Data Scientist｜cooking ｜Exclusive Telegram: t.me/CcPrimary
- Tweets 6,040
- Following 2,012
- Followers 49,148
- Likes 7,146
Tinkle retweeted
如果你不懂天才的伟大，我不会说服你😘
Most tech giants in the 2000s built their infrastructure and product as one entangled unit. Amazon had the foresight to separate out AWS as an API layer, of which Amazon retail was the first of many users. Today, AWS generates more profit than all of Amazon's other business lines combined.
Hyperliquid is built with the same philosophy. Housing all of finance requires thoughtfully designed, open financial primitives. Each primitive should obey the Unix principle of "Do one thing and do it well." Talented builders then have the foundation to chain these together to create magical applications.
HyperCore borrowing is an example to highlight this philosophy in action.
Most other platforms implement portfolio margin by marking an account's collateral to market value with an LTV haircut, creating borrowed assets without an explicit lender. This system is simpler to implement, but misses a golden opportunity for composability.
Hyperliquid instead begins with a borrow/lend protocol on HyperCore. Every borrowed asset is sourced from a supplier, so risk is isolated within the borrow/lend primitive instead of platform-wide. HyperCore's portfolio margin system is implemented as an orchestration layer that composes borrow/lend, with other primitives such as perps, spot, and outcome trading.
This decomposition has several nice corollaries:
1. Today's announcement of manual borrowing is not a new feature, but simply an extension of the underlying primitive. Borrowers on day one have access to 400M and growing of supplied liquidity.
2. Portfolio margin users earn interest on their idle stablecoin collateral. This is not a new feature, but a natural byproduct of composing trading with lending.
3. System safety is easier to reason about when perp and borrow/lend margining are independent.
In the same way that math theorems almost prove themselves when the right abstractions are defined, composable designs just feel right.
1
1
1,769
VergeX is making AI-powered trading more accessible.
Turn natural language into trading strategies, execute them autonomously, and onboard in seconds with a seamless self-custodial experience.
Protected by Privy.
The future of AI trading requires infrastructure that matches its intelligence.
@vergex_ai has integrated @privy_io embedded wallet infrastructure and Account Abstraction.
Users can now create secure non-custodial wallets in seconds through one-click email login, with no seed phrases or setup required, and immediately begin using AI trading agents.
4
4
28
12,944
Tinkle retweeted
Had an awesome time at OpenAI Founder Day Singapore today!
Loved connecting with so many sharp AI founders and builders from across Southeast Asia, and getting direct insights from the @OpenAI team.
The energy was high and the conversations were genuinely inspiring.
Proud that VergeX was part of it — walking away with fresh ideas and motivation to build.
2
1
14
2,841
Tinkle retweeted
GLM-5.2 delivers a substantial leap in app development capabilities, which also represent demanding long-horizon tasks.
Results:
- GLM-5.1: 21/70
- GLM-5.2: 48/70
- Claude Fable 5: 56/70
That's more than a twofold improvement from GLM-5.1 to GLM-5.2.
These come from an internal benchmark of 35 challenging mobile development tasks, each run twice for a total of 70 trials. We measured task completion, defined as core features working without major issues.
77
100
1,449
345,179
程序员圈子都在说GLM的coding plan特别好用，官网基本上抢不到，比演唱会门票还难抢，首个接近opus4.8的模型。
Introducing GLM-5.2: Frontier Intelligence, Open Weights
- Significant improvements in coding and agentic tasks
- Strong long-horizon capabilities with a 1M context window
- Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency
- MIT-licensed open weights
- Same API pricing as GLM-5.1
Tech Blog: z.ai/blog/glm-5.2
Weights: huggingface.co/zai-org/GLM-5…
API: docs.z.ai/guides/llm/glm-5.2
Coding Plan: z.ai/subscribe
Chat: chat.z.ai
5
12
6,150