punk2898 🙌💎
punk2898 🙌💎
Settled in Singapore, high-frequency quantification
53Following
2Kfollowers
Feed
Feed
Pinned
The first live stream is at 8 PM, with just two things: sending red envelopes 🧧 & sending red envelopes 🧧
By the way, let's chat:
- How can ordinary people without funds or skills make their first bucket of gold?
- What is the best way to manage finances for most people?
- What are the mechanisms that are life bugs (life arbitrage)?
- What are the fastest and most probable ways to achieve financial freedom in the crypto world?
And more
If you're interested, follow my personal page ⬆️
If you have any questions, feel free to send them, and we can chat together tonight.
Good questions will receive a separate red envelope 🧧
See you at 8 PM ⏰
$BTC $ETH $SOL $XAUT

Pinned
I created a web version for OKX Planet 🆕
If I were to create a PC version of OKX Planet, it would prioritize solving the pain points of creators, focusing on enhancing the work efficiency and experience of heavy creators.
1. Goals 🎯
Provide a stable, fully functional web interface that makes it easy for creators to create and manage content on their computers.
2. Technical Implementation 💻
Utilize existing APIs to quickly build the web version.
Using AI tools can assist in rapid development.
3. Advantages 👍
Address the inconvenience and inefficiency of mobile operations, allowing creators to better unleash their "combat power."
4. Optimizing Creator Experience
Content Creation: Provide a more comfortable editing interface, supporting richer formatting tools and material management features, which may include:
- More convenient image and video uploading and management.
- Support for advanced formatting like code blocks and tables.
- Features like draft saving and version management.
- Community Management: Make it easier for creators to manage community members, publish announcements, and interact, such as:
1️⃣ Batch management of members.
2️⃣ A clearer notification and message center.
3️⃣ A data analysis panel to help creators understand community activity levels.
- Efficiency Improvement: Reduce the hassle of switching between mobile and computer, allowing all operations to be completed on the PC.
5. Future Outlook (Prototype Design) 📈
I plan to conduct subjective prototype designs for various functions of the planet, which may include:
- More powerful data analysis tools to help creators understand content performance and community dynamics.
- Integration with external tools (like design software and data analysis tools).
- More flexible monetization tools or incentive mechanisms.
Overall, the PC version of OKX Planet I envision will be a creator-centered platform that emphasizes efficiency and experience.
The reason I have such a deep experience using OKX Planet is that I like the tone of OKX products, but the recent iteration speed has indeed been a bit too slow.

Spent two days Vibe Coding an internal tool that can boost work efficiency by 10 times, and this is not an exaggeration
The main products used are:
- Cursor: the most developer-friendly Harness system (if you don't know what Harness is, look it up)
- GPT Image 2: currently the most practical raw image product, without it our business process wouldn't work
- Claude: spent one and a half days out of two communicating with Claude Opus 4.7 on various technical solutions, finally polishing the final version
The deepest insight is that you need to understand the business, the code, and the product to create truly practical products
Starting the next product 😁

Elon Musk retweeted an interview with his good friend Jensen Huang, the founder of Nvidia. Old Huang made a few points:
- AI will increase the world's GDP from 100 trillion to 500 trillion dollars
- AI will eliminate some jobs but will create even more jobs
Interestingly, Musk's retweet and the recent official cooperation announcement with Nvidia mark their honeymoon period, with the matchmaker being Trump. Most likely, many of their collaborations were negotiated on Air Force One
OpenAI is in trouble, karpathy, a founding member of OpenAI, recognized in the AI community as a "living Bodhisattva" level figure and an open-source fanatic, has now joined Anthropic 🤣
Check out his resume:
- One of the founding members of OpenAI
- Former AI Director at Tesla, personally recruited by Elon, led the Autopilot computer vision team, directly driving FSD
- Main lecturer of Stanford CS231n (the legendary deep learning vision course), taught hundreds of thousands worldwide
- Open-source fanatic: projects like nanoGPT, micrograd, char-rnn, directly teaching you how to implement GPT from scratch (various LLM Wiki bloggers just copy his work)
- Wrote a bunch of legendary blogs + produced numerous YouTube videos, explaining the toughest topics like backprop, Transformer, and LLM with exceptional clarity
But this is great news for developers!

