Jeff Wang

[Jeff Wang @HiJeffWang]

长周期任务的洞察特别认同，应该尽快投入长周期任务探索

[jietang @jietang]

Recent thoughts:
The Shift to Long-Horizon Tasks
The most likely breakthrough this year will be in long-horizon tasks. We are moving toward a stage where Large Language Models (LLMs) learn to complete extended, complex missions by interacting with Agent environments. This is perhaps where the true value of LLMs lies. Take cybersecurity as an example: imagine a model that continuously hunts for software bugs and vulnerabilities. While it sounds like a search process, it’s actually the model learning the high-level intuition and methodology of a professional hacker. Unlike humans, AI can run 24/7 without fatigue. It could potentially find exploits at a much higher frequwill ency and claim bounties on platforms like HackerOne or BugCrowd. It sounds fun, but fundamentally, it's a revolution that displaces the hacker. If even hackers are being "disrupted," one can only imagine the impact on general programmers.

From One-Person to None-Person Companies
Building on long-horizon capabilities, Autonomous Agent Systems (AAS) will inevitably become the next frontier. Last year, we were discussing the rise of the "One Person Company" (OPC). I didn't expect us to move so quickly toward the "None Person Company" (NPC). It’s an ironic twist—we might all end up as NPCs in this new ecosystem.

Engineering the Impossible: Memory and Learning
To realize the vision above, we must solve three technical pillars: Memory, Continual Learning, and Self-Judging.
I used to think these would require massive paradigm shifts and years of research. However, the pressure from both the technical and application sides is so intense that we are seeing these capabilities emerge through ingenious engineering "tricks":
Memory: Long context windows (1M+) and RAG have significantly bridged the gap.
Continual Learning: While true continual learning remains difficult, the release cycles are shrinking. Global models are updated monthly; domestic models are catching up. If we reach weekly updates by next year, it will effectively function as continual learning.
Self-Judging: This remains the most elusive, yet models like Opus 4.7 are already demonstrating early self-correction and judgment capabilities.

The Self-Evolving Endgame
The most difficult—and most promising—path is Self-Evolution. The current wave is incredibly fierce. I suspect that models like Claude may have already achieved a baseline for self-training: writing their own code, cleaning their own data, generating synthetic data, and then training on it. It might "waste" some compute, but it saves the most precious resources: human labor and time. In the LLM era, speed is everything. Rapid iteration is what creates the cognitive gap between leaders and followers. Claude’s rumored 2-million-chip cluster for next year is likely dedicated to exactly this: autonomous model self-training.
Technical Summary:
1M Context: Necessary baseline.
Memory & Continual Learning: Prerequisites, likely solved first via "tricky" engineering.
Harnessing Environments: The breakthrough point.
Self-Judging: The tipping point.
Full Self-Training: The endgame.

Redefining AGI and the Industry
If this is the road to AGI, then AGI’s definition should be the sum of all human collective intelligence, not just an individual’s intelligence. It must possess the creative capacity to produce something as profound as the "Theory of Relativity"—meeting the bar set by Hassabis.
During this transition, every APP will need to be reconstructed as AI-native. In fact, we might move past the concept of APPs entirely. The most significant challenge will be the reconstruction of the operating system itself. In the future, you won’t see a traditional desktop; you will see an LLM OS, where applications are "generated on demand." This challenges the 80-year-old Von Neumann architecture and represents a total upheaval of the computer science industry.

The Irreversible Wave
From completing long-horizon tasks to fully autonomous operations, every sector—Security, Finance, Law, E-commerce—will be reshaped. Many friends have reached out lately, asking how to transform their enterprises to keep pace with AI. But few truly realize that this irreversible process has already begun. As this massive technical wave hits, we must be prepared to act, but we must also start thinking seriously about how to regulate it.

