaceutical sales rep costs the employer between $134,000 and $219,000 a year. The software is 1% to 5% of the cost of the person using it.
Microsoft raised the price of a Microsoft 365 E3 seat from $36 to $39 per user per month on July 1 of this year, and E5 from $57 to $60. Add Copilot at $30 and a fully loaded E5 seat costs $1,080 a year. Against a knowledge worker costing $75,000 to $120,000 all-in, that is roughly 1% of the employee.
This is the part the compute maximalists cannot see. These companies are not selling labor. They are selling the rails on which labor runs, at a price so far below the value created that the buyer never bothers to negotiate hard, and with switching costs so high that the buyer could not leave even if he wanted to.
Do people try to leave? Constantly. And they almost always fail. (Ask me how I know!)
Panorama Consulting Group's tracked studies of enterprise resource planning replacements put average cost overruns at 189% across industries. Gartner projects that by 2027, more than 70% of recently implemented ERP initiatives will fail to fully meet their original business goals. Ripping out a core enterprise system is one of the most reliably disastrous things a large company can attempt, and it was true before anyone had heard of a transformer model.
The incumbents are not being disintermediated by artificial intelligence. They are selling it!
Microsoft passed 30 million paid Copilot seats in the June quarter, up from 15 million in January. Tech wizards like Leo hate copilot. Just like they hated Windows ’97. And everything else Microsoft has ever built. So what? Accenture alone bought 740,000 of them. Bayer, Johnson & Johnson, Mercedes-Benz and Roche have each deployed more than 90,000. Microsoft's commercial remaining performance obligation — contracted revenue not yet recognized, which is the closest thing software has to a railroad's signed freight contracts — stands at $678 billion, up 84% year over year!
Adobe's AI-first annual recurring revenue passed $500 million in the quarter ended May 2026 and tripled year over year. Salesforce's Agentforce went from $800 million of annual recurring revenue in the January quarter to $1.2 billion by April, up 205%. Veeva is giving its AI agents away free inside Vault CRM through 2030, which is the single most revealing data point in the set: Veeva does not need to monetize AI, because Veeva's moat is the validated record, not the intelligence applied to it.
Aschenbrenner thought AI would eat the applications. Instead the applications are selling AI as an upsell on top of a subscription the customer cannot afford to cancel – because it costs nothing compared to the value it delivers.
These software companies are computing toll booths: they’re what enterprises pay to implement compute. And, as compute gets cheaper, they will generate vastly more revenue, not less. The proof is sitting there in their earnings and cash flows: they’re riding on lower and lower cost of compute, which makes their business more and more efficient.
· Adobe: 36.6% operating margin, 35.6% return on invested capital, capital expenditure of $179 million on $23.8 billion of revenue — 0.75% — and $9.85 billion of free cash flow.
· Veeva: 28.7% operating margin, 68.5% return on invested capital, a 44.3% free cash flow margin, and effectively no capital expenditure at all.
· Salesforce: $41.5 billion of revenue, roughly $14.4 billion of free cash flow, capital expenditure of about 1.4% of revenue, and $72.4 billion of contracted backlog.
· Intuit: $18.8 billion of revenue, roughly $6.1 billion of free cash flow, $124 million of capital expenditure.
Veeva earns 68 cents a year on the dollar. And invests nothing it growing its business.
Adobe currently trades at about 11 times trailing earnings. Salesforce at about 13. Intuit at about 14. These are the multiples of a dying industry, applied to businesses converting a third to nearly half of every revenue dollar into free cash.
This enormous mispricing was manufactured by people who like Aschenbrenner, believed these businesses were doomed. But they aren’t.
And that’s not all.
Aschenbrenner assumed that because a technology is transformative, the capital that builds it will earn its cost.
There is no relationship between those two things. In fact, it’s more likely not to be true.
Leo’s own essay contains the tell: "Over the past year, the talk of the town has shifted from $10 billion compute clusters to $100 billion clusters to trillion-dollar clusters. Every six months another zero is added to the boardroom plans."
He wrote that as a bull case. But it isn’t. That is a recipe for a financial disaster.
https://t.co/gqUIUAzmYP, Inc. (NASDAQ: AMZN) spent $131.8 billion of capital expenditure in 2025 against $139.5 billion of operating cash flow. That is 94.5% of everything the business generated, poured back into the ground, in a single year. Its 2026 cap ex guidance is $220 billion.
Alphabet Inc. (NASDAQ: GOOGL) spent $91.4 billion in 2025, 55.5% of operating cash flow, and guides to $195 billion to $205 billion this year.
Meta Platforms, Inc. (NASDAQ: META) spent $72.2 billion, 62.4% of operating cash flow, and guides to $125 billion to $145 billion.
Microsoft spent $115.9 billion in the fiscal year that just ended, against $182.9 billion of operating cash flow. Capital expenditure was 34.9% of revenue, up from 18.1% two years earlier. Free cash flow fell to $67.0 billion from $74.1 billion in fiscal 2024, on revenue that grew by more than a third over the same span. Microsoft is running harder and generating less cash. That is what a huge capital cycle does even to the best business in the world.