Elon Musk is still awesome, hitting three targets with one arrow 🎯
1️⃣ Colossus 1 leased to Anthropic for one year, generating $10 billion in revenue
2️⃣ Colossus 2 quietly leased to companies like Cursor for model training, also bringing in tens of billions in revenue annually
3️⃣ SpaceX 612 goes public, and the financial statements show explosive annualized revenue and profit margins (everyone else is losing money but I'm making profits, yet people still don't buy my stock)
Awesome, really impressive, can't help but admire 👍


小互
Performance comparable to Opus, but 30 times cheaper?
Cursor releases self-developed encoding model Composer 2.5
In terms of scores: Composer 2.5 all fall within the same range as Opus 4.7, with a maximum difference of less than 1 point.
In terms of price: Opus 4.7 costs about $15 per million input tokens and $75 per million output tokens, while Composer 2.5 is 10 times cheaper for input and 30 times cheaper for output.
Cursor claims that Composer 2.5 has significant improvements over Composer 2 in intelligence and behavioral performance, especially in long-duration tasks, complex instruction adherence, and smooth collaboration.
Long tasks can continuously progress across rollouts spanning hundreds of thousands of tokens without easily going off track.
Complex instruction adherence is more reliable, communication style and engagement level calibration are more stable, and the intensity of work is better adjusted.

Nearly three years have passed, and picking up this book again brings a whole new feeling. The hardest part is knowing you're wrong, but still having to grit your teeth and persevere.
Did this book help me lose less money? Definitely yes.
But you only regret not having enough knowledge when you actually need it; true understanding comes from practice.
Many truths can only be understood by experiencing them yourself.
punk2898 🙌💎
21 Ways to Fail at Startups, Helping You Avoid Losing 1 Billion 💰
You don’t necessarily have to start a business, but this can help you steer clear of these companies
This is a book I read 8 years ago called "The Little Failures of Startups"
But even now, I still occasionally revisit it
❌ 5 Major Startup Pitfalls
📈 21 Classic Cases
🙅 10+ Industry Lessons
💬 45 Authoritative Analyses and Reviews
Don’t care about what others say about this book
Just listen to me and that’s enough 😎
I started my first startup in 2015. Since it’s a startup, I definitely needed to cram a lot of entrepreneurial knowledge. I read many books, but one had a particularly big impact on me, which is "The Little Failures of Startups."
Strictly speaking, this book isn’t really a book; it’s a collection of interviews with various star founders—of course, those who failed, learning from history.
Even now, I still flip through it from time to time to see if I’m making the same mistakes they did.
The strange thing is, you can only truly learn lessons by experiencing them yourself; just reading books isn’t enough, but at least it gives me a reference.
⚠️ Note:
💪️ They don’t expect entrepreneurs to avoid failure, but when you encounter failure, tell yourself to keep going
🎙️ Since these are interviews, the subjects might not always tell the truth, so don’t be misled by the author when reading
📖 Su Dongpo had a way of reading called the "Eight-Sided Reading Method," where you read with different purposes each time and gain different insights
--
Table of Contents (Check if it suits you):
Chapter 1: The Lament of Resource Dependents
1. Resources are both a pie and a pit
2. Losing the battle with investors
3. The abandoned giant whale
4. Thoughts on the transformation of traditional enterprises
5. The sudden death of "China’s Best Business Model"
Chapter 2: The Costly Price of Trial and Error
1. I was defeated by success
2. A 90s entrepreneur eager for quick success
3. Lessons from county-level internet startups
4. The split of "Grassroots Bull Blog"
5. The life-and-death crisis of Shengshi Hexing
6. Drinking poison to quench thirst leads only to death
Chapter 3: Unreliable Business Models
1. Fresh O2O exploration
2. The "cost black hole" I created myself
3. The echo of product pioneers
4. The model dilemma of Shede.com
5. The fleeting dream of the car wash app Cherry
Chapter 4: The Aftermath of Expansion
1. Jingrui Education’s "Northern Expedition"
2. The deadly trap of scale growth
3. "Franchise stores" cost me millions
4. The death of a profitable fast-food restaurant
5. Absurd mergers and barbaric bankruptcies
6. The setback of Bashan Farmer
Chapter 5: The Founder’s Self-Redemption
1. Don’t twist yourself, and don’t twist others
2. The self-redemption of Six Rooms
3. The near-death experience of a zero-fund small team
4. The life-and-death group treasure
5. Entrepreneurs who can kneel down
Questioning US stocks → Understanding US stocks → Joining US stocks
This is the Nasdaq weekly K-line, spanning BTC bull and bear markets, this is the world's largest capital pool, rising amid doubts
- What about the US-Iran war? It doesn't hinder my AI narrative
- What if the 10-year US Treasury yield rises? I keep going up
This week, Fed dynamics, economic data, and USD/JPY will change, but I will keep rising





