第67期 karpathy/micrograd

今日摘要

GitHub openai：Official JavaScript TypeScript library for the OpenAI API

GitHub anthropics：Open source repository of plugins primarily intended for knowledge workers to use in Claude Cowork

GitHub karpathy：Inference Llama 2 in one file of pure C

GitHub openai：OpenAI Privacy Filter

GitHub karpathy：LLM Council works together to answer your hardest questions

总结 + 观点：A tiny scalar-valued autograd engine and a neura…｜中文观点：对 karpathy/micrograd 来说，更值得判断的是它会不会进入团队默认工具链，…

总结 + 观点：Reference and an example for the Bluetooth API f…｜中文观点：如果 anthropics/claude-desktop-buddy 能减少集成成本、维护…

总结 + 观点：Neural Networks: Zero to Hero｜中文观点：karpathy/nn-zero-to-hero 更值得从实际采用价值来判断，而不是只看它…

总结 + 观点：Introducing GPT-5.5, our smartest model yet—fast…｜中文观点：Introducing GPT-5.5 的核心不在新鲜感，而在它是否能提升工程效率、部署稳…

总结 + 观点：GPT-5.5 System Card｜中文观点：GPT-5.5 System Card 更值得从实际采用价值来判断，而不是只看它有没有制造…

openai/openai-node

来源：GitHub openai

标签：#github_orgs #extended

作者：

原文：Official JavaScript TypeScript library for the OpenAI API

链接：https://github.com/openai/openai-node

观点：对 openai/openai-node 来说，更值得判断的是它会不会进入团队默认工具链，而不是短期讨论热度。

anthropics/knowledge-work-plugins

来源：GitHub anthropics

标签：#github_orgs #extended

作者：

原文：Open source repository of plugins primarily intended for knowledge workers to use in Claude Cowork

链接：https://github.com/anthropics/knowledge-work-plugins

观点：anthropics/knowledge-work-plugins 的核心不在新鲜感，而在它是否能提升工程效率、部署稳定性或开发者工作流。

karpathy/llama2.c

来源：GitHub karpathy

标签：#github_orgs #extended

作者：

原文：Inference Llama 2 in one file of pure C

链接：https://github.com/karpathy/llama2.c

观点：karpathy/llama2.c 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

openai/privacy-filter

来源：GitHub openai

标签：#github_orgs #extended

作者：

原文：OpenAI Privacy Filter

链接：https://github.com/openai/privacy-filter

观点：openai/privacy-filter 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

karpathy/llm-council

来源：GitHub karpathy

标签：#github_orgs #extended

作者：

原文：LLM Council works together to answer your hardest questions

链接：https://github.com/karpathy/llm-council

观点：karpathy/llm-council 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

karpathy/micrograd

来源：GitHub karpathy

标签：#github_orgs #extended

作者：

原文：A tiny scalar-valued autograd engine and a neural net library on top of it with PyTorch-like API

链接：https://github.com/karpathy/micrograd

观点：对 karpathy/micrograd 来说，更值得判断的是它会不会进入团队默认工具链，而不是短期讨论热度。

anthropics/claude-desktop-buddy

来源：GitHub anthropics

标签：#github_orgs #extended

作者：

原文：Reference and an example for the Bluetooth API for makers in Claude Cowork Claude Code Desktop

链接：https://github.com/anthropics/claude-desktop-buddy

观点：如果 anthropics/claude-desktop-buddy 能减少集成成本、维护负担和迁移阻力，它才有进入生产栈的价值。

karpathy/nn-zero-to-hero

来源：GitHub karpathy

标签：#github_orgs #extended

作者：

原文：Neural Networks: Zero to Hero

链接：https://github.com/karpathy/nn-zero-to-hero

观点：karpathy/nn-zero-to-hero 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

Introducing GPT-5.5

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Introducing GPT-5.5, our smartest model yet—faster, more capable, and built for complex tasks like coding, research, and data analysis across tools.

链接：https://openai.com/index/introducing-gpt-5-5

观点：Introducing GPT-5.5 的核心不在新鲜感，而在它是否能提升工程效率、部署稳定性或开发者工作流。

GPT-5.5 System Card

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：GPT-5.5 System Card

链接：https://openai.com/index/gpt-5-5-system-card

观点：GPT-5.5 System Card 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

Automations

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Learn how to automate tasks in Codex using schedules and triggers to create reports, summaries, and recurring workflows without manual effort.