Moody's projects hyperscaler capital expenditure of $785 billion in 2026 and close to $1 trillion in 2027, funded in part by roughly $175 billion of debt issuance this year. Where will the money come from…?
Oracle: fiscal 2026 capital expenditure of $55.7 billion, free cash flow of negative $23.7 billion, capital expenditure at 82.6% of revenue, long-term debt up from $76.3 billion to $124.7 billion, and $248 billion of future data-center lease obligations not yet on the balance sheet.
CoreWeave: $5.13 billion of 2025 revenue, $14.9 billion of capital expenditure, negative $7.25 billion of free cash flow, net debt at 8.1 times EBITDA, term loans at 11% to 15%, a weighted-average short-term borrowing rate of 12.3%, and a $1 billion private placement in April 2026 at 9.75%.
Meta's Hyperion campus in Louisiana is financed through a special purpose vehicle in which Blue Owl Capital holds 80% and Meta holds 20%, funded by $27.294 billion of senior secured notes at a 6.581% coupon maturing in 2049. The noteholders have no pledge on the physical data center. Their credit is Meta's promise to pay rent starting in 2029, plus a residual value guarantee. Twenty-seven billion dollars of debt, secured by a lease, sitting off the balance sheet.
And… like the EU’s finance minister explained two decades ago… “when it gets serious, you have to lie.”
Microsoft extended server useful lives from three years to four, then to six, adding about $3.7 billion to fiscal 2023 operating income. Alphabet did the same, adding about $3.0 billion. Amazon added about $2.5 billion in 2024. Meta added $2.59 billion in 2025. Oracle added $573 million. Every one of those is a non-cash increase in reported profit produced by an assumption about how long a chip stays useful. It’s a lie.
But not everyone is lying. Effective January 1, 2025, Amazon shortened the useful life of a subset of its servers and networking equipment from six years back to five, citing, in its own 10-K, "the increased pace of technology development, particularly in the area of artificial intelligence and machine learning." That cost it $1.4 billion of additional depreciation and $1.0 billion of net income.
Amazon is the operator with the longest and hardest-won experience running data centers at scale, and Amazon is the one telling you the hardware wears out faster than the schedules assume.
How could all of this spending possibly pay off?
Bain & Company's global technology report puts it at roughly $2 trillion of annual artificial intelligence revenue by 2030, and calculates that even if every dollar of on-premise IT budget shifted to the cloud and every dollar of AI productivity savings were reinvested, the industry would still be about $800 billion short. Sequoia Capital's David Cahn, who has been running the same arithmetic since 2023, has escalated his estimate from $200 billion to $600 billion to roughly $840 billion.
Against that: OpenAI's audited 2025 revenue was $13.07 billion, with an operating loss of $20.92 billion. Anthropic's 2025 revenue was $10 billion. Combined, $23 billion.
And of every dollar spent on Nvidia systems, roughly 72 to 75 cents is Nvidia's gross profit. Data center is now 88% of Nvidia's revenue. The margin is not in the build-out. The margin is in selling to the build-out.
What’s about to happen is obvious, because it has happened before.
Between 1865 and 1873 the United States built the most consequential physical network in its history and destroyed an enormous amount of capital doing it.
Track mileage went from 35,085 miles in 1865 to 52,922 in 1870 to 74,096 by 1875. Construction peaked at 7,439 miles laid in 1872. Railroad capital reached roughly $4.5 billion at a time when the entire banking system's capital was $720 million and the federal debt was $2.3 billion. In January 1870, of 896,596 shares traded on the New York Stock Exchange, 781,340 — 87% — were railroad shares. From 1870 to 1874, roughly 70% of all railroad securities issued in London were American. American rail bonds paid 6.5% when British consols paid far less, and European capital came for the yield.
Every argument you hear today was made then, too. The railroads will transform the country. Yep, they did compress distance and cost of transportation in a way that seemed impossible only a few years earlier. And it didn’t make any difference.
On September 18, 1873, Jay Cooke & Co. failed. Cooke had contracted to place $100 million of Northern Pacific 7.3% gold bonds, but sold less than $20 million. He ended up effectively owning 75% of the railroad he was supposed to be financing. And it failed. The New York Stock Exchange closed for ten days — the first closure in its history.
By 1876, 134 railroads were in default on $500 million of bonds out of roughly $2 billion outstanding. By 1877, 20% of American railroad track mileage was in receivership. European investors are estimated to have lost around $600 million between 1873 and 1879.
A very large fraction of the capital that built the American rail network was lost.
And where the roads survived, competition took the returns. Revenue per ton-mile fell from 1.88 cents in 1870 to 0.73 cents in 1900, a decline of about 61%. Rate wars on the New York-to-Chicago corridor drove the through rate from $1.88 down to 25 cents, then 20 cents, and no pooling agreement stabilized the worst of it until late 1885.
Every additional mile of track made the network more valuable to America and less valuable to the men who had paid for it.
