科技推特精选 - 2026年5月24日

2026年5月24日科技每日简报

Today's top tech conversations are led by @dkundel, whose post about 'If you want to build Codex int...' garnered the highest engagement. Key themes trending across the top stories include mythos, https, claude, anthropic, about. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. dkundel (Group Score: 179.8 | Individual: 45.2)

Cluster: 6 tweets | Engagement: 467 (Avg: 77) | Type: Tech

If you want to build Codex into your app just point Codex at https://t.co/6brLq4EFwU and let it handle the rest ❤️ fully open-source incl. sign in with ChatGPT\n\nQT @thsottiaux: A little secret. About 5% of our production traffic is on the Pi harness, about another 5% is on OpenCode. Reminder you can use your ChatGPT account in a flourishing set of other tools.

We’ll continue to make Codex awesome, but you have options.

See 5 related tweets

@thsottiaux: A little secret. About 5% of our production traffic is on the Pi harness, about another 5% is on Ope...
@reach_vb: PSA: You can use your ChatGPT account to use Codex where you work - OpenCode, Pi, OpenClaw and more....
@rileybrown: This is the way.\n\nQT @thsottiaux: A little secret. About 5% of our production traffic is on the Pi...
@0xSero: You can make a religion out of this https://t.co/KnghDE0IuL\n\nQT @thsottiaux: A little secret. Abou...
@pmarca: Interesting.\n\nQT @thsottiaux: A little secret. About 5% of our production traffic is on the Pi har...

2. peterwildeford (Group Score: 143.1 | Individual: 60.7)

Cluster: 5 tweets | Engagement: 1149 (Avg: 86) | Type: Tech

Claude Mythos alone is finding more vulnerabilities than were found from all sources combined in prior years 👀 https://t.co/QcDBGo12Zh\n\nQT @AnthropicAI: Patching these vulnerabilities will make us safer. But the software industry will need to adapt to the volume of vulnerabilities that models like Claude Mythos Preview will be able to find.

We discuss this in our initial update on Project Glasswing: https://t.co/3cSgHHZXgG

See 4 related tweets

@scaling01: Claude Mythos absolutely destroys GPT-5.5 in ExploitBench and ExploitGym

Mythos finds 18 arbitrary ...

@Techmeme: Anthropic says Claude Mythos Preview has been used to find more than 10,000 high- or critical-severi...
@zephyr_z9: Very good\n\nQT @peterwildeford: Claude Mythos alone is finding more vulnerabilities than were found...
@Cointelegraph: 🚨 UPDATE: Anthropic’s Project Glasswing uncovered over 10,000 high- and critical-severity software v...

3. SemiAnalysis_ (Group Score: 138.9 | Individual: 42.5)

Cluster: 4 tweets | Engagement: 774 (Avg: 227) | Type: Tech

FACT ALERT 🚨 : In modern agentic coding, 42% of the time is spent on CPU doing tool use such as editing files, running Bash scripts, running lints, etc. The economy of traditional cloud computing charges at $per cpu core. In the economy of agents, the business model is$ per token thus to increase token revenue, you need to increase the amount of CPUs power u have so that you can generate your tokens.

See 3 related tweets

@tunguz: Here is one big reason why this matters. Time spent on non-LLM inference tasks is only going to incr...
@tunguz: Very important.\n\nQT @SemiAnalysis_: FACT ALERT 🚨 : In modern agentic coding, 42% of the time is sp...
@firstadopter: Did someone say CPU CPU CPU?\n\nQT @SemiAnalysis_: FACT ALERT 🚨 : In modern agentic coding, 42% of t...

4. eng_khairallah1 (Group Score: 100.4 | Individual: 53.8)

Cluster: 3 tweets | Engagement: 1018 (Avg: 78) | Type: Tech

RT @sairahul1: Anthropic pays $750,000+ a year for engineers who know how to build LLMs from scratch.

Stanford just released the exact lecture that teaches it - 1 hour 44 minutes, free, straight from CS229.

Bookmark and watch it this weekend.

It'll teach you more about how ChatGPT & Claude actually work than most people at top AI companies learn in their entire careers.

See 2 related tweets

@sairahul1: Anthropic pays $750,000+ a year for engineers who know how to build LLMs from scratch.

Stanford jus...

@RoundtableSpace: STANFORD JUST DROPPED A FREE MASTERCLASS ON HOW LLMS ACTUALLY WORK

Covers the foundations behind ...

