10 BOOKS SERIOUS AI RESEARCHERS ACTUALLY RECOMMEND (NOT THE ONES EVERYONE POSTS)
Every AI reading list says the same five names. The people actually building these systems read deeper than that. Here's the shelf they point to when nobody's performing for an audience.
1. Probability Theory: The Logic of Science - E.T. Jaynes
The book researchers quietly call life-changing. Jaynes reframes probability not as gambling odds but as the mathematics of reasoning under uncertainty, which is exactly what every modern model is doing. Dense, opinionated, and the closest thing the field has to a sacred text. Almost nobody outside the work has heard of it.
2. Information Theory, Inference, and Learning Algorithms - David MacKay
The book that unites information theory and machine learning in one place, written by a Cambridge physicist who made it genuinely fun. Free online, full of puzzles, and on the shelf of nearly every researcher who actually understands why their models compress and predict the way they do.
3. Reinforcement Learning: An Introduction - Sutton and Barto
The foundation under everything from AlphaGo to how modern models get fine-tuned with human feedback. Researchers don't recommend it because it's trendy. They recommend it because the ideas in it keep turning out to be the ideas that matter, decades later. Also free.
4. The Book of Why - Judea Pearl
A Turing Award winner's argument that today's AI is stuck because it confuses correlation with causation, and a map of what real reasoning would require. The book that names the exact ceiling current systems keep hitting. Researchers cite it constantly. The public reads past it.
5. Vision - David Marr
A neuroscientist's framework for how any system, brain or machine, processes information, written before deep learning existed and somehow predicting the questions it would face. The "levels of analysis" idea in here quietly shapes how serious people think about what a model is even doing.
6. Gödel, Escher, Bach - Douglas Hofstadter
The cult book about how meaning and selfhood emerge from systems following simple rules. It won a Pulitzer and then got name-dropped to death, but almost nobody finishes it. The ones who do think differently about intelligence forever. The real one, not the summary.
7. Metaphors We Live By - Lakoff and Johnson
The argument that human thought runs on metaphor, not cold logic, and that you can't build a mind on first-order logic alone. Researchers working on why language models grasp meaning the strange way they do keep circling back to this one. A genuine left-field pick.
8. The Society of Mind - Marvin Minsky
One of AI's founding figures arguing that intelligence isn't one thing, it's a swarm of dumb little processes working together. Written as hundreds of one-page ideas. Out of fashion for years, now looking prophetic in the age of multi-agent systems. Insider catnip.
9. How to Solve It - George Pólya
A 1945 book on mathematical problem-solving that quietly shaped how a generation of researchers think about breaking down hard problems, and that keeps surfacing in papers on how to make models reason. The bridge between human heuristics and machine reasoning.
10. The Mathematical Theory of Communication - Claude Shannon
The original paper that invented information theory and, with it, the entire conceptual ground that machine learning stands on. Short, brutal, and foundational. Researchers revere Shannon the way physicists revere Newton. Most reading lists skip the source and quote the descendants.
The popular books tell you what AI might do. These tell you how the people building it actually think. The difference is the whole point.
The day I stopped trying to predict the market was the day my trading improved.
Mark Douglas taught that successful trading isn’t about being right.
It’s about managing risk, thinking in probabilities, and executing your edge without fear.
re: cutting losses...
Psychology: Douglas + Kiev will get u aligned (article)
In practice: this is laughably silly but the best "how to improve" and become automatic...is reps.
1/5
muscle memory, habit loops, brain grooving, these are real for cutting losses
reps are what create a habit, and cutting losses is unnatural...u have to imagine urself re-wiring ur brain in the other direction
u can buy and sell 1 lots all day either while learning to use stops, or executing manually
people "freeze up" all the time in a fast tape, or a crisis, or when u didnt cut a loss...beyond common, blanking on the term for it
after someone misses their 10% stop, they disengage and just watch, mesmerized, as it goes down 50%
when plan A fails, that was it in their mind. no backup plan, just watching a loss balloon in size
2/5
if ur muscle memory is sharp u might snap back to reality
but so many accounts have been blown by people who said "I'll be fine"
then traded too much size in a fast market, and the size of the unrealized dollar loss on the P&L freezes them like a deer in headlights
and they have maybe five lifetime "reps" in cutting a loss, this isn't gonna be #6
there is a reason even the best traders use stops
they know that you cannot trust yourself, your mind, your emotions, to "be fine"
3/5
but the only way you acquire this as a skill, is to start small and slowly trade larger and larger, because you will hit mental resistance as dollar P&L and % losses grow larger
seeing a unrealized $10k or $100k on paper, ur brain will do the "I could buy a car with that" and suddenly you're in another galaxy of ur mind.