The AI build-out will have the same problem – but it will be much, much worse. Compute will be a pure commodity.
Nobody disputes that the models are transformative. The problem is, that’s true of all of them.
Which of the second-derivative names Aschenbrenner owned has route control, like a monopoly railroad? Bitcoin miners with retrofitted substations? Rented compute resold at a spread? Memory, an industry that has never once earned its cost of capital through a full cycle? Those are not toll booths. Those are the Northern Pacific just before bankruptcy.
The railroads made a fortune – but not for their investors.
Adams Express Company was incorporated in 1854 with $1.2 million of capital. It did not own a single mile of track. It bought space on other men's trains and moved parcels, money and valuables on them. By 1866 its capital was $10 million and it was paying an 8% dividend quarterly. By 1875 its capital was $12 million. It paid an unbroken $8 per share annual dividend from 1869 forward — straight through the depression that put a fifth of American rail mileage into receivership, and straight through the next one in the 1890s.
American Express Company (NYSE: AXP) declared a $6 dividend in 1869, cut it to $3 in the depression year of 1877, restored it to $6 by late 1881, and held it there for the rest of the century. An 1888 board report showed ten-year net earnings of $26.24 million.
By 1890, the express companies were handling more than 115 million packages a year over 174,535 miles of railroad and steamship routes. And they didn’t own a single locomotive or a single boat.
Pullman's Palace Car Company was organized in 1867 with $1 million of capital. It did not own track either. It owned the sleeping cars and leased them to the railroads. Capital grew to $36 million by the early 1890s with nearly $25 million of accumulated surplus. Dividends ran 9.5% to 12% from 1867 to 1871 and 8% annually for decades after. In 1879, with 464 cars out on lease, it earned gross revenue of $2.2 million and net profit of almost $1 million.
Pullman put out $1 million of equity and earned $1 million a year on a network that cost other people billions and bankrupted a third of them.
Adams Express converted itself into a closed-end investment fund in 1929 and is still listed today as Adams Diversified Equity Fund (NYSE: ADX). The company that rented space on the railroads outlived almost all of them.
I’d bet a lot of money that Leo had never heard of any of these businesses.
But for people who are experienced in putting capital at risk, the pattern is not subtle or hard to understand. When an economy builds an expensive new network, the capital that builds the network earns a poor return because competition, obsolescence and overbuild strip it away. The businesses that ride on the network at near-zero incremental capital cost, and that own the customer relationship, the data or the standard, keep the profit.
I’ve seen this entire act before, during my career.
In the five years after the Telecommunications Act of 1996, carriers poured more than $500 billion into fiber, switches and wireless networks. By the early 2000s no more than 2% of North American long-haul capacity was in use. Global Crossing raised roughly $20 billion, built 100,000 miles of undersea fiber, filed for bankruptcy in January 2002, and saw its assets change hands for about $250 million — roughly 1.25 cents on the dollar of invested capital. WorldCom filed six months later, at the time the largest bankruptcy in American history.
Who got the value? Google, Amazon and Netflix, which built businesses on top of bandwidth that had become nearly free because somebody else had already gone bankrupt providing it. By 2018 and 2019, Google and Facebook were funding roughly four of every five dollars of new transatlantic cable investment — buying the rails only once the rails were cheap and only once they owned the applications that made the rails worth owning.
Leopold Aschenbrenner is not stupid. He is the opposite of stupid, which is part of the problem. He is a brilliant technologist who has never had to make a payroll, never had to explain to an auditor why the electronic records changed, never had to decide whether to spend eighteen months and $40 million ripping out a working system to save $200,000 a year in license fees.
He looked at enterprise software and saw code. A businessman looks at enterprise software and sees the thing his company cannot operate without for a single day, priced at 1% of the employee who uses it, backed by a validated audit trail he would have to rebuild from scratch in front of a regulator, and running on a contract he signed for three years.
An investor who has read a balance sheet from 1874 sees $220 billion of annual capital expenditure, an 8-times-levered reseller of rented compute borrowing at 12%, $27 billion of data-center debt hidden in a special purpose vehicle, and useful-life assumptions that the most experienced operator in the business is quietly walking back.
The kid believed the technology determines the return. But it never has.
It’s the capital structure that determines the returns: who controls the standards, who controls the customer, and who owns the data? Yes, the A.I. models will change everything. But that does not mean the people building the machines will be paid for it.
The money will be made where it was made in 1874 and again in 2004: by the toll booths riding on top of somebody else's ruinous capital expenditure.