链接：https://openai.com/academy/codex-automations

观点：对 Automations，更该看它能不能改善多步骤协作、记忆管理和稳定交付，而不是只看 demo 效果。

Top 10 uses for Codex at work

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Explore 10 practical Codex use cases to automate tasks, create deliverables, and turn real inputs into outputs across tools, files, and workflows.

链接：https://openai.com/academy/top-10-use-cases-codex-for-work

观点：对 Top 10 uses for Codex at work，更该看它能不能改善多步骤协作、记忆管理和稳定交付，而不是只看 demo 效果。

Plugins and skills

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Learn how to use Codex plugins and skills to connect tools, access data, and follow repeatable workflows to automate tasks and improve results.

链接：https://openai.com/academy/codex-plugins-and-skills

观点：对 Plugins and skills，更该看它能不能改善多步骤协作、记忆管理和稳定交付，而不是只看 demo 效果。

Working with Codex

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Learn how to set up your Codex workspace, create threads and projects, manage files, and start completing tasks with step-by-step guidance.

链接：https://openai.com/academy/working-with-codex

观点：Working with Codex 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

Codex settings

来源：OpenAI Blog

标签：#ai_engineering_blogs #core

作者：

原文：Learn how to configure Codex settings, including personalization, detail level, and permissions, to run tasks smoothly and customize your workflow.

链接：https://openai.com/academy/codex-settings

观点：对 Codex settings，更该看它能不能改善多步骤协作、记忆管理和稳定交付，而不是只看 demo 效果。

Uncle Bob: It's Over

来源：Hacker News Frontpage

标签：#research_community #core

作者：

原文：Article URL: https://old.reddit.com/r/vibecoding/comments/1srfqm0/uncle_bob_its_over/ Comments URL: https://news.ycombinator.com/item?id=47998601 Points: 27 Comments: 16

链接：https://old.reddit.com/r/vibecoding/comments/1srfqm0/uncle_bob_its_over/

观点：Uncle Bob: It's Over 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

For thirty years I programmed with Phish on, every day

来源：Hacker News Frontpage

标签：#research_community #core

作者：

原文：Article URL: https://christophermeiklejohn.com/ai/personal/phish/flow/agents/2026/05/03/rift.html Comments URL: https://news.ycombinator.com/item?id=47998225 Points: 75 Comments: 60

链接：https://christophermeiklejohn.com/ai/personal/phish/flow/agents/2026/05/03/rift.html

观点：For thirty years I programmed with Phish on, every day 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

A Desktop Made for One

来源：Hacker News Frontpage

标签：#research_community #core

作者：

原文：Article URL: https://isene.org/2026/05/Audience-of-One.html Comments URL: https://news.ycombinator.com/item?id=47997947 Points: 27 Comments: 8

链接：https://isene.org/2026/05/Audience-of-One.html

观点：A Desktop Made for One 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

Quoting Anthropic

来源：Simon Willison

标签：#ai_engineering_blogs #core

作者：

原文：We used an automatic classifier which judged sycophancy by looking at whether Claude showed a willingness to push back, maintain positions when challenged, give praise proportional to the merit of ideas, and speak frankly regardless of what a person wants to hear. Most of the time in these situations, Claude expressed no sycophancy—only 9% of conversations included sycophantic behavior (Figure 2). But two domains were exceptions: we saw sycophantic behavior in 38% of conversations focused on spirituality, and 25% of conversations on relationships. -- Anthropic How people ask Claude for personal guidance Tags: ai-ethics anthropic claude ai-personality generative-ai ai llms sycophancy

链接：https://simonwillison.net/2026/May/3/anthropic/#atom-everything

观点：Quoting Anthropic 更值得从实际采用价值来判断，而不是只看它有没有制造新的讨论热度。

Security Through Obscurity Is Not Bad

来源：Hacker News Frontpage

标签：#research_community #core

作者：

原文：Article URL: https://mobeigi.com/blog/security/security-through-obscurity-is-not-bad/ Comments URL: https://news.ycombinator.com/item?id=47997486 Points: 44 Comments: 48

链接：https://mobeigi.com/blog/security/security-through-obscurity-is-not-bad/

观点：从 Security Through Obscurity Is Not Bad 看，后续更应关注安全事故是否改变企业采购、接入和上线前的合规门槛。