[Berryxia.AI @berryxia]

刚刷到CJ Zafir 发了一条关于 fine-tuning 小模型的帖子，看下来觉得这波建议特别实在。

他直接说，如果你也喜欢玩开源模型 fine-tuning，那先听听这些：

从 1B、2B、4B、8B 这些小模型开始练手，别一上来就冲 27B 以上。

云 GPU 用 Google Colab Pro 就够了，A100 80GB 一小时才 0.6 美元左右，小模型完全够用。

数据集自己造，用 Codex 5.5 先规划，再配 DeepSeek v4 Pro 生成每一行数据。

底座模型推荐 Unsloth 的 instruct 版本，Hugging Face 上直接拉，fine-tuning 笔记也用他们的做参考，直接丢给 Codex 让它帮你改成你想要的配置。

他建议花一天时间把这些东西过一遍：SFT、RL 训练（GRPO、DPO、PPO 这些）、LoRA / QLoRA、量化类型、本地推理引擎（llama.cpp）、KV cache 和 prompt cache。

他说就直接上手吧，Claude、Codex、ChatGPT 都能给你设计第一步的完整计划。

最后他还提到，未来技术会越来越往 5B 到 15B 的 Expert Language Models 走，并非一味堆通用大模型，所以 fine-tuning 这门手艺现在学特别值。

很多公司愿意花 5 万美元以上，让你帮他们用自家数据训个性化模型。

整条帖子的意思就是：fine-tuning 其实谁都能入门，调模型、测模型、用模型，慢慢就能把这变成一份靠谱的事业。

感兴趣的可以看看，还���有意思。

[Vincent | 信号＞噪音 @VincentLogic]

Obsidian + Claude Code = 你的个人生活操作系统
收藏夹吃灰？笔记太乱？
想拥有 Andrej Karpathy 同款知识管理系统？
这套 Obsidian + Claude Code 的“第二大脑”搭建指南，帮你把死知识变活！👇

1️⃣ 核心架构：三层大脑回路
底层 (Obsidian): 本地存储，Markdown格式，数据主权在你手里。
中层 (Claude Code): AI 思考与行动引擎。它读取 Obsidian 的内容，帮你分析、提取、执行。
顶层 (Skills): 自动化工作流。从 GitHub 下载现成的 Skill，让 AI 自动跑任务。
👉 以前 Obsidian 是静态笔记本，现在它变成了会思考的助手。
2️⃣ 工作流第一步：信息摄入 (Input)
别手动复制粘贴了！
Web Clipper: 浏览器插件，一键保存网页、YouTube 视频（含字幕）、公众号文章到 Obsidian Inbox。
Daily Note: 每天自动生成日记模板，记录睡眠、运动、灵感。
Podwise: 播客转文字，无缝接入知识库。
3️⃣ 工作流第二步：AI 分析与处理 (Process)
这是 Claude Code 发威的地方！
/todo 命令: AI 扫描你的日记和上下文，自动生成“艾森豪威尔矩阵”（重要/紧急），帮你排好今日任务。
/weekly learning: 每周复盘。AI 分析你一周的笔记，提取认知改变、盲点和新想法。
/summary: 和 AI 聊完天，一键总结对话精华，存入知识库，不浪费任何灵感。
4️⃣ 工作流第三步：整理与检查 (Organize)
/lint 命令: 定期给知识库做体检。AI 检查概念定义是否冲突、有没有孤岛笔记（没链接的笔记），生成健康报告。
知识沉淀: AI 自动把 Inbox 里的碎片信息，整理成 Concepts（概念）和 Summary（摘要），形成结构化知识网。
5️⃣ 一人 IP 创作系统 (Bonus)
这套逻辑同样适用于做自媒体！
Step 1: 全网选题扫描，收集 Input。
Step 2: Claude 分析创作库，发现读者真正需要的内容（笔记数据分析）。
Step 3: AI 辅助 SOP，自动生成初稿，你只做精修和判断。
6️⃣ 高手心得 (Key Takeaways)
一个入口 (Inbox)，多个终点。
结构够用就行，别追求完美分类。
AI 建议，你拍板。
对话也是知识，记得用 /summary 保存。
不绑定任何工具，AGENTS.md 兼容多种 Agent。