[Codez @0xCodez]

IBM just released a 1-hour course on building agentic knowledge graphs from scratch:

• 00:00 - Introduction to knowledge graphs
• 05:35 - Building your first agentic graph
• 19:59 - Agentic memory powered by graphs
• 30:39 - Graphs for multi-agent orchestration

This 1-hour watch will replace 10 paid courses on agentic engineering.

Watch it today, then learn how to become a knowledge graph engineer in the article below.

[Lunar @LunarResearcher]

Google just released a free 1-hour AI engineering course.

How to build agents in 2026:

00:00 - Context engineering
10:00 - Building AI agents
31:52 - Agentic loops
43:22 - Building an MCP server
51:20 - Prompt engineering
59:10 - The future of AI development

Most people learn how to use AI.

This shows how to build the systems behind it.

Worth more than most $300 AI courses.

Bookmark and watch it today

Then read the article below

[Aakash Gupta @aakashgupta]

These guys literally dropped a masterclass on the basics of technical AI:

3:22 - How LLMs actually work
5:24 - What temperature is
7:05 - What context windows are
7:59 - Building a tool from scratch
10:28 - Why MCP exists
12:03 - What makes an agent an agent
12:43 - The power of skills
16:47 - What LangChain does
20:50 - How RAG kills hallucinations
23:36 - When you don't need a framework
25:26 - How routing works
28:34 - Why routers save millions
31:35 - Mock interview: LLM or ML?
35:50 - Why models hallucinate

[Neo Kim @systemdesignone]

If you want to become world-class with AI agents, then watch these 10 videos (save them):

[Khairallah AL-Awady @eng_khairallah1]

this is f*cking gold

How to build your first AI agent (Full guide)

if I had this a year ago, I would've shipped my first agent in a day instead of 2 weeks

in the right hands, this changes everything:

[Nicholas Mugalli @RealNickMugalli]

🇨🇳The founder of Moonshot the Chinese company that created Kimi K3 to overthrow Anthropic and OpenAI lead just gave a 40 minute masterclass on their progress.

The best explanation I’ve seen on Chinese cheap models. Free game!! Must watch…

[The Hair King @jamesbondwisdom]

washing your hair with an anti dandruff shampoo to regrow hair sounds like nonsense until you read the clinical study showing ketoconazole shampoo improved hair density and thickness comparably to 2% minoxidil over 6 months. then it sounds like a $12 shampoo doing a prescription drug's job

ketoconazole is an antifungal compound that was designed to treat dandruff. but researchers noticed something unexpected during trials. patients using it were growing thicker hair. not just cleaner hair. measurably thicker individual hair shafts and more of them

the mechanism is dual. first it eliminates malassezia fungus on the scalp which causes the chronic low-grade inflammation that accelerates follicle miniaturization. second, and this is the part nobody expected, ketoconazole appears to disrupt DHT activity at the follicle level through a mechanism that's still being fully mapped

- the study compared ketoconazole 2% shampoo used 2-4 times per week against 2% minoxidil. hair shaft diameter and the proportion of hairs in the growth phase improved comparably in both groups
- it works on contact meaning you don't need to leave it on for hours. 3-5 minutes on the scalp during a shower is sufficient for the antifungal and anti-DHT effects
- unlike minoxidil there is no dependency. if you stop using ketoconazole shampoo your hair doesn't fall out in response. minoxidil cessation causes accelerated shedding of every hair it was maintaining
- the anti-inflammatory effect reduces the perifollicular inflammation found in every biopsy of androgenetic alopecia. you're treating one of the root causes not just a symptom
- it's available over the counter as Nizoral in most countries. no prescription needed. no doctor visit required. $8-12 per bottle lasting 2-3 months
- you can stack it with every other treatment. use ketoconazole to clean the scalp and reduce DHT, then apply rosemary oil or peppermint oil after. the clean follicle environment amplifies everything applied afterward

the most effective hair loss shampoo ever studied sits on the drugstore shelf next to the Head and Shoulders. nobody buys it for hair growth because nobody told them it works for hair growth. because there's no profit margin in telling people a $12 shampoo does what a $40/month prescription does

[Joe Murgia @TheUfoJoe]

"Anyone who leaked a [UFO] report...could be prosecuted under the Espionage Act...life in prison, or death."

"Maybe we can reverse engineer this so we will have this incredible edge over the rest of the world."

🔥 Dr. Phil Had Me at The Bolender Memo 🔥

(Dr. Phil continued to kill it yesterday and this was better than the last one! Link to full video (21:52) is in the replies.)

"Our government has been lying to us for more than 80 years. Ask yourself why? Why do they not want you to know this is going on?"

(He starts out by showing that Google searches for "UFO" have allegedly nearly tripled since last Friday night, along with searches for UAP being up 400%. And "Dr. Phil UFO" is one of the fastest rising searches in the country. If all of that is true, it's a very good thing. Especially since he's been putting out some really good videos.

He mentions the latest release (#4) of UFO/UAP files from the Pentagon last Friday.)

