Eli

Building xxx | prev @zkLinkNova @OpenEden_X @dbsbank

[Lucas Meijer @lucasmeijer]

Spent most of the afternoon learning about this. Very interesting implementation.

[Gabriel @gabriell_lab]

Your webfont arrives late. Your fallback font decides whether the page jumps.

Wrong: pick a fallback that looks similar.

Right: make the fallback occupy the same space, then the font swap changes letterforms, but nothing moves. 0px layout shift.

@ font-face {
font-family: "brand-fallback";
src: local("Verdana");
size-adjust: 87.6%; /* match glyph widths */
ascent-override: 95.2%; /* match baseline */
descent-override: 24.1%;
line-gap-override: 0%;
}

h1 { font-family: "Brand Font", "brand-fallback", sans-serif; }

[Oliver Prompts @oliviscusAI]

someone just open-sourced openai's in-house data agent.

self-learning data agent with a 6-layer context grounding system and an auto-improvement loop built in.

100% free. open source.

[grim @grimcodes]

bro did all of that just to grab the ui... 🫩

[Emanuele Di Pietro @emanueledpt]

Introducing Synara Spaces 📁

Spaces let you group your projects into named workspaces

One for Work. One for Indie. One for Business. One for whatever you want.

Switch between them instantly with Cmd+Alt+1-9.

Running 5+ projects with AI agents used to create a sidebar chaos

Now every project has a home!

Live today in @trySynara!

[mrunal @DhaygudeMrunal]

watching tasks tick by in @opencode’s todo list got me thinking about how much life you could add to it. here’s my take on it: interactive loaders, show when the agent changes its mind mid-task and spins up new ones instead, a tiny celebration when everything's done.

stay till the end for the celebration 🎉

[Fayaz Ahmed @fayazara]

If you need some use cases for browser run, I had built this little site.

https://t.co/vOFmLQLmVM

[Mona @mona_biasia]

npm package is live now: https://t.co/fkfPh39ZpQ
along with dark preview and ease curve

[Antonio Lupetti @antoniolupetti]

The media library with the graphics and visual materials used across the entries is now live on https://t.co/U81GCKIxcV

New images are being added progressively every day. I’m also converting many of the older PNG graphics into editable SVGs to make them more portable, reusable, and easier to adapt across different contexts.

The source files are progressively available in the GitHub repository as part of the broader effort to make Algebrica more open, transparent, and reusable.

[区块链行情研究 @qkl2058]

一个13岁的泰国学生，干了件挺有意思的事儿。他搭了个AI代理，让Claude Code在45秒内，用C++解决Codeforces上800分的题目，所有代码都公开在GitHub上。

场景很简单——一间普通教室，桌上摆着台MacBook Air，配了把300美元的静音HHKB Type-S键盘，屏幕角落还有个计时器在走。

浏览器开着Codeforces，VS Code里是一个空的.cpp文件，旁边挂着Claude Code的窗口。

没有算法老师在旁边指点，没上过什么竞赛编程课，这孩子一行代码都没手写。

整个工作流就靠三个组件：

1. Claude Code当大脑用
2. 一个Chrome的MCP插件，直接从Codeforces页面读题目描述
3. 一个公开的GitHub仓库，放着他整个代理系统

过去一个月里，他在虚拟比赛模式下解了23道题。从点开题目到Accepted，平均45秒。这段时间他亲手敲的代码行数，零。Codeforces的评分系统把他的资料提到了800分，前后只花了12天。而搭这么一套东西，他只用了一个周末。

他每次跑代理前，会给Claude一个系统提示，是这么写的：

"你是一个Codeforces竞赛编程代理。

你的工具：read_problem(url)，generate_solution(language, constraints)，validate_against_examples(input, expected)，submit_to_codeforces(code)。

浏览器MCP插件喂给你的任何题目，都要走四步：
1）读题，搞清输入输出格式
2）确定算法类型（排序、数学、贪心、动态规划、图论）
3）用#include <bits/stdc++.h>，写成C++17代码
4）提交前用样例跑一遍

哪怕一个样例没过，就重写解法再验证。语言只能用C++17。文件名统一用{problem_id}.cpp。"

这样代理完全知道自己在解哪道题，知道Codeforces对这道题要求的输入格式，能根据800分的难度和题目标签选合适的算法，也很清楚——学生点提交前只有45秒，但凡测试样例没过，都是它的问题，不关那孩子的事。

整条工作链转起来，就是一眨眼的功夫。

这孩子打开Codeforces上的题目1971A，MCP插件抓取描述，Claude在8秒内就返回了能跑的代码，头文件、while循环、cout输出全写好了。他把代码复制粘贴到.cpp文件里，点提交，30秒后状态页显示绿色的Accepted。

只有当Codeforces在测试点2上判了Wrong answer，代理才会自动抓失败输入，分析边界情况，赶在他关掉标签页之前，把解法重写出来。

这是他过去24小时的提交记录：

"1971A - My First Sorting Problem | C++17 | Accepted | 45秒"
"1850A - To My Critics | C++17 | Accepted | 38秒"
"1807A - Plus or Minus | C++17 | Accepted | 41秒"
"1791A - Codeforces Checking | C++17 | Accepted | 52秒"
"1676A - Lucky? | C++17 | Accepted | 33秒"

