Curious Penguin

Explorer/Investor/Observer in Healthcare/Crypto/AI

Passionate about interdisciplinary research

[Curious Penguin @CuriousPenguin]

@zephyr_z9 郭嘉定的价

[Curious Penguin @CuriousPenguin]

@labubu_trader 期待，快点写！

[Curious Penguin @CuriousPenguin]

MS expect $1.2/1.4tr capex from hyperscalers in 27/28. I expect a lot more capex from private credit on data centers rented by hyperscalers — US has $3tr real estate private credit funds, can do you think they will continue to invest in human houses or silicon houses?

[Curious Penguin @CuriousPenguin]

Prime shows last week the 1st week of HF net buy in past few weeks(since week of Jun22), mainly short cover but small long buy. Start turning bullish on sentiment. Fundamentals not worry me anyway.

[Curious Penguin @CuriousPenguin]

@labubu_trader Today counts as capitulation?

[Curious Penguin @CuriousPenguin]

A very hard working team…

[Curious Penguin @CuriousPenguin]

Nice to have the competition

[ClaudeDevs @ClaudeDevs]

We've reset 5-hour and weekly rate limits for all users.

[Curious Penguin @CuriousPenguin]

在看AI的应用，发现一个有趣的小例子：隐适美 $ALGN 公司内部现在有个AI审核整牙的进展 -- 原先是每隔固定时间，患者去诊所看看要不要换下一副牙套；现在则是患者拍牙齿的照片，AI审查是否矫正进度达标，达标了就去诊所换下一副，还没达标（比如懒惰了老偷偷不带）就不用去。这里的技术、AI做的事儿都不算难，但非常有意思，算是个性化医疗的第一步（进度个性化）；以后也许我们会看到更多治疗方案也个性化的案例。Very positive on what AI can bring to our society!

[Curious Penguin @CuriousPenguin]

Watched another Naval's recent video on AI Industrial Revolution, again below is my 100% human written takeaways （人类手搓笔记）:

- Choose the right thing to work on, esp. when AI give you 100-1000x leverage; Need better judgment to decide what to work on in the first place；
- Prompts (input quality) matter now but will matter less and less with models getting smarter, as long as there is a solvable problem; for the unsolvable domains, human need to work closely with AI;
- AI can now generate software; in the future (soon) it will generate step files and PCB layouts ----- going to hardware design territory;
- Also, AI already enables hardware companies to write better software to control their hardware products (I totally agree here --- for medical field that I'm very familiar with, most CT/MRI's GUI is terrible);
- One reason for China to go open-source model is that China wants to use this to give leverage to the hardware economy (China already quite dominant here --- 中国政府永远在下一盘大棋);
- Except for coding, Gemini($GOOG) is still the best industrial production model;
- Regulatory barriers are one of biggest holdback for physical AI to develop; FDA big slow down in Medical AI; Madeup random regulations are all over the place and contradicting each other; However it's also noted that regulators' asymmetric risk led them to be conservative;
- On US healthcare system - you are basically running a small communist society inside a larger capitalist society; private market in healthcare can lead to faster advancement;
- Autonomous AI: running a lot of autonomous security research, optimization work; if there is anything left to automate, do it;
- Number of code may have gone up 10x now in past year -- it's great video game that real stuff comes out;
- What human can uniquely do? Creativity? already arguments in the group; can art value be taken by computer? no clear answer; if art is not just certain data distribution but conveys feeling and purpose, then it's still in our human hands;
- Naval (not all speakers) are very pro-human and believe human with AI will only create more wealth, and hire more people; Naval said if someone is smart, creative and use AI, he wants to hire them more than ever because there is leverage to get from there; AND a lot of them don't need to be hired -- they create their own thingl;
- Naval believes by the end of the day, everyone will be generalists and can do anything; Nurture your creativity, taste and judgment, and use AI to leverage yourself to everything;

花了1个多小时手搓note，速度慢了，但是吸收/内化效率高了非常多。以后会保持这个学习习惯。Hope you enjoy reading my takeaways -- should be better than AI's.

[Naval @naval]

Full podcast episode with @rauchg, @maxhodak_, and @bscholl.

40 minutes of unreleased material.