"My team and I were granted early access, exclusive access to those documents before they went public."

"Our government has been lying to us, by omission AND by intentional misdirection, for 79 years."

"Look at these documents. You go back as far as 1947 where they have had clear information that there are objects not of this Earth, both technologically, metallurgically, performance-wise. But yet, that has been hidden from us. They've denied that, they've actually threatened people from talking about this. Threatened with careers, imprisonment, and some, with death, because they would consider it treason."

(He backs it up with documents. See below.)

Dr. Phil: "For 80 years, every time there was a legitimate UAP sighting and the U.S. government had a chance to get in there, they essentially told us, 'Nothing to see here. Move along.' To be clear, a UAP sighting is not proof of little green men or alien life. It means something happened, something was observed, something occurred for which we have no explanation. That's it.

"Now, let's talk about that for a second. Things happen, and we don't have any explanation for it. We don't have technology that explains that. There's something that is observed in the sky, going at a speed, stopping, making a sharp turn, reversing direction, accelerating, changing altitudes, and we don't have anything on this Earth that can do that. That's what's called unexplained.

"Now, do we know where it's from? How it does that? Well, if we did, it would be explained. But we don't. We go look at all of our secret weapons. We go look at what we know through intelligence, other countries have. And let me tell you, a lot of these (laughs) - they're not close calls. They are not close calls. And some of the things that have been observed, we damn sure didn't have anything like that in the 40s or the 50s or the 60s, and we still don't, now in 2020s.

"But at the very same time, our government was telling us, 'Nothing to see here: weather balloon, reflection off of an airplane, just a weather anomaly.' And, the government was simultaneously threatening its own people with criminal penalties if they ever disclosed UAP information.

"If somebody that was credible, that had seen this, spoke about it, they were threatened with all kinds of penalties. We were gaslighted. 'No big deal here. Probably weather balloons, misidentified aircraft.' And a jumpy public that watched, 'Close Encounters' one too many times, thinking, 'Look, what's really going on here?'

"Behind the curtain, the government was spending generations of time and resources protecting this information. Was it happening? Yes, it was happening! Do we have proof of it happening? Yes, we have proof of it happening. And what I mean by that is we have this on radar. We have credible, military pilots reporting it.

"We have aircraft that have guns, and when you open a gun and go live on a fighter, there's a camera that activates, so you have video of what the gun is shooting at. You have gun cameras. If they see one of these things in front of them, they open their weapons, in case they need them, and so it shoots video of what they're seeing.

"Now let's talk about some of the proof. JANAP 146 - Joint Army, Navy, Air Force publication 146 - made it a criminal offense for military personnel and commercial airline pilots to discuss UFO sightings outside official channels. The penalty, 'up to 10 years in prison and a $10,000 fine.'

"So, you might see some guy down on a lake, drinking beer, that talks about something he saw over the lake. Yeah, they don't mind that guy talking. But credible people? Trained observers with instrumentation? No. They say something, they're going to prison.

"Well, that took effect upon receipt. No hearings, no debate. The regulation says, 'All persons aware of the contents or existence...are governed by...espionage laws.' So not just the pilot who filed a report under JANAP 146, anyone who leaked a report. Radio operators, airline staff, anyone in the chain could be prosecuted under the Espionage Act.

"You can see this yourself. Declassified copies are online today, including on the NSA's website. Now what is the Espionage Act? That's the same law used to prosecute spies. Section 793: Up to 10 years in federal prison for every violation. Section 794: If the information reaches a foreign power, life in prison, or death.

"Imagine you're a TWA captain in 1955. You see something over the Pacific you just simply can't explain. You file your report like the regulation requires. And from that moment, talk to a newspaper, tell your own wife, puts you in the same legal category as a spy.

"Ask yourself why. Why are they so interested in muzzling all this conversation? Why do they not want you to know this is going on? Why are you not entitled to know what's happening in the air around you?

"Now, let's fast forward 20 years, 1971. Oliver Harry Turner was an Australian nuclear scientist and intelligence officer, head of the nuclear branch of Australia's Joint Intelligence Organization. He was asked to assess the American response to the growing UAP issue.

"If you're thinking, what does an Australian know about U.S. military secrets? Well, the possibility of life beyond this planet is bigger than any one country. Australia and the rest of the world has a legitimate interest in what the U.S. knows. And Australia is one of our closest intelligence allies. What we now call Five Eyes.

"The Five Eyes countries are the United States, Britain, Canada, Australia, and New Zealand. These countries have shared their most sensitive intelligence with each other since World War II. When a senior, Five Eyes nuclear intelligence officer writes a report about what the United States knows, well, that's serious.

"He was outside the American classification system. He had no career to lose. He pieced this together from official CIA, Air Force, Congressional, and Project Blue Book records. Now this report was written May 27, 1971. Original classification: Secret. Title: Scientific and Intelligence Aspects of the UFO Problem. Report declassified by the National Archives of Australia in 2023.

