yizheng.eth(周易正💹🧲)

Hermes Agent 架构的文档推荐直接看官方的文档，写的还算清楚。 然后用 Codex 或者 Claude Code 打开项目代码库，直接让 Agent 给你解释代码库，如果不清楚的你随时追问，这样比较好就是你可以问任何你想知道的问题，它会通过检索项目文档和代码，帮你解释的清清楚楚。 官方文档：hermes-agent.nousresearch.com/docs/developer…

PLEASE BLOCK THIS USER. they have been continuously promoting ador. it's ppl like this who do not need extra promo for spreading the agenda of a label that's responsible for kicking dani out and suing her. thank you.

Anthropic 的哲学家@AmandaAskell 最近参加了一个访谈，在访谈中她分享了自己探索好奇领域的一个方法。 提示词大概是： 我希望你从「xx」领域里选一个大概研究生水平的概念。然后我希望你通过写一个寓言的方式，间接地把这个概念完整讲出来。最好一直到快结尾时，人才会慢慢意识到这个概念究竟是什么。然后在故事之后，再补一段解释，把你刚才真正要讲的概念说清楚。

this OpenClaw bot watches NASA wildfire satellites. when a fire starts, it finds every at-risk home nearby, renders fire hardening fireproofing upgrade on their actual house, and mails them a postcard, all on autopilot. here's how contractors in fire zones can close $30k–$60k fireproofing jobs with this: - pulls live fire detections from NASA FIRMS satellites every 5 minutes - finds every home in the fire perimeter from public records - captures the property via Google Street View - classifies roof type + fire vulnerabilities using AI vision - generates a NASA satellite image of the neighborhood with real fire hotspots overlaid - renders a fire hardening retrofit on their actual house - prints a postcard with the NASA fire map + before/after retrofit + QR code every step from fire detection to mailbox runs without a human reply "FIRE" + RT and i'll send you the full guide so you can build this too (must be following so i can DM)

@everestchris6 Fire

You can now fine-tune Gemma 4 31B for free in our notebook! 🚀 Training the 31B parameter model is completely free with Kaggle and Unsloth. GitHub: github.com/unslothai/unsl… Guide: unsloth.ai/docs/models/ge… Gemma-4-31B Notebook: kaggle.com/code/danielhan…

🚨An Obsidian vault template that gives Claude Code persistent memory across every session -- your goals, decisions, wins, and context -- all loaded automatically before Claude writes a single line. It's called Obsidian Mind. And it's not a note-taking setup. It's an external brain designed specifically for engineers using Claude Code as a thinking partner. Every session starts with Claude already knowing your current projects, focus areas, past decisions, and patterns -- no re-explaining, no wasted turns. Here's how the system works: → CLAUDE.md -- operating manual loaded automatically every session, defines session start and end workflows → North Star.md -- your goals and focus areas, grounds every suggestion Claude makes → Memories.md -- persistent knowledge Claude retains and updates across sessions → SessionStart Hook -- injects the full vault file listing into every Claude session so Claude knows what exists from the first message → Work notes -- link to decisions, link to competencies, link to evidence in the brag doc → Brag Doc -- running log of wins, linked to the work notes that prove them → Competency notes -- one per skill from your org's framework, backlinks accumulate as you ship Here's the wildest part: Performance review prep becomes reading the backlinks panel on each competency note. Every work note you linked to a competency over the quarter is right there. The evidence graph builds itself as you work -- not at review time. Clone it, open it as an Obsidian vault, run `claude` in the directory, fill in North Star.md. That's the setup. 100% Open Source. MIT License. (Link in the comments)

分析claude code源代码第一步，先跑起来。这次泄漏的代码只是source map还原后的结果，缺少很多脚手架和私有package。本来我打算自己搞一下，结果发现已经有好心人搞定了。 github.com/claude-code-be… （我运行时有点小bug，把错误信息发给codex就搞定了） 尽早fork下载本地，估计存活不了太久😂

I wanted to share a bunch of my favorite hidden and under-utilized features in Claude Code. I'll focus on the ones I use the most. Here goes.