Obsidian 负责“记”，Claude Code 负责“想”和“做”。
人类只做高价值的筛选与决策，AI 接管所有繁琐节点的编织。

[宝玉 @dotey]

这个可视化看claude code原理、功能的网站不错，尤其是 Agent Loop 动画，很直观
https://t.co/z7P2DB8TCm

[Sanbu 散步 @sanbuphy]

是不是觉得 karpathy 的 autoresearch 很酷？又看到一个类似的自动化工具：ArgusBot。
本质是一个 7*24 小时循环迭代任务 的科研执行系统。它可以让任务持续进行，实时汇报进展、接受指令。
我自己平时做科研/工程时，经常会遇到需要反复盯进度的问题，有了这个方便很多。
https://t.co/zroknjWBfE

[The Deep View @theDeepView]

We spend all week researching the top AI tools so you don't have to...

Get insights, tools, and news for free when you join our community of 150,000+ readers.

Click the image below to subscribe for free 👇

[Andrej Karpathy @karpathy]

The most dramatic optimization to nanoGPT so far (~25% speedup) is to simply increase vocab size from 50257 to 50304 (nearest multiple of 64). This calculates added useless dimensions but goes down a different kernel path with much higher occupancy. Careful with your Powers of 2.

[Simon Wardley @swardley]

A complete economic understanding of cloud and serverless, what matters and what is different in one simple diagram.

[DigitalOcean @digitalocean]

Get more CPUs, memory, & SSD storage for less. Test out our new Optimized Droplets with a free $100 credit. https://t.co/ONsAK1OUu0

[Kaggle @kaggle]

We're running a 5-Day SQL Scavenger Hunt to get you started with SQL basics starting on Feb 11. It's a good opportunity to ramp up your skills while exploring our new BigQuery integration (making it blazingly fast to query terabytes of data). Sign up here: https://t.co/0u5PslfJsj

[Dr. Ganapathi Pulipaka 🇺🇸 @gp_pulipaka]

Install Dlib (The Easy, Complete Guide). #BigData #DeepLearning #MachineLearning #DataScience #AI #RaspberryPi #IoT #Analytics #CPP #Python #macOS #Ubuntu #Linux #XCode
https://t.co/P9qG4f9BSD

[Andrew Ng @AndrewYNg]

This is a beautiful summary of an "AI is the new electricity" talk. Thank you @TessFerrandez! https://t.co/GqZuca6v7P

[Google DeepMind @GoogleDeepMind]

Reflections on 2017 from the @DeepMindAI founders @demishassabis @mustafasuleymn @ShaneLegg https://t.co/TY8kYWbXPT

[David Robinson @drob]

Wow- NLP analysis discovers more than a million fake anti-net-neutrality comments. Brilliant data science detective work by @thisismetis student Jeff Kao. https://t.co/p9vooF5GPd #datablog

[François Chollet @fchollet]

Quick tutorial on using tf.keras with the TF Estimator API and the TF Dataset API: https://t.co/FOrGq0oUYT

[Ian Goodfellow @goodfellow_ian]

While GANs have been great at generating realistic images from a single category (one GAN for faces, another GAN for buildings) they've always struggled to fit all 1,000 classes of ImageNet with a single GAN. This ICLR submission has done it: https://t.co/PxDqplEymF-

[Stanford NLP Group @stanfordnlp]

Pytorch implementations of many of the deep learning NLP models discussed in cs224n by Kim SungDong. #dlearn #NLProc https://t.co/aGVpONdoVe

[Ian Goodfellow @goodfellow_ian]

GANs for generating images of how clothes will fit. Only two of these images are photos. https://t.co/FYCBoveqh9

[Dustin Tran @dustinvtran]

Finally had time to watch this insightful talk on "AI impact on jobs" by Michael Osborne. Highly recommend it. https://t.co/S7xofUmvCG

[Durk Kingma @dpkingma]

"Variational Inference and Deep Learning: A New Synthesis", written by yours truly, is now available for D/L here: https://t.co/bp7ZsWrC8f.