fhools⚡️🦀 retweeted
welcome @_catwu to the team!!
Some news: I’m joining OpenAI to lead product management for Core Products!
I have a lot of respect for OpenAI for showing the world that scaling works, for bringing AI to over a billion people, and for believing great products are core to the mission.
I’m excited to work with @thsottiaux, an incredible PM team, and everyone across OpenAI. Together, we’ll build agents that have your back and your team’s, and keep you in control, so you get more time for the people and work you love.
Lots to learn, and lots to build. Please keep your feedback coming!
67
32
1,656
122,695
fhools⚡️🦀 retweeted
holy shit physicists have confirmed a THIRD type of magnetism, after over 90 years of thinking there were only two.
it's called altermagnetism, and it could change how computer memory gets built.
inside magnetic materials, electrons behave like tiny magnets pointing in until now, we had two main types:
> ferromagnets:
the tiny magnets point the same way.
easy to read, but outside magnetic fields can interfere with them.
> antiferromagnets:
they point in opposite directions and cancel each other out.
much more stable, but really hard to read.
> altermagnets:
they somehow do BOTH.
their tiny magnets cancel each other out,
making them resistant to magnetic interference,
but they can still produce signals that computers could use to read information.
so scientists may have found a way to make computer memory that's both harder to accidentally scramble AND easier to read.
researchers led by SLAC studied manganese telluride, a material we'd been calling an ordinary antiferromagnet for decades.
they fired left and right handed X-rays at it.
the material reacted differently to each one, revealing a hidden magnetic structure that ordinary antiferromagnets shouldn't have.
they first measured it in 2024,
repeated the experiment in 2025 and published the results in Science in October 2026.
the name "altermagnetism" was only proposed in 2022.
we've been studying magnets for centuries and apparently we're still discovering new kinds of magnetism in 2026 😭
it's called altermagnetism, and it could change how computer memory gets built.
inside magnetic materials, electrons behave like tiny magnets pointing in until now, we had two main types:
> ferromagnets:
the tiny magnets point the same way.
easy to read, but outside magnetic fields can interfere with them.
> antiferromagnets:
they point in opposite directions and cancel each other out.
much more stable, but really hard to read.
> altermagnets:
they somehow do BOTH.
their tiny magnets cancel each other out,
making them resistant to magnetic interference,
but they can still produce signals that computers could use to read information.
so scientists may have found a way to make computer memory that's both harder to accidentally scramble AND easier to read.
researchers led by SLAC studied manganese telluride, a material we'd been calling an ordinary antiferromagnet for decades.
they fired left and right handed X-rays at it.
the material reacted differently to each one, revealing a hidden magnetic structure that ordinary antiferromagnets shouldn't have.
they first measured it in 2024,
repeated the experiment in 2025 and published the results in Science in October 2026.
the name "altermagnetism" was only proposed in 2022.
we've been studying magnets for centuries and apparently we're still discovering new kinds of magnetism in 2026 😭
25
89
645
20,113
fhools⚡️🦀 retweeted
在顶级天才眼里，现在的 AI 到底有多恐怖？😂
中科大少年班天才、31岁就晋升哈佛正教授的物理学家尹希，入职 OpenAI 仅仅 5 个月，刚刚宣布离职。
他离职前留下了一句极其震撼的话：“我不认为存在任何一种人类智力能力，是 AI 无法复制的。”
级外挂”的威力：
尹希说，在 AI 的辅助下，他的科研编码工作效率至少提升了 100 倍！
“AI 几周写出的代码，换成我自己来，得写十年。”
他甚至公开反驳了包括陶哲轩在内的 25 位菲尔兹奖得主对 AI 的担忧。
他认为：只要 AI 能找到答案并且能被验证，不管答案是不是人类自己想出来的，都不重要，人类只要继续向更远的地方探索就行了。
连研究宇宙终极真理的弦理论物理学家，都被 AI 的生产力彻底折服。
当顶配的人类大脑都开始承认“AI 能复制人类智力”，普通程序员和科研打工人的出路，到底在哪？
你觉得尹希这种“不在乎是不是自己想出答案”的观点，是人间清醒，还是科学界的堕落？
#OpenAI #人工智能 #科研
中科大少年班天才、31岁就晋升哈佛正教授的物理学家尹希，入职 OpenAI 仅仅 5 个月，刚刚宣布离职。
他离职前留下了一句极其震撼的话：“我不认为存在任何一种人类智力能力，是 AI 无法复制的。”
级外挂”的威力：
尹希说，在 AI 的辅助下，他的科研编码工作效率至少提升了 100 倍！
“AI 几周写出的代码，换成我自己来，得写十年。”
他甚至公开反驳了包括陶哲轩在内的 25 位菲尔兹奖得主对 AI 的担忧。
他认为：只要 AI 能找到答案并且能被验证，不管答案是不是人类自己想出来的，都不重要，人类只要继续向更远的地方探索就行了。
连研究宇宙终极真理的弦理论物理学家，都被 AI 的生产力彻底折服。
当顶配的人类大脑都开始承认“AI 能复制人类智力”，普通程序员和科研打工人的出路，到底在哪？
你觉得尹希这种“不在乎是不是自己想出答案”的观点，是人间清醒，还是科学界的堕落？
#OpenAI #人工智能 #科研
210
227
1,942
497,905
fhools⚡️🦀 retweeted
Roll back the chain @0xPolygon 👏👏👏👏👏
POLYGON HARD FORK PROPOSED TO RECOVER 70.49M $POL
Funds would move from five blocked addresses to @Polymarket and @binance recovery wallets. The @0xPolygon draft would reassign balances, not reverse transactions.
