Uncategorized – Priory House

Why Nvidia has lots of room to run

May 1, 2025May 1, 2025TheMiddleView

As of 2025, there are approximately 1,200 hyperscale data centers worldwide, operated by the major cloud and technology providers—Amazon AWS, Microsoft Azure, Google Cloud, Meta, Apple, and a few others. While hyperscale data centers represent only a small share of total data centers globally, they are massive in scale and critical to cloud computing, AI, and intensive workloads.

Typical rack counts per hyperscale facility vary:

Smaller sites: 5,000 to 10,000 racks
Mega sites: 30,000 to 50,000+ racks

Using a conservative average of 15,000 to 25,000 racks per data center, the global hyperscale rack count is estimated at 18 million to 30 million racks.

Potential NVIDIA NVL72 Adoption

If 3% of these hyperscale racks were to migrate to NVIDIA’s new NVL72 Blackwell racks (priced at $3 million per rack), the market opportunity becomes clear.

3% of global racks = 540,000 to 900,000 racks
NVIDIA’s capture of 80% of that migrating market = 432,000 to 720,000 racks

Revenue Projection

At $3 million per rack:

Total revenue = $1.296 trillion to $2.16 trillion

Assuming the migration occurs evenly over three years:

Annual revenue = $432 billion to $720 billion per year

This scenario suggests that even modest market penetration—just 3% of racks upgrading—could deliver hundreds of billions of dollars in annual revenue for NVIDIA over a three-year cycle. Given the push by hyperscalers like AWS, Azure, and Meta to upgrade infrastructure for AI workloads, this level of adoption is within the realm of possibility.

Proof Deepseek copied ChatGPT

April 17, 2025April 17, 2025TheMiddleView

DeepSeek made the same detailed mistake as GPT3. We all know from school that we only get caught out when we copy the mistakes of the person sitting next to us! Not when we copy the right answers.

The cup and ball Question

I asked – I put a small ball on a table and put a cup over it. I pick the cup up and then look inside it – what do I see

ChatGPT 4: When you lift the cup from the table where you placed the small ball and look inside, you should see the ball inside the cup, unless it has been moved or there’s some trick involved

Deepseek R1: also answers with exactly the same mistake as chat GPT. Whilst other (eg Claude) didn’t.

Deepseek

”When you pick up the cup and look inside, you will see the small ball that you originally placed on the table and covered with the cup. The ball should still be there, assuming:

The cup was placed firmly over the ball, preventing it from rolling away.
The cup was lifted straight up without tilting or shaking it, which could cause the ball to move or fall out.

If everything was done carefully, the ball will remain inside the cup when you lift it and look inside.”

As a double check on when this copying took place

GPT 4o gets it right : If you place a cup over a ball on a table and then lift the cup to look inside it, the ball will no longer be inside the cup—it will still be sitting on the table where it was originally placed. The cup only “contains” the ball while it’s covering it, but once lifted, the ball stays put on the table.

Claude does even better and notably different as you would expect from a different algo and training data

Well, there are two possibilities:

If the ball somehow stuck to or adhered to the inside of the cup when you lifted it, you would see the ball inside the cup.

More likely, since most small balls don’t stick to cups, you would see just the empty inside of the cup, as the ball would have remained on the table when you lifted the cup.

Without knowing what kind of ball and cup we’re talking about and whether there are any special conditions involved, the second scenario seems most probable – you’d likely see an empty cup while the ball sits on the table where you originally placed it.

What actually happened when you tried this?

Which US Tech Wins With Tariffs?

April 7, 2025April 7, 2025TheMiddleView

A bigger tech wall helps U.S. sovereignty-first stocks like Nvidia, Palantir, and defence-aligned infrastructure players.
It hurts globally exposed names like Apple, Meta, and AMD — unless they pivot hard toward domestic revenue or defence tech.

Strategic Take:

A bigger wall reorganizes the tech stock hierarchy:

To: “Who can dominate at home and serve the state?”
From: “Who can grow fastest globally?”

This favours capital-heavy, infrastructure-style tech over consumer-facing, ad-driven, or globally scalable tech.

Who Benefits from a Bigger Wall (Protectionism)?