5. TheAhmadOsman (Group Score: 95.5 | Individual: 31.9)

Cluster: 3 tweets | Engagement: 80 (Avg: 135) | Type: Tech

UNBELIEVABLE RESOURCE

The bible for understanding LLMs is NOW AVAILABLE online to read (FOR FREE)

Covers all the concepts below, no experience needed and anyone from any background can understand it

Tokens / Tokenizers
Transformers
Attention
KV Cache
Prefill vs Decode
Decoding Controls / Samplers
Agents / Tools
Model Packages
Chat Templates
Long Context
Multimodality
RAG
Fine-tuning

Then connects that to running models locally

What "Local AI" Really Means
Open-weight vs Opensource
Quantization
VRAM Math
Hardware tiers
File formats / load safety
Runtimes / serving modes
Model selection
Privacy
Failure modes
Benchmarks
Practical setup paths

You should read this, and if you cannot now then you most definitely wanna bookmark it for later

Opensource / Local AI FTW\n\nQT @TheAhmadOsman: https://t.co/uRDNr16FHA

See 2 related tweets

@TheAhmadOsman: DROP EVERYTHING

The bible for how LLMs work is now available online to read FOR FREE

This is for y...

@ivanfioravanti: I’ll have to create a version for Apple Local LLMs only 🤷🏻‍♂️\n\nQT @TheAhmadOsman: UNBELIEVABLE RES...

6. chandrarsrikant (Group Score: 86.4 | Individual: 35.4)

Cluster: 3 tweets | Engagement: 118 (Avg: 1883) | Type: Tech

🚨MC Interview: High-Agency Generalists will define the AI era, says Claude Code co-creator Sid Bidasaria

Claude Code co-creator Sid Bidasaria is part of a growing cohort of Indian-origin leaders helping shape the future of Anthropic, one of the world’s most influential AI companies.

Other prominent Indian-origin executives at the company include CFO Krishna Rao and CTO Rahul Patil.

An alumnus of The Doon School, Dehradun, Bidasaria went on to earn a bachelor’s degree in computer engineering from University of Illinois Urbana-Champaign. Early in his career, he interned at Tech Mahindra, Impetus and Bank of America.

Before joining Anthropic in August 2024, he held engineering and product roles at companies including Citrix, Rubrik, Robinhood and Propelo.

When Anthropic launched its agentic coding tool Claude Code, the target audience was clear: engineers. But as non-technical workers increasingly adopt it for everyday business workflows, the company is expanding its horizon to cater to broader professional use cases.

Bidasaria spoke to Moneycontrol at the Anthropic HQ in Silicon Valley. He said AI is blurring the lines between engineers product managers, data scientists and designers.

By @tsuvik

https://t.co/xEXUJFx3TD

See 2 related tweets

@chandrarsrikant: "Previously, you had fairly distinct boundaries. Engineers wrote code, product managers focused on p...
@chandrarsrikant: 🚨 On AI edge this weekend, an interview with Sid Bidasaria: From Tech Mahindra intern to Claude Code...

7. PeterDiamandis (Group Score: 84.1 | Individual: 32.2)

Cluster: 3 tweets | Engagement: 3166 (Avg: 1147) | Type: Tech

Starship was the greatest thing humanity ever built without AI assistance. It will be the last. Everything after this will be designed with AI. The era of human-only engineering is over.

See 2 related tweets

@XFreeze: Starship is a marvel of engineering

What you’re seeing today is something others can only dream of ...

@XFreeze: Starship may end up being the last truly gigantic engineering marvel built entirely by human hands a...

8. trevin (Group Score: 83.7 | Individual: 56.4)

Cluster: 2 tweets | Engagement: 738 (Avg: 43) | Type: Tech

RT @perplexity_ai: Today we're open-sourcing Bumblebee, a read-only scanner for macOS and Linux.

It checks developer machines for risky packages, extensions, and AI tool configs.

Connected to Computer, it can trigger deeper scans whenever a new supply-chain risk emerges.

https://t.co/FOaWnF1yQy https://t.co/wXauD4wDOT

See 1 related tweets

@BrianRoemmele: We have been using the open source tool Bumblebee at The Zero-Human Company over the last day and fi...

9. synthwavedd (Group Score: 80.0 | Individual: 30.2)

Cluster: 3 tweets | Engagement: 1053 (Avg: 534) | Type: Tech

i am absolutely THRILLED to announce that it appears they're beginning to make strides on UI de-slopification with GPT-5.6! 🥹

here's your first look on a prompt without any UI guidance ("default") - we're getting somewhere... https://t.co/4LwkVaULWG\n\nQT @synthwavedd: iris-alpha

See 2 related tweets

@kimmonismus: Looks like GPT-5.6 release is very close. Really looking forward to it. 5.5 already is an insanely g...
@AndrewCurran_: RT @synthwavedd: i am absolutely THRILLED to announce that it appears they're beginning to make stri...