little shit, like its 10 AM and you fucked up, u locked in a $20k realized loss.
you're done for the day, but that shit is staring at you.
suddenly its 4 PM and you lost $60k on a daily limit of $10k
you might cut a trade loss well, but then not know how to maintain discipline on these types of loss limits
4/5
the first 5-7 years are probably just variations of blowing up and breaking discipline over and over, in new ways and at larger sizes
talk to ANY trader with 10-20 years experience, they'll echo everything here
the difference is in who has the humility to recognize: I cannot trust my mind or my emotions
and they begin automating and mitigating and building a system to protect themselves from themselves
The guys who say "I have a strong mind" stay at day one, make no changes, and just blow up for the next ten years instead of acknowledging that markets are custom-designed to activate every single human thinking error
It's not that you're weak, the human mind is weak
But I've learned that things like risk management, cannot be taught or written about or spoken on
And that the guys who blow up every day for ten years, they don't want to change
5/5
this is one gigantic data center in your brain that will ruin your life if not automated 5000 different ways and disciplined -- little shit, the inability to keep a promise you make to yourself is everpresent in life.
"I'll eat a spoonful of ice cream" whole tub gone
drugs, drinking, this same brain area controls all these same impulses
This is why they say the market is a mirror, and why your P&L shows you who you are
If you can't cut a loss, the rest of your life will reflect that
--
again, best way to improve is start with 1 lots
once you have those down, you size up to a level where you "feel" a loss, you don't feel a 1-lot
your mind and emotions will activate in new ways as size gets larger, its a process
if you want to make $10M, at one point, you'll have a day where you eat a $1M realized loss
a huge part of all this is staying out of your own way
to each their own, my best advice is automate absolutely everything...all positions have automatic stops, period.
the risk management articles my pinned tweet are thorough as hell
but start here with Douglas and Kiev - they're aces
good luck trading
USA (Use Stops Always)
ex-CEO of Goldman Sachs just revealed his entire portfolio - 98% equities, 90% single stocks - and he trades every single day from his iPad
"the difference between somebody who's really really good and somebody who can't make it is not that great"
"I thought SpaceX was overpriced at $100 billion - they're now proposing $1.75 trillion"
"I've known people in high office - after they finish speaking they say how did I do - they want affirmation - people are a lot more insecure than you think"
bookmark & watch the full podcast ↓
Bank of America: China's AIDC investment could reach USD 327 billion by 2030; focus on the dual themes of 'power grid + materials'
Bank of America Securities significantly raised its forecast for China's AI data center (AIDC) capital expenditure to $327 billion by 2030 and systematically outlined structural investment opportunities for traditional sectors—including copper, PCB materials, optical fiber, and transformers—to integrate into the AI value chain. Related stocks received a series of buy ratings.
According to reports, China’s AIDC capital expenditure over the next five years could reach approximately RMB 2 trillion.
Per Wind Trading Desk sources, Bank of America forecasts that China’s AI-related capital expenditure will grow from approximately USD 140 billion in 2026 to USD 327 billion by 2030, representing a compound annual growth rate (CAGR) of 24%, accounting for roughly 20% of global AI capital expenditure by then. Meanwhile, Bank of America’s global team simultaneously revised upward its 2030 global AI capital expenditure forecast to over USD 1.7 trillion, a substantial increase from USD 260 billion in 2025.
Global data center installed capacity is projected to expand from approximately 100 GW today to nearly 300 GW by 2030. Rack power density is expected to surge from the traditional server range of 10–15 kW to 100–120 kW under current NVIDIA platform roadmaps, with potential to exceed 1 MW in next-generation systems. This shift is driving upstream material and power equipment demand into a structurally ascending trajectory.
Power Demand: China’s data centers will consume 318 terawatt-hours (TWh) of electricity by 2030
According to data from the International Energy Agency (IEA), global data center electricity consumption in 2025 will approach 500 TWh, accounting for approximately 1.6% of total global electricity use. Bank of America estimates this figure will grow at a compound annual rate of 22%, reaching 1,208 TWh by 2030—about 3.7% of global electricity consumption.
In China, data center installed capacity is projected to rise from 29 GW in 2025 to 77 GW by 2030, with corresponding electricity consumption increasing from 121 TWh to 318 TWh—accounting for roughly 2.5% of national electricity use. Bank of America also notes that due to chip-related constraints prompting some internet companies to relocate computing capacity to Southeast Asia, actual domestic data center power consumption in China may lag behind the true growth rate of AI demand.