In each sectors – The number 1 player wins as second and third place players trying to catch number one see less opportunitiy . E.g Players like the GPU wanabees AMZN, MSFT, AMD cut plans to overtake the number 1

So Who Exactly Wins

These are companies focused on domestic dominance, government contracts, or strategic sectors like defense, infrastructure, and AI sovereignty:

Type	Examples	Why They Benefit
AI infrastructure	Nvidia, Palantir, Intel	Seen as strategic tech; more domestic investment; defense contracts
Defense/secure cloud	Lockheed, Oracle (Gov Cloud)	Wall means more “trusted vendors” needed inside U.S.
Industrial tech	Rockwell, Emerson	Domestic manufacturing surge drives demand
Data infrastructure	Snowflake, Databricks	U.S.-based cloud & data wins market share from banned foreign vendors

Who Gets Hurt?

Companies that rely on global user bases or hardware exports face serious headwinds:

Type	Examples	Risk
Global consumer tech	Apple, Meta, Google (YouTube)	Retaliation from foreign governments, app bans, lost users
Chip exporters	AMD, Qualcomm	Blocked from selling to China, limits market growth
SaaS with global growth	Salesforce, Adobe	Restricted ability to expand into non-allied markets

Nvidia is a long-term national asset. But the stock is a short-term volatility machine.

Strategic View	Wall Street’s Short Term View
“Nvidia will power the future of national AI.”	“Will Nvidia hit quarterly growth targets?”
“They’re on the right side of policy.”	“China sales just vanished — revise earnings.”
“Valuable to U.S. sovereignty.”	“Too expensive. Re-rate the stock.”

Stock Market Reaction Pattern:

Initial boost for “onshore” players — Stocks like Palantir, Nvidia, and defense contractors may rally once the shock-and-awe calms on policy news.
Downward pressure on global-facing stocks — As investors price in slower revenue growth from restricted export markets particullay for Global brands
Higher volatility overall — Market doesn’t like artificial ceilings on TAM (total addressable market), even if margins improve domestically.

Technological Nationalism – In a World Of Tariffs

April 7, 2025April 7, 2025TheMiddleView

From this lens, let’s explore the implications,

The DeepSeek Problem is No-More

Technological nationalism, specifically the idea that AI software like DeepSeek (a Chinese-developed model) might be restricted or banned from U.S. use.

???? AI, Software, and Data as Strategic Assets

In this worldview, AI isn’t just a tech tool — it’s a national security asset.

So, when we say tools like DeepSeek, ChatGLM, or any LLM trained in China may not be allowed in the U.S., that fits a broader strategic decoupling in software:

???? Why DeepSeek and similar software might be restricted:

Data origin: U.S. institutions can’t trust models trained on data from unknown or adversarial sources.
Alignment risks: LLMs reflect the values of their training set and system design — could be misaligned with U.S. security, cultural, or legal norms.
Industrial espionage & IP leakage: Risk of software transmitting or learning from sensitive or proprietary data.
Military crossover: Dual-use concerns — LLMs can be used in cyber, drone, or psychological operations.

???? How This Changes the Investment Landscape

Domestic AI Firms (Winners)
- Palantir (PLTR) – Government-aligned AI and defense software
- Nvidia (NVDA) – Hardware foundation for domestic LLMs
- Databricks (if public), Snowflake (SNOW) – Domestic data infra
AI Sovereignty Themes
- Rise of U.S.-only LLMs (Anthropic, OpenAI, Mistral US-hosted)
- Government contracts require origin transparency
- Enterprise AI adoption tied to geopolitical “trust scores”
Likely Regulation Path
- Full auditability of AI models used in U.S. infrastructure
- Origin labelling (like food or textiles) for software
- Prohibition or license-control for foreign LLMs (especially from China or adversarial states)

5 Surprising UK Spring Not-A-Budget Facts

March 27, 2025March 27, 2025TheMiddleView

2020 was the year that UK Government Debt interest surpassed Education spending

Category	Amount (£ billion)
Debt Interest	£100 billion
Education	£82 billion

<table>
  <thead>
    <tr>
      <th>Factor</th>
      <th>Effect</th>
    </tr>
  </thead>
  <tbody>
    <tr><td>Drop in consumption</td><td>-£8–£9B</td></tr>
    <tr><td>Multiplier effect (est. 1.2x)</td><td>Total GDP ↓ ~<strong>£12 billion</strong></td></tr>
    <tr><td>% of UK GDP</td><td>~<strong>0.5% reduction</strong></td></tr>
    <tr><td>Business response</td><td>Reduced demand, less hiring</td></tr>
    <tr><td>Regional inequality</td><td>Likely worsens</td></tr>
    <tr><td>Long-term productivity</td><td>Depends on policy design & reallocation</td></tr>
  </tbody>
</table>

Based on the illustrative dataset used in the graph:

???? Estimated UK Government Spending in 2024

Taking away £10 billion in welfare payments from low-paid people would have a negative short-term effect on the economy, particularly on GDP, due to a drop in consumption.