"On June 9th of 2026, whistleblower David Crusch (Yes, he said Crusch) stood on Capitol Hill and told the public to read pages seven through sixteen."

~

David Grusch: "There is a declassified 1971 Australian, formally-classified, Secret assessment that a couple years ago was put in the Australian National Archives. I encourage people to read page seven through sixteen, and that was the nuclear branch chief of the Australian government discussing the U.S. cover-up and the involvement of the CIA back in the 70s. And that's actually a little-known document that is publicly available."

~

Dr. Phil: "Now here's the kicker: foreign intelligence describing an American cover-up is now referenced in the files that our government is just now releasing. And here are six key findings in the Turner report.

"Number one, what Turner called the facade of ridicule. Turner documents that early Air Force intelligence concluded, 'Some of these objects, 'had flight characteristics' that could best be explained as having 'extraterrestrial origin.'

(The actual language says:

"The early analysis of UFO reports by USAF intelligence indicated that real phenomena were being reported which had flight characteristics so far in advance of U.S. aircraft that only as extra-terrestrial origin could be envisaged." )

Dr. Phil: "Instead of telling the public, the CIA and Air Force adopted a deliberate debunking policy.

"Now let that sit with you for a minute. Instead of telling the public, the CIA and Air Force adopted a deliberate debunking policy. We've got to get these people believing this isn't real. We've got to debunk this.

"January 1953, Turner's own words: 'By erecting a facade of ridicule, the U.S. hoped to allay public alarm, reduce the possibility of the Soviet taking advantage of UFO mass sightings...and act as a cover-up so the U.S. can develop vehicles that emulate UFO performances.'

"What's the point? Well, the point is, they were thinking, 'All right, let's keep this secret,' like we're the only ones seeing this, 'and maybe we can reverse engineer this so we will have this incredible edge over the rest of the world.'

"That's a great goal, I guess, if you can go from flying-prop planes or early jets to this incredible speed. If these are extraterrestrial, and the nearest galaxy is Andromeda, which it takes two and a half million years to get to, flying at the speed of light, we're pretty far from being able to do that.

"Today, in 2026, can we move at the speed of light? No. If we could, it would take two and a half million years to get to the next galaxy. We can't move at the speed of light, even now in 2026. But that was the goal. They'll find one of these and reverse engineer it."

(I don't know whether or not any black program has tech that can move at the speed of light and I doubt Dr. Phil knows, either. Someone should show him what Lacatski said about being able to reverse engineer some of this acquired (alleged non-human) tech but "not to its full extent.")

Dr. Phil: "Finding number two. He then talks about Project Sign. This was the U.S. Air Force's first official UFO investigation set up in late 1947. Its analysts reportedly concluded that extraterrestrial origin was the best explanation. Air Force Chief of Staff General Hoyt Vandenberg rejected that conclusion, and copies of the report were ordered destroyed. Destroyed!

"And per Turner, in February 1949, members of Project Sign, 'either volunteered to leave or were compelled to leave,' and they were replaced by people, 'willing to ridicule the concept of UFOs.'

"Think about this! We have sightings by legitimate observers with scientific instrumentation, and the people who are doing the observing are voluntarily leaving or compelled to leave, and replaced by people willing to ridicule the concept of UFOs.

"Finding number three, what I spoke about earlier: JANAP 146. Up to 10 years in prison and a $10,000 fine for discussing sightings outside official channels. And per the regulation's own text, Chapter One, Section 102, it covered not just military personnel, but U.S. and Canadian civilian and commercial pilots.

"That's legal force over airline pilots, arguably the most credible witnesses in the sky! What jurisdiction they had over Canadian pilots, I have no idea, but they listed 'em.

"Turner documents a meeting between military intelligence and airline pilots at the Roosevelt Hotel in Hollywood. At that meeting, pilots were, 'coerced' to keep their sightings out of public view and inside official channels. Am I overstating it to say that there's been a cover-up, that we're being lied to?

"Finding number four. But they missed the retirees. JANAP 146 only covered active service. Once you retired, you could talk. And three very senior men did so between 1953 and 1960. Admiral Roscoe Hillenkoetter, the first director of the CIA, Captain Edward Ruppelt, the man the Air Force put in charge of investigating UFOs, Major Dewey Fournet, the Pentagon's project officer on UFOs. Per Turner, all three publicly stated the U.S. government knew UFOs were extraterrestrial and was withholding the fact from the public! So when those three retired, they told the truth!

"Finding five. The government then silenced retirees. Per Turner, the revised regulation JANAP 146e made UAP disclosure by retirees an offense under the Espionage Act.

Then finding number six. 1969, 17 years, the Air Force ran a public-facing UFO investigation called Project Blue Book. If you wrote your congressman about a sighting, it went into the Blue Book. And in 1969, the Air Force shut down the Blue Book and told the country, 'We looked at more than 12,000 sightings. No problems. No national security threat. No need for the Blue Book.'