Three key factors are driving the sharp increase in power demand: rapid growth in AIDC workloads; higher per-chip power consumption as GPUs replace CPUs; and significant system-level power expansion required to support rack power densities rising from 10–15 kW to 100–120 kW or higher.
AI Materials: Structural Supply Tightness Across Five Key Categories
Regarding copper, Bank of America forecasts that China’s data center-related copper demand will grow from 341 kilotonnes (kt) in 2025 to 1,190 kt by 2030—a CAGR of 28%—raising its share of total Chinese copper demand from 2.1% to 6.4%. This incremental demand will primarily stem from data center operations (650 kt), grid expansion (504 kt), and power plant construction (36 kt). Addressing market concerns about 'optical fiber replacing copper,' Bank of America argues that copper remains irreplaceable in power transmission, short-distance interconnects, and intra-server connections. Coupled with an anticipated global copper supply deficit of 491–754 kt between 2026 and 2027, the fundamental support for copper prices remains robust.
In PCB materials, structural overcapacity persists in low-end copper foil. However, AI servers are driving an increase in PCB layer counts and higher specifications for high-frequency signal transmission, sharply expanding demand for high-end copper foil. High barriers to capacity conversion, equipment bottlenecks, and customer certification cycles typically exceeding one year make it difficult to close the supply-demand gap in the near term, supporting sustained pricing and profitability for high-end products.
Supply of high-end electronic glass fiber (specialty yarns with low Dk and low CTE) has historically been dominated by a few Japanese manufacturers such as Nittobo (3110 JP). However, Chinese producers have gradually achieved breakthroughs after years of R&D. Currently, there are five qualified domestic suppliers, with Taishan Fiberglass, a subsidiary of Sinoma Science & Technology, leading the market. Bank of America has assigned a Buy rating to Sinoma Science & Technology, forecasting its specialty fiberglass capacity to expand from the current 24 million meters to 94 million meters by 2027.
Fiber optic demand is accelerating its shift from traditional telecom toward AIDC applications. While overall capacity remains sufficient, a structural supply bottleneck exists in preforms—the key upstream raw material for high-end fiber optics—due to their long capacity expansion cycle and high technical barriers, thereby underpinning fiber optic prices. Bank of America has assigned a Buy rating to Jiangsu Zhongtian Technology, expecting its compound annual EPS growth rate from 2026 to 2027 to reach approximately 75%.
In magnetic materials, tungsten, and uranium: high-performance neodymium-iron-boron (NdFeB) permanent magnets benefit from dual drivers—liquid cooling systems in AIDC infrastructure and humanoid robotics—maintaining structurally tight supply conditions. Tungsten has entered the AI value chain due to growing demand for high-density PCB drilling, with China’s resource controls keeping supply inelastic. Uranium is viewed as a core strategic resource for scalable, zero-carbon, and stable baseload power in the AI computing era. Bank of America’s global commodities team forecasts uranium prices to rise by 47% and 29% year-over-year in 2026 and 2027, respectively, driven by a structural supply deficit of 2%–7% annually and demand growing at a compound annual rate of approximately 4%.
AI Power Supply: China Holds Unique Competitive Advantages
Bank of America emphasizes that China possesses multiple structural advantages in AI power supply: commercial electricity tariffs are 30%–60% lower than those in the U.S. and EU; available reserve capacity margin stands at approximately 30%, higher than the U.S.’s sub-25% and the EU’s ~15%; transmission and distribution infrastructure has an average age of less than 20 years (compared to over 40 years in the U.S. and Europe), ensuring greater grid stability; and China operates a comprehensive network of 46 ultra-high-voltage transmission corridors alongside a robust power equipment manufacturing ecosystem. Additionally, China has announced plans to expand nuclear power installed capacity from 62 GW by end-2025 to approximately 110 GW by 2030 under its 15th Five-Year Plan.
Opportunity 1: Transformers. Bank of America forecasts China’s transformer exports to grow by 30% year-over-year in 2026, with domestic grid investment rising by 12% to approximately RMB 715 billion. The global transformer shortage is expected to persist through at least 2029, with lead times for high-voltage transformers stretching up to three years. China is effectively filling overseas capacity gaps thanks to its robust supply chain.
Opportunity 2: Gas turbines. Bank of America’s Global Industrials team projects annual global gas turbine orders of approximately 120 GW from 2026 to 2028. Current new order lead times extend to 3–6 years, creating a window for Chinese manufacturers to enter the market on the basis of competitive pricing and shorter delivery timelines (as short as 13 months). Dongfang Electric is the only Chinese company capable of exporting medium- and large-scale gas turbines. Its G50 50MW units have already secured sales of 10 units to a Canadian data center client and have been exported to Kazakhstan and Indonesia. Management plans to scale export capacity to 23 units by end-2027 and 45 units by end-2029.