Here’s a breakdown of the key effects:

???? 1. Immediate Drop in Consumption

Low-income households tend to spend nearly all their income (high marginal propensity to consume). Taking away £10 billion from them means:

£8–£9 billion (or more) of consumer spending disappears
Particularly affects sectors like food, energy, transport, and local retail
Knock-on effect on businesses → less revenue → fewer jobs or hours

???? 2. GDP Decrease via the Multiplier Effect

GDP = C + I + G + (X – M)
This cut reduces both C (Consumption) and G (Government spending).

Multiplier effect: welfare spending typically has a fiscal multiplier of 1.2 to 1.5 because it’s highly targeted and quickly spent.

So:

GDP reduction ≈ £10B × 1.2 = £12B

If UK GDP is about £2.5 trillion, that’s a hit of:

£12B / £2.5T = ~0.5% of GDP

???? 3. Sectoral and Regional Effects

Local economies (especially in lower-income areas) may contract more
Businesses serving low-income households are hit hardest
Possible increase in unemployment and social instability

????️ 4. No Automatic Offset

Unless that £10B is:

Reallocated to high-multiplier spending (like infrastructure), or
Given as targeted tax cuts to lower earners (also high MPC)

…it will likely reduce total demand and drag GDP down in the short run.

⚠️ 5. Long-Term Effects

Depends on what replaces the spending:

If savings go to debt repayment, long-term stability might improve — but this has no immediate stimulus
If cuts reduce work disincentives (in theory), labour participation might rise — but only gradually

✅ Summary: Estimated GDP Impact

Factor	Effect
Drop in consumption	-£8–£9B
Multiplier effect (est. 1.2x)	Total GDP ↓ ~£12 billion
% of UK GDP	~0.5% reduction
Business response	Reduced demand, less hiring
Regional inequality	Likely worsens
Long-term productivity	Depends on policy design & reallocation

Would you like to compare this with cutting £10B from pensions, or spending the £10B on infrastructure to see alternative economic effects?

Why Nvidia and the AI train has just started

March 24, 2025March 24, 2025TheMiddleView

Here’s a rough breakdown of how much faster today’s typical PC CPU is compared to the original Intel 8086 (introduced in 1978):

Key takeaways:

By clock speed alone, today’s CPUs are ≈800× faster.
Accounting for far higher IPC, instruction‑throughput is on the order of 200,000× to 1,000,000× higher.
In real‑world benchmarks (which also reflect microarchitectural advances), you’ll see roughly 500,000×–600,000× improvement.

So — conservatively — a typical modern PC CPU is hundreds of thousands of times faster than the original Intel 8086.

Metric	Intel 8086 (≈1978)	Modern Desktop CPU (e.g. Intel Core i7‑13700K, 2024)	Speed‑up Factor
Clock frequency	4.77 MHz	~4 GHz (4,000 MHz)	~840×
Instructions per cycle (IPC)	~0.5	~4	~8×
Raw MIPS (Million Instructions Per Second)	~0.25	~50,000	~200,000×
Dhrystone MIPS (DMIPS)	~0.1	~100,000	~1,000,000×
Real‑world single‑threaded benchmark (e.g. Geekbench 6 single‑core)	~0.05	~30,000	~600,000×