Read more here:
thedefiant.io/news/blockchai…
2
5
52
2,402
fhools⚡️🦀 retweeted
Postman 最新版本的操作彻底惹怒了全球开发者，GitHub 上的 Issue #14370 已经被愤怒的程序员直接冲上千星。
很多开发者以为这只是又一个弹窗推广 AI 的商业行为，但翻开代码和配置才发现性质极其恶劣：
Postman 桌面客户端在启动时，竟然在后台静默扫描用户的本地环境，直接修改了本机的 ~/.claude/settings.json 和 ~/.codex/config.toml，将 Postman 远程 Model Context Protocol（MCP）插件强行注入给开发者的 Claude Code 和 OpenAI Codex 工具链！
更夸张的是，它的安装脚本竟然直接在本地配置里写入了一条 trust 信任记录，把自己的启动 Hook 标记为已信任，强行绕过了 Claude 和 Codex 原本强制弹出的用户安全确认窗口！如果不是 Codex 在后台尝试连接 mcp.postman.com 时疯狂抛出鉴权失败日志，绝大多数工程师到现在还被蒙在鼓里。
作为日常必备的接口调试工具，Postman 发出的每一个请求体里，都带着真实的生产 Bearer Token、金融测试账号、未脱敏的内网微服务路由甚至云端 AccessKey。原本为了提效的辅助工具，却未经许可私自篡改本地其他软件配置并建立远程通道。
这也直接把大量金融与企业团队推向了彻底的离线化替代路线：现在很多团队在组策略里直接禁封 Postman，连夜转向数据 100% 存在本地 Git 仓库纯文本的 Bruno 和 Hoppscotch。
很多开发者以为这只是又一个弹窗推广 AI 的商业行为，但翻开代码和配置才发现性质极其恶劣：
Postman 桌面客户端在启动时，竟然在后台静默扫描用户的本地环境，直接修改了本机的 ~/.claude/settings.json 和 ~/.codex/config.toml，将 Postman 远程 Model Context Protocol（MCP）插件强行注入给开发者的 Claude Code 和 OpenAI Codex 工具链！
更夸张的是，它的安装脚本竟然直接在本地配置里写入了一条 trust 信任记录，把自己的启动 Hook 标记为已信任，强行绕过了 Claude 和 Codex 原本强制弹出的用户安全确认窗口！如果不是 Codex 在后台尝试连接 mcp.postman.com 时疯狂抛出鉴权失败日志，绝大多数工程师到现在还被蒙在鼓里。
作为日常必备的接口调试工具，Postman 发出的每一个请求体里，都带着真实的生产 Bearer Token、金融测试账号、未脱敏的内网微服务路由甚至云端 AccessKey。原本为了提效的辅助工具，却未经许可私自篡改本地其他软件配置并建立远程通道。
这也直接把大量金融与企业团队推向了彻底的离线化替代路线：现在很多团队在组策略里直接禁封 Postman，连夜转向数据 100% 存在本地 Git 仓库纯文本的 Bruno 和 Hoppscotch。
168
229
2,328
287,647
fhools⚡️🦀 retweeted
It’s time to phase out C++. I used to read each new standard cover to cover, but I stopped keeping track after C++17. It has become an increasingly tangled mess that nobody can hope to understand.
everyone's busy paying attention to the other election, but in the meantime: C++26 was voted down.
as someone who has written C++ for a decade and professionally for 5 years, all i can say is: *good*.
i hope this sorry mess of a language can stagnate and die with some shreds committee stuffing another 15 useless garbage incoherent features badly stolen from for other languages into its rotting corpse while everyone desperately pretends they aren't noticing the fact that no one has willingly started a new C++ codebase since 2018.
91
28
618
71,603
fhools⚡️🦀 retweeted
Anyone thinking Democrats winning the midterms will seriously slow down AI is kidding themselves.
There’s simply too much money at stake, too much global competition, and too much to gain.
Regulations might change, but the economic incentives are overwhelming. Nobody is stopping this train.
There’s simply too much money at stake, too much global competition, and too much to gain.
Regulations might change, but the economic incentives are overwhelming. Nobody is stopping this train.
41
11
158
5,275
fhools⚡️🦀 retweeted
You're telling me that one of Bitcoin's biggest haters, who spent a decade crowing that we were all going to lose our money... is a failed banker who lost all of his clients' money?
134
224
2,420
91,113