The AI Industrial Revolution

Part 1: Waste Tokens, Save Time

0:00 Three Frontier Founders
1:27 AI Software Factories
4:15 Waste Tokens, Save Time
5:47 Models Instructing Humans
9:29 Is Pure Software Dead?
12:03 You Don't Get Stuck Anymore

Part 2: Vibe Coding Hardware

14:39 Vibe Coding a Turbine Blade
18:07 Open Source Compounds China's Advantage
20:15 You Always Want the Smartest Model
22:44 Software Still Needs Hands
24:43 Humans Are Becoming Verifiers

Part 3: The Regulatory Frontier

27:53 The Regulatory Red Queen Race
32:32 Why There's No Innovation in Healthcare
36:49 We Need a True 50-State Experiment
40:31 China's FDA Is Beating Ours
43:37 Healthcare Is a Communist Society Inside Capitalism
45:57 Sid's Story: N-of-1 Medicine

Part 4: The Autonomous Company

47:49 Autonomous Infrastructure
51:25 Your Job Is to Train the Agent
54:54 The Next Lord of the Rings
59:08 What's Your Definition of Art?
1:05:00 Can AI Have New Ideas?
1:07:03 A Large Number of Small Teams

[Curious Penguin @CuriousPenguin]

100% human watch and human write for key take from this video (so there will be bad grammar and typo):
- Gary: Best way to know what future like, live in it now by paying for multiple AI and agent (if financially ok to you);
- Naval: Coding is to communicate with machine, hence use AI; writing is to communicate with human, hence write yourself; also writing and talking help shape your thinking;
- Naval: Technology diffusion (in this conext: from US frontier lab to China) is hard to prevent; even China doesn't distill US models, it's basically US' Chinese vs China's Chinese in AI development -- those talents talk and move;
- Naval: AI eat software and China eat hardware except for a few (e.g. SpaceX) -- both software and hardware are becoming commodity;
- Vive: strongly recommend GLM5.2;
- Naval: AI lowers bar for many start-up, yet not for LLM -- LLM requires big capex;
- Group: $GOOG /Gemini has lost it -- they may need to cut 5,000 PMs, PMs are slop; Best talents want OAI/Anthropic's share, sitting with best people, getting to OAI/Anthropic's data set, and not tied down to those PMs and legacy BS; (personally, I like Google's products and Demis, I hope they can turn around);
- Gary: Most CEOs don't want to fire 50% employees (despite financially / profit wise it's the reasonable thing to do);
- Gary and Naval: Half people Gary knew at $META has left; you can't build the culture from buying people; key issue is still big org like Google;
- Naval: Wonder whether there will still be ecosystem for start-ups, because general models (strong and stronger) can beat the specialized model (e.g. Harvey);
- Gary: 2027 will be harness war, people to decide which to use, codex or openclaw or claudecode or...
- Naval: Post 2007 there were a lot of start-ups because there are two mobile systems: iOS and Android; it would be much more difficult for start-ups if there were only one monopoly; Now for AI LLM, at least there are two horses competing, and making each other behave a bit;
- Naval: if only one player left and nationalized, DMV style running, US models will lose to China;
- Naval: Don't think US and China will have war; US' real problem is civil war between nationalists and communists;
- Group: China/CCP is very internally focused;
- Gary: We shouldn't have UBI (universal basic income), instead we should have UBR - universal basic robot；
- Naval: If AI becomes as good as human but not better, human's role will be pokemon trainer and handler, to train and handle AI and robots -- that's what people use AI today are doing. They are more productive than ever and they are not displaced by AI. The only people who are displaced by AI are those who refuse to use AI; If you use AI, you have more work than ever

Aboves are basically my takes after watching - it has its limits that I may have only written down what I can comprehend and think important. But probably still better than AI summary? Anyway, I had a great fun watching this video.

[Naval @naval]

New podcast with @garrytan, @farbood and Daniel Francis.

Live in the Future!