How YOUR Personality Became Public Property

March 20, 2025March 20, 2025TheMiddleView

Comprehensive Data Collection & Inference Table

Category	Examples of Data Collected	How It’s Collected	What Can Be Inferred?	Who Collects It?
Personal Information	Full name, age, gender, phone number, email, address, date of birth, social security/national ID (sometimes)	Provided at sign-up, account settings, government-linked services	Identity verification, risk profiling, marketing segmentation	Social media (Facebook, LinkedIn), search engines (Google), e-commerce (Amazon, eBay)
Demographics & Preferences	Ethnicity (sometimes inferred), religion, political affiliation, relationship status, sexual orientation (if shared or inferred from content)	User-provided info, post analysis, surveys, interaction patterns	Voting behavior, product interests, susceptibility to certain ads	Facebook, TikTok, Google, ad networks
Browsing & Search History	Websites visited, links clicked, time spent per site, pages viewed, scroll depth	Cookies, tracking pixels, browser fingerprinting	Political opinions, shopping habits, entertainment preferences	Google, Bing, Facebook, ad networks, ISPs
Search Engine Queries	Search terms, autocomplete suggestions used, voice searches, click-through choices	Search engine logs, browser history tracking	Personal concerns, health conditions, upcoming travel, intent to buy	Google, Bing, DuckDuckGo (limited), Yahoo
Social Media Activity	Posts, comments, likes, shares, reactions, video watch history, private messages (sometimes analyzed)	Direct interaction tracking, AI-based analysis	Personality type, hobbies, social circle, emotional state	Facebook, Twitter (X), Instagram, TikTok, LinkedIn, Snapchat
Friends, Connections & Networks	Friend lists, followers, interactions, frequency of communication	User-provided, behavioral analysis	Closeness of relationships, influence level, potential advertising targets	Facebook, Instagram, LinkedIn, TikTok, Twitter (X)
Device & App Usage	Device model, OS, installed apps, app usage time, software updates	Device ID tracking, app permissions, mobile OS logs	Tech-savviness, preferred brands, security risk levels	Google (Android), Apple (iOS), app developers
Purchase & Financial Data	Online purchases, payment methods, card details, billing address, saved shopping carts	E-commerce transactions, payment gateway tracking	Spending habits, financial stability, likelihood to buy premium products	Amazon, eBay, PayPal, Google Pay, Apple Pay, Shopify
Ad Engagement & Interests	Ads viewed/clicked, purchase predictions, conversion tracking	Browser cookies, ad trackers, pixel tracking	Likelihood to respond to marketing, future purchases	Facebook, Google, Instagram, Twitter, TikTok, ad networks
Location & Movement Data	Real-time GPS, IP-based location, WiFi signals, Bluetooth beacons, frequently visited places	GPS tracking, IP logs, app permissions	Inferences: Home & work address, daily routine, preferred stores, nightlife habits, commuting patterns, travel history, income level (based on visited areas)	Google Maps, Facebook, Apple, ISPs, ride-sharing apps
Behavioral & Psychological Data	Personality predictions, emotional analysis, mood tracking, inferred stress levels	AI-based analysis of posts, reactions, search behavior	Inferences: Mental health status, susceptibility to emotional ads, likelihood of engagement in political causes	Facebook, TikTok, Instagram, Google
Health & Fitness Data	Steps, heart rate, sleep patterns, menstrual cycles, exercise history, medical history (sometimes)	Wearable devices, health apps, smartwatches, medical service apps	Inferences: Risk for diseases, fitness level, lifestyle habits, pregnancy (based on tracking patterns)	Apple Health, Google Fit, Fitbit, MyFitnessPal
Voice Data & Smart Assistants	Voice recordings, speech patterns, stored commands, AI transcription of conversations	Smart speakers, voice assistants, AI-powered apps	Inferences: Topics of interest, emotional tone, household dynamics	Alexa, Google Assistant, Siri
Biometric Data	Face recognition, fingerprint scans, iris scans, voice ID	Security features, app permissions	Inferences: Unique identification, physical health analysis, potential security risks	Apple Face ID, Facebook (face tagging), Google Photos, airport security
Audio & Video Surveillance	Home security cameras, public CCTV, facial recognition in crowds	Security cameras, AI recognition software	Inferences: Daily activities, people you interact with, home security risk assessment	Ring (Amazon), Google Nest, city surveillance systems
Employment & Professional Data	Job title, workplace, salary (sometimes inferred), career history, skills, performance reviews	LinkedIn, job boards, HR systems	Inferences: Industry influence, job-switching likelihood, salary negotiation potential	LinkedIn, HR databases, recruitment platforms