"But a memo from General Carroll Bolender, the Air Force general, said the reason Blue Book showed no national security threats was because any national security threats were gag ordered under JANAP 146 and were quote, 'not part of the Blue Book system.' The serious reports never stopped; they just moved out of public view.

"Now, I know that was a lot of information. You may need to listen back to that, but those are facts. Those are in the government's documents that have now been declassified!

"If the government has known for decades that unidentified objects are flying through our skies, and therefore, we may not be alone in the Universe, then the greatest revelation in American history has also been the target of a huge cover-up.

"One of the challenges of social media and TV news is taking something this complex and reducing it to sound bites. And you're getting bits and pieces from the media. That's why I invest so much time and energy giving you the real story. I don't want to tell you what you believe or don't believe, but I want to give you the information so you can make up your own mind.

"This information is in the files, we just haven't had access to the files. And then when we get the files, we're given the files without any context. You see a radar screen and you see a blip, and then it moves. Well, what do you have to compare it to? Is it moving fast? Too fast? Unexplainably fast? Without any context, how are we supposed to interpret that?

"Well, I'm digging in, I'm talking to experts, I'm finding out what the scale is. And what we're learning, is we don't have anything that'll move that fast. We don't have anything that'll turn that sharp. We don't have anything that will withstand those kind of G-forces."

[DON “Tobechukwu” Ade 👑 @Row_Haastrup]

This Chinese guy has successfully exposed all magicians 🤣🤣🤣🤣🤣

[T R U T H P O L E @Truthpole]

🚨 Bill Herrmann’s 1978 alien abduction testimony, recorded under hypnotic regression, details his encounter with a UFO and nonhuman beings. The session captures vivid accounts of missing time and direct interaction with the craft.

Herrmann described the event as involving advanced technology and entities that communicated with him during the experience. The case remains one of the more documented abductions from that era, often cited in UAP research.

The regression footage continues to surface in discussions of close encounters and government knowledge of the phenomenon.

[Rony @Ronycoder]

This 1-hour lecture from MIT professor Patrick Winston will teach you more about communication, public speaking, and presenting ideas than most people learn in years.

Bookmark it and give it 1 hour today, no matter what.

[Wimar.X @DefiWimar]

SpaceX valuation: $2.65 Trillion
Amazon valuation: $2.65 Trillion

SpaceX revenue: $18 Billion
Amazon revenue: $720 Billion

[Sumanth @Sumanth_077]

Train your own LLM from scratch!

A step-by-step repo that walks you through building and training a transformer model from scratch using PyTorch. From downloading training data all the way to generating text.

The architecture is built from the ground up following the original "Attention is All You Need" paper. MLP, single head attention, multi-head attention, transformer blocks, and the full transformer model - all coded and explained with detailed diagrams at each step.

Training data comes from The Pile - a diverse 825GB open-source dataset covering books, articles, code, websites, and more. The repo includes scripts to download it, preprocess and tokenize it using tiktoken, store it in HDF5 format, and feed it into training batches.

You can train a 13M parameter model on a single Colab T4 GPU. At 13M parameters the model starts generating proper grammar and coherent short sentences. For billion-parameter training you need at least an A100 or RTX 4090. The repo includes a full GPU compatibility table so you know exactly what's possible on your hardware.

Includes a complete SFT and RLHF guide as a separate notebook for taking your trained model further.

Key capabilities:

• End-to-end pipeline: data download → preprocessing → training → text generation
• Full transformer implementation from scratch with PyTorch
• Trains models from 13M to 2B+ parameters on a single GPU
• Training data from The Pile (825GB, 22 diverse datasets)
��� Tokenization via tiktoken (r50k_base)
• SFT and RLHF guide included

100% open source.

I've shared the link in the replies!

[Arnaud Bertrand @RnaudBertrand]

By the way, public service announcement: if you're one of the numerous people posting about Anthropic's dystopian ways and you're thinking about getting Claude to help you write that post... don't!

Another one of their terms is that you may not use Claude to do anything that "exposes [Anthropic to] reputational harms" 👇

And, if you do, under the - extremely unusual - clause 13 of their terms (https://t.co/z43rJNkvZu), you have PRE-AGREED, by using Anthropic (and accepted their terms), that the harm you've done is irreparable, that you won't oppose Anthropic injunction, and they don't need to prove actual damage.

They can simply go to a judge in a friendly jurisdiction (and of course, their terms precise that any dispute "will be resolved exclusively in the state or federal courts located in San Francisco, California") and:
a) file an injunction that shuts you down
b) make you pay for everything since under section 11 of their terms you agree to indemnify Anthropic for "any and all liabilities, claims, damages, expenses (including reasonable attorneys' fees and costs), and other losses arising out of or related to your breach or alleged breach of these Terms."