Opportunity 3: Diesel engines. Chinese manufacturers of large-bore diesel engines have obtained UL and EPA certifications in the U.S., becoming the first to enter the North American AIDC backup power market. Bank of America forecasts demand for AIDC backup diesel engines in China to reach 8,500 units in 2026, with supply tightness expected to ease in 2027 as capacity expands.
Opportunity 4: Energy storage systems. Bank of America forecasts the global BESS (Battery Energy Storage System) market to grow at a compound annual rate of approximately 23% from 2025 to 2030, with AIDC-related BESS growing at about 27%. By 2030, annual new AIDC BESS installations globally are projected to reach 70 GWh, accounting for roughly 8% of total global new BESS installations.
Opportunity Five: Power Supply Systems. Bank of America forecasts that China’s AIDC power supply systems market (comprising UPS, HVDC, and SST) will grow at a compound annual growth rate (CAGR) of approximately 25% from 2025 to 2030. Nvidia is actively driving its supply chain toward an 800V DC high-voltage direct current (HVDC) architecture along its hardware roadmap to address continuously rising rack power densities. The average selling price (ASP) of high-capacity power supply systems is significantly higher than that of traditional units, and the combination of high R&D barriers and customization requirements creates a strong competitive moat.
Liquid Cooling: Penetration Rate to Jump from 30% to 70%
Liquid cooling is the fastest-growing segment within China’s data center cooling market. As rack power densities continue to exceed the practical limits of air cooling (approximately 40 kW per rack), coupled with increasingly stringent domestic energy efficiency regulations, the adoption of liquid cooling is accelerating. Liquid cooling offers thermal conductivity 20 to 50 times higher than air cooling, enabling power usage effectiveness (PUE) to drop to approximately 1.1.
Bank of America projects that the penetration rate of liquid cooling in China’s data centers will rise from 30% in 2025 to 70% in 2030, with demand expanding from 1.4 GW to 9.5 GW—a compound annual growth rate (CAGR) of 47%. The overall data center cooling market in China is expected to grow from 2025 to reach RMB 70 billion by 2030, representing a CAGR of 36%. Currently, cold plate liquid cooling accounts for over 90% of the market share, while immersion liquid cooling is projected to increase its share from approximately 5% in 2025 to around 17% in 2030, reaching a market size of RMB 16 billion.
AI Metals
Zijin Mining $601899.SS
Jiangxi Copper $600362.SS
CMOC $603993.SS
China Tungsten $000657.SZ
Xiamen Tungsten $600549.SS
JL Mag $300748.SZ
CGN Mining $01164.HK
PCB Materials
Mitsui Kinzoku $5706.T
Tongguan $301217.SZ
Defu Tech $301511.SZ
Nittobo $3110.T
Jushi $600176.SS
Sinoma S&T $002080.SZ
Huntsman $HUN
Dow $DOW
Mitsubishi $4188.T
Sumitomo Bakelite $4203.T
Shengquan $605589.SS
Dongcai $601208.SS
Electronic/Semi Materials
Murata $6981.T
Samsung Electro-Mechanics $009150.KS
Fenghua $000636.SZ
Three-Circle $300408.SZ
Huate $688268.SS
Guanggang $688548.SS
Anji $688019.SS
Tongcheng $603650.SS
Thermal Materials
Juhua $600160.SS
Dongyue $00189.HK
Envicool $002837.SZ
Shenling $301017.SZ
Delta Electronics https://t.co/AU9d0DefFz
AVC (Asia Vital Components) $3017.TW
Vertiv $VRT
YOFC (Yangtze Optical Fibre and Cable) $601869.SS
Hengtong $600487.SS
Zhongtian $600522.SS
Corning $GLW
Furukawa Electric $5801.T
Pacific Quartz $603688.SS
The Quartz Corp
The photonics market is expected to grow from $3B in 2025 to $91B in 2028.
The deep dive below covers everything you need to know about the industry, key players, risks and much more.
There are 10+ visuals included to help you fully understand it all. Hope you like it!
87 minutes of wisdom can change your entire trading game:
You don’t need 100 trades… you need ONE super stock. The market doesn’t reward effort — it rewards precision. While most chase noise, the patient trader waits for the perfect breakout.
Inspired by William O’Neil.
Michael Pollan explained his entire writing process to me in 70 minutes. Afterwards, he said: "Man, I just gave you a whole semester's worth of lessons in a single conversation."