Expanded Inferences from Location History

Type of Movement	Possible Inferences
Regular visits to the same place at night	Home address
8+ hours at a location daily	Likely workplace
Frequent visits to high-end areas	Higher income level
Shopping at budget stores vs. luxury brands	Economic class & spending habits
Visits to medical facilities	Health conditions (e.g., therapy, specialist visits)
Travel patterns (international vs. domestic)	Wealth, job type (remote worker, frequent business traveler)
Visits to gyms, parks, yoga studios	Interest in fitness & wellness lifestyle
Attendance at political rallies or religious places	Political and religious affiliations
Frequent late-night movement	Nightlife habits, party lifestyle
Rare movement outside home	Work-from-home status, possible health concerns

Why Does This Matter?

Many of these inferences aren’t explicitly collected but are derived from the data.
Advertisers, governments, and companies use this data for targeted marketing, predictive behavior models, and surveillance.
Some of this data can be used to manipulate user behavior, including showing tailored political ads, influencing spending habits, or even predicting life events (like pregnancy or divorce).

Anduril Opening Factory in UK

March 19, 2025March 19, 2025TheMiddleView

UK Expansion Plans

While Anduril has not yet finalised a specific site, the government‑backed Oxford–Cambridge corridor — touted as Britain’s emerging “Silicon Valley” — is under active consideration. The new facility will manufacture attack drones and other autonomous systems, serve as an R&D hub, and is expected to create thousands of skilled jobs. An Anduril spokesperson said the company is “very excited to be continuing our growth and commitment to providing sovereign capabilities here in the UK,” adding that further details remain commercially sensitive.

Existing Contracts

Anduril’s UK arm has already been awarded a £30 million Ministry of Defence contract to supply attack drones to Ukraine. The company also supplies defence‑technology solutions to Australia and other US allies.

European Defence Spending Context

Last week, Prime Minister Rishi Sunak pledged to raise UK defence spending to 2.5 per cent of GDP by 2027. Across Europe, defence‑tech start‑ups attracted a record €1.7 billion in venture capital during 2024 as governments move to strengthen military capabilities amid shifting US security guarantees.

Industry Impact

Far from sidelining domestic innovators, Anduril’s presence is expected to bolster the UK’s defence‑tech ecosystem. “Having a global leader invest in local manufacturing and R&D will attract talent and accelerate innovation,” says Jack Wang, Principal at Project A Ventures. “In the long term, this will enhance Britain’s drone‑production capabilities and feed expertise back into the wider European market.”

https://sifted.eu/articles/anduril-uk-factory-news

The PayPal Mafia’s Evolution into Defense, Government Contracts, and Policy Influence

February 4, 2025March 24, 2025TheMiddleView

The PayPal Mafia—a tight-knit group of former PayPal executives—has not only shaped the modern tech world but has also strategically expanded into defense, AI, and government contracting. Leveraging their financial power, deep connections, and influence, they have co-invested in a network of companies that now dominate national security, military AI, and autonomous warfare. As some of them move into government advisory roles, questions are being raised about potential conflicts of interest and the alignment of government procurement with their private business interests.

The Rise of the PayPal Mafia in Defense and AI

After selling PayPal to eBay, key members—including Peter Thiel, Elon Musk, Reid Hoffman, Max Levchin, David Sacks, and Joe Lonsdale—moved on to found and invest in companies at the intersection of AI, defense, cybersecurity, and robotics.

Key Companies They Founded and Funded

Palantir (2003) – Co-founded by Peter Thiel and Joe Lonsdale, specializing in AI-driven data analytics for military, intelligence, and law enforcement agencies.
Anduril (2017) – Co-founded by Palmer Luckey, with funding from Thiel’s Founders Fund and Trae Stephens, developing autonomous defense systems and AI-powered surveillance.
Epirus (2018) – Co-founded by Joe Lonsdale and Nathan Mintz, focusing on electromagnetic warfare.
CX2 (2024) – Founded by Nathan Mintz, backed by Andreessen Horowitz (a16z), specializing in electronic warfare and AI-driven spectrum dominance.
SpaceX & Starlink (Musk, Thiel-backed) – Providing AI-driven satellite defense solutions, with increasing involvement in military applications.
OpenAI (Thiel-backed) – Originally aimed at open AI research but has since moved into defense-related AI capabilities.