In other words, if you use Claude to help you talk shit about Anthropic publicly, their terms say you pay their lawyers to go after you and you've already pre-agreed you've lost the case.

Oh, and cherry on the cake: in the odd case the judge were like "are you crazy, this is insanely abusive, you Anthropic are the ones at fault here," according to their terms Anthropic's maximum liability is... $100.

[Jafar Najafov @JafarNajafov]

Google DeepMind engineers leaked an internal prompting method that nobody outside the company uses.

It's called "inverse prompting" and it feels illegal to know.

Instead of describing what you want, you describe what you don't want first.

My hallucination rate: 31% → 4%

Here's how it works:

[Arnaud Bertrand @RnaudBertrand]

As someone who partially grew up among European elite kids like him, this reminds me just how incredibly hollow some of them are.

For a quick background, I went to one of the poshest high schools in France (Janson de Sailly, for those who know) and, afterwards, to what was at the time - and probably still is - the most expensive undergraduate school in Europe (EHL in Lausanne, Switzerland).

Needless to say, many of my classmates were from unbelievably privileged backgrounds. Just in my classroom in Lausanne I had the son of a (very famous) Russian oligarch, the son of Italy's largest real estate developer and the son of Spain's largest real estate developer (funnily, the latter two were flat mates).

Another classmate of mine came from the richest family in Naples, Italy and - while we were at school - his father (known in Naples under the nickname "Il Sultano") got arrested for having bribed half of Naples's city council - which, if you know Naples, ought to tell you something.

These were the kids I was doing group projects on business ethics with (literally) 😅

Anyhow, my story, and probably my luck, was that - before going to high school in Paris - I was raised in very normal public schools in the South of France where my friends were anything but wealthy. Their parents were farmers and everyday workers.

Which means - and I'd come to realize this was very important in life - that it was easy for me to understand how big a mistake it is to see money as identity and meaning - and to confuse someone's net worth with their actual worth.

What really struck me at the time was the contrast with my "poor" classmates of earlier in my life. They couldn't define themselves by what they had - by definition - and this forced them to reach deeper for their identity: their skills, knowledge, humor, etc.

Rich kids can skip that entire process, and the tragedy is that most of them do: they reach for the readymade identity that money provides. I remember being incredibly frustrated by many of my classmates, like "ok, I get it, your dad is rich and you own a lot of nice things but who are YOU, what else is there?" The answer, more often than not, was nothing.

To be fair, there were exceptions. One of my classmates I was most impressed by came from one of Zurich's wealthiest families (which, if you know Zurich, means insanely wealthy) yet he was almost OCD in not showing he had money: driving the shittiest car imaginable, living in a small studio, etc. He was very intellectual, very contrarian, and clearly at war with the idea that his family's wealth ought to define who he was.

I only discovered who he actually was when I started my first company and he approached me to invest: to discuss the investment I went to one of his family homes, which it turned out was a literal palatial castle on the shores of Geneva lake. The guy had decided to live in a small rundown studio when he literally had a castle sitting empty a 5-min drive away.

THAT I was impressed by: it's easy to see that money isn't meaning when you don't have any. To see it when you have more than almost anyone - when everyone around you is organized around the opposite assumption - is much harder. But to actually live it, to choose the studio when you have the castle keys in your pockets - with no audience to applaud you for that - that shows real depth.

At the end of the day, I think, the real distinction isn't between rich and poor but between people who exist from the inside out and people who exist from the outside in.

Wealth just happens to make it incredibly easy to be the latter, to skip the work of becoming someone and settle for a borrowed identity that glitters from the outside but is hollow all the way through. A Potemkin village identity.

This is actually a real societal issue, and magnified by social media (with idiotic posts like this one 👇): the more "outside in" folks out there, the less people with genuine internal anchors, the more fragile everything becomes.

When you think about it, everything that genuinely matters in a society is built by people who think for themselves: they take the world in, pass it through something genuinely their own, and give back something that didn't exist before: an idea, a conviction, a stand.

Every reform, every invention, every act of moral courage in history came from someone with an internal anchor strong enough to resist the current. Remove those people and all you have left is the current.

This isn't new, by the way. Most ancient traditions warn against exactly this, from the Bible (the golden calf story) to Confucius, who built his entire ethics around the distinction between the exemplary person (the Junzi, 君子) - oriented around internal cultivation and righteousness - and the petty person (Xiaoren, 小人), oriented around profit and gain. The junzi builds himself from the inside, the xiaoren chases what's outside.

So please, do not make the mistake of being impressed by wealthy people flaunting their wealth. Don't focus on the glitter, focus on the hollowness it's trying to hide.

[Thinkaipath @thinkaipath]

🚨 BREAKING: Claude can now turn hours of research into minutes — for free.

Use these 9 prompts to analyze papers, connect ideas, and find hidden insights.

(🔖 Save this).