Timestamps:
0:48 Does coffee make you more creative?
3:37 Do drugs make you creative?
10:45 How psychedelics shape your thinking
12:57 Journaling
16:29 Michael's routine
18:55 Writing a 1st draft
20:34 Why write in the first person?
26:09 The Cow Story
34:08 Balancing stories and facts
39:02 Metaphors are a cheat code
42:48 Finding good questions
51:25 Reading novels
55:40 How Michael does research
1:01:43 What makes a good character?
1:08:35 Art expands consciousness
1:11:05 Michael's top writing lessons
You'll find the full conversation below. If you'd rather watch on YouTube, or listen on Apple / Spotify, check out the reply tweets.
How to fix your confidence:
Focus inwardly
Become aware of the sensations in your body as you notice how good it feels to breathe slower.
Do this as you feel your feet planted firmly on the floor and begin to relax every muscle in your body head to toe.
Within a few minutes you'll notice your body feels heavier and heavier.
Once that sensation arrives, you will be able to visually go within yourself to create permanent upgrades in your subconscious: remembering who you truly are.
Now, ask yourself this question:
"Can I remember the last time I truly felt like I won?"
The moment you acted upon an idea and surprisingly it paid off.
Despite voices of doubt.
The fact you transformed a deep calling, invisible idea in to physical reality.
Maybe it felt as confirmation that you are enough.
Go back there now.
Where were you sat, stood, walking? What was in your hand? Who, if anyone, was near you?
What did you tell yourself in that split second before the thought “don’t get too comfortable” crept in?
Second by second, let yourself feel any winning sensations, maybe a warmth in the chest, maybe somewhere else that says “we did it”
Stay with it a moment longer than your mind wants you to because in a moment your subconscious mind is going to predict the future.
It understands the fact you can remember what it feels like to be confident as proof you already are, so obviously it begins to ask itself:
Who would I become if experienced this every single day?
How would it feel if I felt certain in myself regardless of outside circumstance?
Would I believe this is your potential or just much closer to it?
Enjoy it.
Commit the next 30 days to practice this and see if you find the answers to those questions.
Cogito Ergo Sum
$SPCX shares are priced at $135 for its $2 trillion IPO.
Its return is 100x-200x by 2035.
These 20 companies will benefit the most:
1. $BKSY ~$34
AI-ready Earth observation satellites feed SpaceX orbital intelligence layer.
2. $SPIR ~$20
Space data analytics monetizing SpaceX's growing orbital constellation.
3. $ACHR ~$5
Air mobility networks integrate with Starlink's low-latency infrastructure.
5. $SATL ~$7
High-resolution imaging complements SpaceX orbital AI compute constellation data.
6. $VIAV ~$50
Optical networking components critical for Starlink ground station upgrades.
7. $OUST ~$40
Sensor fusion tech supports SpaceX booster catch reusability automation.
8. $GILT ~$15
Satellite ground infrastructure scales alongside Starlink enterprise deployments.
9. $POET ~$11
Optical interposer chips slash data center power costs inside COLOSSUS AI cluster.
10. $ARQQ ~$12
Quantum encryption securing Starshield government classified orbital networks.
11. $TWST ~$74
Synthetic biology tools accelerate SpaceX long-term Mars life support research.
12. $LUNR ~$30
NASA lunar lander tech directly supports SpaceX Moon base buildout.
13. $AEVA ~$24
LiDAR sensors enable autonomous Starship landing and booster catch precision.
14. $KTOS ~$60
Defense tech partner powering Starshield national security satellite contracts.
15. $IONQ ~$58
Quantum compute layer powering next-gen orbital AI satellites.
16. $RDDT ~$178
Real-time social data feeds Grok's truth-seeking AI via X integration.
17. $RKLB ~$115
Small payload launch fills exact gaps Falcon can't efficiently serve.
18. $ASTS ~$97
Direct-to-phone satellite broadband. Starlink's closest competitor and partner.
19. $MTSI ~$375
RF semiconductors power Starlink phased-array antenna signal processing.
20. $BWXT ~$200
Nuclear propulsion R&D aligns with SpaceX Mars mission power requirements.
I'm definetly a buyer of $SPCX IPO and want to get it super cheap.
♻️ RESHARE this post and write 1 comment, I'll DM you the PRICE I want to buy $SPCX at this month.
$SPCX lists tomorrow.
Everyone’s asking the same thing. I’ve gotten 20+ texts from friends/family:
“If I can’t buy SpaceX directly, what can I do?”
Mapped it below:
- Direct owners
- Comps
- Suppliers
- Infrastructure
- Orbit economy
Save/share. idc
NFA