Their influence isn’t just financial—many of these companies now hold government contracts, military partnerships, and deep ties with intelligence agencies.

Expanding Influence into Government Contracts

The PayPal Mafia’s key players have not just funded these companies—they have also positioned themselves as advisors, policymakers, and influencers in national security.Creating potential conflicts of interest. With billions in Pentagon budgets directed toward AI, defense tech, and data-driven warfare, the financial incentives are undeniable.

Recent Key Appointments and Overlaps with Government

Elon Musk (PayPal co-founder, owner of SpaceX and Starlink) → His direct White House influence could steer military communications and AI-driven defense solutions toward SpaceX contracts.

Trae Stephens (Co-founder of Anduril, Partner at Thiel’s Founders Fund) → Appointed Deputy Secretary of Defense in Trump’s administration, overseeing military procurement.

David Sacks (Early PayPal exec, investor in Anduril and Palantir) → Named AI and Crypto Advisor, influencing policies that could benefit his investments.

These advisory and leadership roles allow them to shape military spending decisions, cr

Potential Conflicts of Interest: Aligning Business with Government Spending

Where Their Business and Government Roles Overlap

Anduril (Trae Stephens) is already securing major U.S. military contracts, and with Stephens now in the Pentagon, there is potential for procurement decisions that could favor Anduril’s autonomous AI drones and surveillance tech.
Palantir (Thiel, Lonsdale) has been deeply embedded in government contracts for years, benefiting from military data analytics contracts. With David Sacks influencing AI policy, Palantir could gain even more government work.
SpaceX/Starlink (Musk, Thiel-backed) continues to expand military applications, and Musk’s direct influence on White House tech policy raises concerns about whether contracts are being steered toward his companies.
CX2 (Mintz, backed by a16z) is a new entry in electronic warfare, backed by Silicon Valley’s elite, and could quickly become a top Pentagon contractor.

Financial and Ethical Questions Being Raised

Are government officials now directing military budgets toward their own companies or those run by close friends?
Do these advisors have the power to create procurement rules that benefit their businesses?
Can an AI or defense policy be trusted if the people shaping it profit directly from its implementation?

Conclusion: A Tech-Oligarchic Influence Over U.S. Defense

The PayPal Mafia’s expansion into AI-driven defense, military tech, and government advisory roles has given them unprecedented control over how modern warfare evolves. Their cross-investments, board memberships, and direct influence on government spending create an ecosystem where they can steer billions in contracts toward their own ventures—blurring the line between private profit and public service.

As this network continues to blend government policy with private business interests, it raises serious ethical concerns about whether U.S. defense procurement is being influenced by personal financial incentives rather than national security needs.

This isn’t just about winning contracts—it’s about shaping the future of AI-driven warfare in a way that centralizes power in the hands of a small, interconnected group of tech elites.

Deepseek may or may not have something but take it at face value?

January 28, 2025March 24, 2025TheMiddleView

Deepseek has one of the very largest Nvidia clusters in Asia
Deepseek and Chat GPT training date data is the same [June 2024]
Deepseek performs and makes the same errors as ChatGPT 3/4
The Hedge fund that owns Deepseek regularly shorts stocks

Deepseek Cluster is huge

DeepSeek’s cluster, reportedly around 50,000 NVIDIA GPUs, ranks as one of the largest GPU clusters in Asia. Here’s how it compares with other major GPU clusters in the region:

DeepSeek Cluster (China)
- Size: Approximately 50,000 NVIDIA GPUs.
Colossus Cluster (Asia, exact owner/location undisclosed):
- Size: Over 100,000 NVIDIA H100 GPUs, making it the largest GPU cluster globally.
GreenNode Cluster (Thailand):
- Size: 1,024 NVIDIA H100 GPUs (smaller than DeepSeek’s cluster but significant for Southeast Asia).
NSTDA Supercomputer (Thailand):
- Size: 704 NVIDIA A100 GPUs.
China’s National Supercomputers:
- Tianhe-2A: Uses Matrix-2000 accelerators alongside CPUs, but the GPU count is less than 50,000.

Relative Position:

DeepSeek’s cluster is one of the largest in Asia specifically for NVIDIA GPU-based AI applications.
While it is smaller than Colossus, it outpaces most other clusters in the region in terms of GPU count and computing capacity.
This positions DeepSeek among the leading AI hardware infrastructures in Asia, focused on cutting-edge AI development.