NEW AI report from Google. Every prior intelligence explosion in human history was social, not individual. These authors make the case that the AI "singularity" framed as a single superintelligent mind bootstrapping to godlike intelligence is fundamentally wrong. This is directly relevant to anyone designing multi-agent systems. They observe that frontier reasoning models like DeepSeek-R1 spontaneously develop internal "societies of thought," multi-agent debates among cognitive perspectives, through RL alone. The path forward is human-AI configurations and agent institutions, not bigger monolithic oracles. This reframes AI scaling strategy from "build bigger models" to "compose richer social systems." It argues governance of AI agents should follow institutional design principles, checks and balances, role protocols, rather than individual alignment. Paper: arxiv.org/abs/2603.20639 Learn to build effective AI agents in our academy: academy.dair.ai

.@milady_bsc launches the first x402 payment facilitator on BNB Chain! This lets ai agents pay merchants using @UTechStables without needing approvals every time, while keeping everything non-custodial and under your control Note: This post is for informational purposes only and not financial advice. DYOR.

@NotebookLM @stevenbjohnson Finally.

For everyone who saw our basketball post and thought, "I'd rather be in the library..." this one's for you. One of our MOST requested features: You can now upload ePub files as sources 🥳! Time to bring your books, study guides, and novels to life. What are you uploading first?

We have OpenClaw at home

Kimi直接集成OpenClaw，这招狠不狠

New Engineering blog: We tasked Opus 4.6 using agent teams to build a C compiler. Then we (mostly) walked away. Two weeks later, it worked on the Linux kernel. Here's what it taught us about the future of autonomous software development. Read more: anthropic.com/engineering/bu…

Prompt 1: The Expert Synthesizer "You are a [field] expert with 15 years of experience. Analyze these sources and identify the 3 core insights that practitioners in this field would immediately recognize as groundbreaking. For each insight, explain why it matters and what conventional wisdom it challenges." This forces depth over breadth. The output is immediately usable.

@crypto_oldwang @grok 整理評論區16个提示發给我

我不明白为什么人们不用Gemini来做股票交易。 每个人都有图表。每个人都有新闻。很少有人有清晰的思维。Gemini帮助解决最后这部分。 以下是16个实用提示，用于股票交易和投资：

@heygurisingh Claude

Most “Claude tutorials” are useless. They teach prompts. Not systems. So I packaged everything into one guide and I’m giving it away FREE for 24 hours: Ultimate Claude Mastery Guide by Guri Singh 80+ Chapters 1000+ Tools 2000+ Prompts Reply “Claude” and I’ll DM it. Must follow to receive.

@zstmfhy 一定要的

没想到那个Z-Library一键同步到 NotebookLM 的 skill 这么火， 我当时直接把最大痛点甩出来： “下书是爽了，但手动下载PDF/EPUB + 再一步步上传到NotebookLM，文件一大就卡死、限制还多、折腾到吐血……” 结果没想到爆了20多万浏览，评论区直接沦陷，全是“想要”、“+1” 更离谱的是，NotebookLM官方账号 @NotebookLM 居然亲自跑来回复了！ 就仨字：😍😍😍（截图放下面，大家感受下官方三个爱心砸脸的快乐～😍）这含金量，简直间接盖章了吧！开心到当场起飞！ 好消息是：这次把矛头对准了影子图书馆里的“究极Boss”——Anna's 1⃣ 直接扔一个 Annas 书籍链接（或其他合法来源） 2⃣ 自动下载 PDF/EPUB• 一键新建 NotebookLM 笔记本 + 上传文件 📷 3⃣ 完事后 Claude 就能对话式“读”整本书，问啥答啥，还带引用来源～学术党、自学者、爱读书的大佬们，你们有没有这个痛点？ 有需求的朋友点个赞/回复“要！”或者“+1”，我看反馈多的话，马上开源仓库 + 写详细教程！（合规提醒：只用于合法免费资源哦～）