Ming

Investor, one person, some LLMs, many agents.

[Ming @huangkaiming]

@Ueueueuwn 178

[Ming @huangkaiming]

@weiyux2021 每天打球，等待

[Ming @huangkaiming]

@xiaomustock 跟了

[Ming @huangkaiming]

AI算力扩张→光互连刚需→康宁产能被巨头提前锁定+玻璃基板潜在第二曲线，估值重估正在进行中。
https://t.co/BSF6UElfQy

[Ming @huangkaiming]

AI 的代码是生活是效率，自己写代码是享受，是思考是优美散文集。

[Howie Xu @H0wie_Xu]

Is code quality still important, or should we focus purely on outcomes?

I'm still an engineer deep down, so I used to believe code quality was paramount - maybe the only thing that mattered. Clean architecture, proper testing, maintainable systems. These felt sacred.

But I've been rethinking this philosophy over the past three months, largely because of OpenClaw.

By traditional standards, OpenClaw's codebase is deemed to be no-good. No comprehensive testing, minimal code review, built entirely through "vibe coding." Any senior engineer would look at it and cringe.

Yet Jensen Huang called it "the most important software in human history."

Jensen clearly has motivations (he wants people burning tokens). But he's not someone who makes careless statements, and I think there's truth in his assessment.

This forced me to examine code quality from a different perspective.

Consider this historical parallel: today's Java code would look absolutely terrible to someone from the assembly language era. The overhead, the abstraction layers, the "inefficiency" - it would seem like garbage to a programmer optimizing every register and memory allocation.

But Java's "write once, run everywhere" philosophy won despite the performance costs because the outcome - developer productivity and platform independence - mattered more than code elegance.

I think we're seeing the same pattern with AI-generated code. Yes, it's inefficient. Yes, it burns resources. Yes, it violates many software engineering principles.

But if it produces better outcomes faster, does that change the equation?

Six months ago, I would have firmly defended traditional software engineering practices. Today, I'm convinced that outcome-driven development is more important than code quality purists want to admit.

This doesn't mean abandoning all engineering discipline. But it does mean recognizing that different contexts require different optimization targets.

If you can deliver valuable functionality to users 10x faster with "messy" AI-assisted code versus "clean" hand-written code, the trade-off might be worth it.

The future of software development may care less about elegant implementation and more about effective problem-solving.

Quality still matters. But defining quality purely by code aesthetics rather than user value is becoming an outdated approach.

#SoftwareEngineering #CodeQuality #AI #VibeCoding #OpenClaw #ProductDevelopment #TechPhilosophy #Innovation #SoftwareDevelopment #OutcomeDriven

[Ming @huangkaiming]

Define AI agent roles via Markdown and simulate team collaboration. Lightweight, zero-dependency, and ready to fork — adapt it to your own project workflow as a small team or solo developer.
https://t.co/x2OIEJz7xP

[Ming @huangkaiming]

@somi_ai @Saboo_Shubham_ Broadcast

[Andrej Karpathy @karpathy]

Nice - my AI startup school talk is now up! Chapters:

0:00 Imo fair to say that software is changing quite fundamentally again. LLMs are a new kind of computer, and you program them *in English*. Hence I think they are well deserving of a major version upgrade in terms of software.
6:06 LLMs have properties of utilities, of fabs, and of operating systems => New LLM OS, fabbed by labs, and distributed like utilities (for now). Many historical analogies apply - imo we are computing circa ~1960s.
14:39 LLM psychology: LLMs = "people spirits", stochastic simulations of people, where the simulator is an autoregressive Transformer. Since they are trained on human data, they have a kind of emergent psychology, and are simultaneously superhuman in some ways, but also fallible in many others. Given this, how do we productively work with them hand in hand?
Switching gears to opportunities...
18:16 LLMs are "people spirits" => can build partially autonomous products.
29:05 LLMs are programmed in English => make software highly accessible! (yes, vibe coding)
33:36 LLMs are new primary consumer/manipulator of digital information (adding to GUIs/humans and APIs/programs) => Build for agents!

Thank you again for the invite @ycombinator and congrats again on an awesome events! I'll post some links/references in the reply.

[Ming @huangkaiming]

🚨 为什么这很重要？
MCP 正在成为 AI 工具世界的“REST 接口”，
为整个生态提供一致的接入规范，
让开发者专注于造好工具，而不是反复造轮子。

[Ming @huangkaiming]

极简图解：MCP与Function Calling
简短结论：它们 不是竞争标准，而是互补关系！

📌 Function Calling 解决的是：
让大模型知道 何时 该调用外部工具
构造调用工具所需的参数结构
主要作用在 单一应用内部
实际怎么跑工具、工具在哪里，得你自己搞定。#MCP #LLMs

[Ming @huangkaiming]

❗️关键区别：
1. Function Calling：告诉 AI “用什么工具”、“什么时候用”
2. MCP：告诉 AI “这些工具放哪了”、“怎么稳定找到并用上”

🧠 更直观的理解：
Function Calling 就像 AI 说：“我现在需要搜索”
MCP 就是：“搜索工具在哪？怎么统一接入给我用？”

[Ming @huangkaiming]

课程内容：
-针对普通用户普及人工智能的课程；
-面向工程师的深度学习课程；
-文本、图片和视频生成课程；

[Ming @huangkaiming]

OpenAI 刚刚推出了免费的人工智能课程。 针对初学者和工程师的课程。 而且它是 100% 免费的。 以下该课程的相关信息：

[Ming @huangkaiming]

您可以通过以下方式访问 OpenAI Academy： https://t.co/OmhU1pZo7L 正在直播。 它是免费的。 它可能会永远改变你的职业轨迹。

[Sam Altman @sama]

this is the most interesting year in human history, except for all future years

[Ming @huangkaiming]

@LabSpeculation 学习学习

[Ming @huangkaiming]

@liaohaibo 身不由己的时间里都是没有创造力的。

[Ming @huangkaiming]

@nangongyuan 和你一样的经历

[Ming @huangkaiming]

@BTCwukong 眼熟