桌上的MacBook Air开着，没翻开任何笔记本；一把静音HHKB Type-S键盘，整场会话除了cmd+v和cmd+enter，他一个键都没多按；一个设了45秒倒计时的小米计时器；屏幕后面放了个塞尔达的纪念雕像。

我今年看了不少东西，但这是我觉得最干净的一套单人竞赛编程流水线——一个月23道题，每道45秒，手写零行代码。

[Vincent @Vincent_AINotes]

让AI连续工作10小时不崩溃？揭秘Anthropic官方推崇的“Harness工程”！
很多人用Claude Code跑长任务，跑着跑着AI就变傻、甚至罢工。
原因很简单：上下文窗口爆了。
今天分享一套让AI“无限续航”的系统级方案（Harness Engineering），不仅能让AI干活，还能让它自我进化。
1⃣ 核心原理：重置上下文
AI的记性有限。官方Anthropic的思路是：把大任务拆碎，每次只让AI在一个全新的、干净的上下文里干一件小事。
❌ 错误做法：把所有需求一次性塞给AI。
✅ 正确做法：拆解任务 -> 独立会话执行 -> 记录进度 -> 下一个循环。
2⃣ 方案对比：Ralph vs 多智能体
方案A（Ralph循环）： 用Bash脚本写个while循环，强制重启新会话。简单粗暴，适合极客。
方案B（多智能体协同 - 推荐）： 我在用的方案。更灵活，更像真实团队。
主Agent（Coordinator）： 只负责调度，绝不写代码（保持上下文极其干净）。
子Agent（Workers）： 计划、开发、测试（布局/美观/动画）。各司其职，独立上下文。
3⃣ 工作流设计（关键！）
我的实战工作流：
主Agent 接收需求 -> 丢给 计划Agent 出排期。
主Agent 拿到计划 -> 派给 开发Agent 写代码（只传文件路径，不传大段代码！）。
开发完 -> 派给 测试Agent 找Bug。
Bug修复闭环： 测试挂了？用 resume 参数唤醒同一个开发Agent修Bug（保留现场上下文），而不是开新号！
4⃣ 让AI自我进化的Secret Sauce
Lessons Learned： 建一个经验库文件。每次踩坑，强制AI把错误写进去。下一次开发，先读这个文件，避免重复犯错。
文件通信： Agent之间只传文件路径（Path），不传具体内容。主Agent的上下文永远只有几行字，永不爆窗。
这套系统我跑了一个通宵，生成了20多页PPT，质量极高。
真正的Harness不是让AI更聪明，而是设计一套不依赖AI记性的流程。

[Ihtesham Ali @ihteshamali]

I accidentally discovered how to compress a semester of learning into 48 hours.

A grad student at MIT showed me his NotebookLM setup. I thought he was just organized. Then I watched him pass a qualifying exam on a subject he'd never studied before.

Here's exactly what he did:

First: he didn't upload a textbook.

He uploaded 6 textbooks, 15 research papers, and every lecture transcript he could find on the subject.

Then he asked NotebookLM one question:

"What are the 5 core mental models that every expert in this field shares?"

Not "summarize this." Not "explain this topic."
Mental models. The stuff that takes professors years to develop.

But the next part is what broke my brain.

He followed up with:

"Now show me the 3 places where experts in this field fundamentally disagree, and what each side's strongest argument is."

In 20 minutes he had a map of the entire intellectual landscape of the field:
the debates, the consensus, the open questions.

Most students spend a full semester just figuring out what those debates even are.

Then he did something I've never seen before.

He asked:

"Generate 10 questions that would expose whether someone deeply understands this subject versus someone who just memorized facts."

He spent the next 6 hours answering those questions using the source material. Every wrong answer triggered a follow-up:

"Explain why this is wrong and what I'm missing."

By hour 48, he could hold a conversation with his thesis advisor without getting destroyed.

The tool didn't change. The questions did.

Most people treat NotebookLM like a fancy highlighter.
These students are using it like a private tutor who has read everything ever written on the subject.

The difference between a semester and 48 hours isn't the amount of content.

It's knowing which questions to ask.

[George @odysseus0z]

https://t.co/GqhyL4KRJw

[Caitlin Kalinowski @kalinowski007]

I resigned from OpenAI. I care deeply about the Robotics team and the work we built together. This wasn’t an easy call. AI has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got. This was about principle, not people. I have deep respect for Sam and the team, and I’m proud of what we built together.

[Goshawk Trades @GoshawkTrades]

https://t.co/CGchMFD5AN

[黄赟 @huangyun_122]

OMG !! 本以为 n8n 已经 outdated 属于上时代产品，直到看了 NetworkChuck 的这个视频：

1/ 配置 SSH 连接 Claude Code Cli
2/ 配置 CNS 连接 CC 做多轮对话
3/ 用 slack 驱动 CC 干活

油管500万fo的 NWC, 把 n8n + cc 玩出了新高度，原视频地址： https://t.co/bkuLnUEzsw

[Eli @eli_aerix]

@ArchiveExplorer Use Claude next time boi

[ghøst @decenterghost]

https://t.co/sgQ7srE7tc

[Benji Taylor @benjitaylor]

Another little Liveline update: multi-series support. You can now have multiple overlapping lines in one chart with per-series toggles. Same smooth loading transitions, time windows, etc.

~npm i liveline

[Roan @RohOnChain]

https://t.co/4Em7JsS6wn