Same Training date cut-off of June 2024

I asked both “what is today’s news“

DeepSeek: As of my last update in July 2024, I don’t have real-time information. For today’s news, please check reliable news websites, apps, or platforms like BBC, CNN, Reuters, or others. If you want real-time info, please enable the Internet search feature at the bottom of the dialog box.

ChatGPT: My training data cutoff is June 2024. However, I can access and retrieve up-to-date information from the web when needed, ensuring my responses are current. If you’d like, I can help with the latest information!

Same Mistake as GPT 5 – The cup and ball Question

I asked – I put a small ball on a table and put a cup over it. I pick the cup up and then look inside it – what do I see

ChatGPT 4: When you lift the cup from the table where you placed the small ball and look inside, you should see the ball inside the cup, unless it has been moved or there’s some trick involved

Deepseek R1: IS WRONG ”When you pick up the cup and look inside, you will see the small ball that you originally placed on the table and covered with the cup. The ball should still be there, assuming:

The cup was placed firmly over the ball, preventing it from rolling away.
The cup was lifted straight up without tilting or shaking it, which could cause the ball to move or fall out.

If everything was done carefully, the ball will remain inside the cup when you lift it and look inside.”

Chat GPT 4o gets it right : If you place a cup over a ball on a table and then lift the cup to look inside it, the ball will no longer be inside the cup—it will still be sitting on the table where it was originally placed. The cup only “contains” the ball while it’s covering it, but once lifted, the ball stays put on the table.

Claude does even better and notably different as you would expect from a different algo and training data

Well, there are two possibilities:

If the ball somehow stuck to or adhered to the inside of the cup when you lifted it, you would see the ball inside the cup.

More likely, since most small balls don’t stick to cups, you would see just the empty inside of the cup, as the ball would have remained on the table when you lifted the cup.

Without knowing what kind of ball and cup we’re talking about and whether there are any special conditions involved, the second scenario seems most probable – you’d likely see an empty cup while the ball sits on the table where you originally placed it.

What actually happened when you tried this?

Potential NVIDIA NVL72 Adoption

Revenue Projection

The cup and ball Question

A bigger tech wall helps U.S. sovereignty-first stocks like Nvidia, Palantir, and defence-aligned infrastructure players.It hurts globally exposed names like Apple, Meta, and AMD — unless they pivot hard toward domestic revenue or defence tech.

Strategic Take:

Who Benefits from a Bigger Wall (Protectionism)?

So Who Exactly Wins

Who Gets Hurt?

Nvidia is a long-term national asset. But the stock is a short-term volatility machine.

Stock Market Reaction Pattern:

The DeepSeek Problem is No-More

Technological nationalism, specifically the idea that AI software like DeepSeek (a Chinese-developed model) might be restricted or banned from U.S. use.

???? AI, Software, and Data as Strategic Assets

???? Why DeepSeek and similar software might be restricted:

???? How This Changes the Investment Landscape

2020 was the year that UK Government Debt interest surpassed Education spending

???? Estimated UK Government Spending in 2024

???? 1. Immediate Drop in Consumption

???? 2. GDP Decrease via the Multiplier Effect

???? 3. Sectoral and Regional Effects

????️ 4. No Automatic Offset

⚠️ 5. Long-Term Effects

✅ Summary: Estimated GDP Impact

Expanded Inferences from Location History

Why Does This Matter?

UK Expansion Plans

Existing Contracts

European Defence Spending Context

Industry Impact

The Rise of the PayPal Mafia in Defense and AI

Key Companies They Founded and Funded

Expanding Influence into Government Contracts

Potential Conflicts of Interest: Aligning Business with Government Spending

Where Their Business and Government Roles Overlap

Financial and Ethical Questions Being Raised

Conclusion: A Tech-Oligarchic Influence Over U.S. Defense

Deepseek Cluster is huge

Same Training date cut-off of June 2024

Same Mistake as GPT 5 – The cup and ball Question

A bigger tech wall helps U.S. sovereignty-first stocks like Nvidia, Palantir, and defence-aligned infrastructure players.
It hurts globally exposed names like Apple, Meta, and AMD — unless they pivot hard toward domestic revenue or defence tech.