00:00 Guest Intros
02:35 Live in the Future
03:58 Will AI Outsmart us?
07:43 In the Anthropic Breadline
09:59 The Tech Genie Is Out
12:33 We Invested in COVID?!
14:25 Good Writing Is Novelty
18:50 Living Like It’s 2028
24:32 Truth dot ai
30:18 Does China have the Weights?
35:38 Everyone has AI Anxiety
39:32 Have Your Agent Talk to My Agent
42:01 What if Open Source takes the Lead?
44:03 The Sun is Setting on Google
48:00 Ride the AGI
50:46 Will There be Startups?
54:05 Defending Taiwan
1:00:05 The California Empire
1:01:26 If the U.S. Falls
1:03:11 Universal Basic Robot
1:06:01 Humans as AI Handlers

[Curious Penguin @CuriousPenguin]

A bit more front-line user color to add to Jukan's points: my friend's small Chinese company actually use DeepSeek and Zhipu's direct API rather than deploying the opensource model freely - the reason is that the model iterated so fast that it's better just use official API to keep it up, rather than deploy a newer version of model every few month.

[Jukan @jukan05]

I don’t understand why supporters of Chinese open source keep pretending not to see what is happening right now: even Chinese “open-source” players are raising prices or gradually shifting toward more closed models.

I think China’s token dumping has already bottomed.

Just look at Zhipu, the company behind GLM-5.2. Even Zhipu has raised prices several times this year.

Chinese LLM developers cannot ignore ROI forever.

Open source is not the same thing as cheap API pricing.

[Curious Penguin @CuriousPenguin]

Everyone: "Nvidia's revenue-share model is brilliant recurring income!"
Me: it's AIG selling CDS with extra steps.
Better than AIG on one dead cloud (hardware margin cushions it).
Exactly AIG on a dead market — every small cloud rides the same AI-capex beta. One down, all down.
$NDVA Rent in good times. Steamroller in bad ones. 🧵

Read my article on detailed comparison from different perspective

[Curious Penguin @CuriousPenguin]

https://t.co/zKr3Z3rqRr

[Curious Penguin @CuriousPenguin]

https://t.co/zKr3Z3rqRr

[Curious Penguin @CuriousPenguin]

So a lot of hire in FDE right？Making $PLTR CEO mooooooore pissed

[TrendSpider @TrendSpider]

BREAKING: Microsoft $MSFT is committing $2.5 billion and 6,000 employees to a new subsidiary called "Microsoft Frontier Co."

The new group will focus on assisting clients with implementing AI technologies.

[Curious Penguin @CuriousPenguin]

Previously you need to buy the house (GPU) to live in, now you can rent it and pay the rent with your monthly salary --- Let's hope Jensen verify his tenants carefully to cherry pick the high earners...

$NVDA #GPU

[Wall St Engine @wallstengine]

$NVDA is reportedly offering financial backstops to smaller GPU cloud providers in exchange for a cut of their cloud revenue, per The Information.

Nvidia guarantees it will rent unused GPU capacity if customers can’t fill it, helping neoclouds finance GPU purchases and data centers.

The program is reportedly called the AI Compute Partnership internally.

Nvidia previously agreed to buy CoreWeave’s unsold capacity through 2032, worth $6.3B at the time, and has since guaranteed another $3.5B tied to customer data-center leases.

[Curious Penguin @CuriousPenguin]

Had dinner with an industry legend tonight.

The alpha:
Top VCs are now specifically hunting sub-10-person AI-native companies. Picture a "professional services firm" that's actually 5 people, a massive server room, and 200 Claude Max accounts.

Big companies can't compete on iteration speed — their structure is the anchor. Layoffs don't fix it. So the smartest people are walking out the door and joining these tiny AI-agent shops.
The talent migration has begun.

You worry if AI bubble has come, I say the big demand wave is yet to come

$Anthropic $Claude

[Curious Penguin @CuriousPenguin]

US companies more and more like Chinese companies 😅

[Zephyr @zephyr_z9]

Is this some kind of bribe for regulatory favors??
Will Anthropic be forced to do this too?
Will SpaceX & Google be forced to do this too?

[Curious Penguin @CuriousPenguin]

Wow, below sounds very familiar to me — In China, companies in strategic industries often have state-owned capital investment.

[Andrew Curran @AndrewCurran_]

OpenAI is proposing handing over a 5% stake to the Trump administration according to the Financial Times.