10. scaling01 (Group Score: 79.4 | Individual: 22.4)

Cluster: 4 tweets | Engagement: 298 (Avg: 196) | Type: Tech

The compute story is cope. The gatekeeping story is cope.

Mythos is genuinely much stronger than anything we've seen so far, and if Anthropic simply let it loose instead of starting Project Glasswing there would be millions-billions of dollars in damages.\n\nQT @scaling01: I don't understand how people are still coping about Mythos.

Here's a few benchmarks: SWE-bench Pro: Mythos -> 77.8%, GPT-5.5 -> 58.6% HLE: Mythos -> 56.8%, GPT-5.5 -> 41.4%

UK AISI cyber ranges:

"The Last Ones": Mythos -> 6/10, GPT-5.5 3/10
"Cooling Tower": Mythos -> 3/10, GPT-5.5 0/10

ExploitBench:

Mythos -> 18 Arbitrary Code Executions
GPT-5.5 -> 0 Arbitrary Code Executions

ExploitGym:

Mythos -> 157 exploits (289.3 LLM calls)
GPT-5.5 -> 120 exploits (375.4 LLM calls)

XBOW same story. Mythos has much higher odds of finding vulnerabilities within smaller token budgets.

See 3 related tweets

@scaling01: Anthropic is codemaxxing OpenAI is mathmaxxing

the question is, which is going to be more useful an...

@scaling01: and the fact that companies are using Mythos and not GPT-5.5 for finding vulnerabilities

the gove...

@ShanuMathew93: Elon went from calling Anthropic misanthropic and closing them to posting memes of monkeys saying My...

11. Reuters (Group Score: 78.8 | Individual: 29.9)

Cluster: 3 tweets | Engagement: 311 (Avg: 101) | Type: Tech

Nvidia CEO Jensen Huang sees long-term opportunities in China despite tech tensions with the US, noting the country is part of a $200 billion market forecast for data centers and CPU https://t.co/r8Z4nCebO2 https://t.co/CgWHRl4J4Z

See 2 related tweets

@ReutersBiz: Nvidia CEO Jensen Huang sees long-term opportunities in China despite tech tensions with the US, not...
@MarioNawfal: 🇨🇳 China just unveiled a new gaming GPU designed to compete directly with NVIDIA and AMD.

The timin...

12. edzitron (Group Score: 78.4 | Individual: 27.4)

Cluster: 3 tweets | Engagement: 300 (Avg: 1167) | Type: Tech

https://t.co/xtFzEngE1g\n\nQT @loffredojeremy: I just ran into a guy I know in NYC who said he’s doing a temporary gig for OpenAI. According to him, OpenAI is paying hundreds of people and families in New York to install 360-degree cameras throughout their homes that record virtually everything. Vacuuming, washing dishes, cooking, and everything else. He said his job is to go around collecting the memory cards and replacing them with new ones.

According to him, OpenAI told workers like him that the recordings are being used to help develop a smart home device. But he said the strange part is that all of the people supervising the project are behavioral psychologists.

See 2 related tweets

@ShriramKMurthi: Didn't @matt_levine write many months ago that OpenAI was hiring a bunch of people in finance on tem...
@sandygrains: RT @loffredojeremy: I just ran into a guy I know in NYC who said he’s doing a temporary gig for Open...

13. chddaniel (Group Score: 78.0 | Individual: 47.3)

Cluster: 2 tweets | Engagement: 263 (Avg: 21) | Type: Tech

BREAKING: ai slop is now 100% optional.

Introducing Shipper - the first AI business builder to replace a 9-5 salary.

It’s the foundation for a one-person billion dollar company, coordinating AI workers across coding, sales, marketing, operations, and design.

🚨 Giving away unlimited FREE credits for the next 24 hours, randomly.

(until they run out)

Want access?

• Retweet • Leave a 'Shipper' comment

See 1 related tweets

@chhddavid: so you're basically telling me Claude Opus 4.7 can now build web apps and mobile apps without....

-...

14. garrytan (Group Score: 74.9 | Individual: 42.8)

Cluster: 2 tweets | Engagement: 867 (Avg: 295) | Type: Tech

Bob McGrew has a framework I keep thinking about: in the AI future there are only two jobs. The Lone Genius and the Manager.

That's it. Everything else gets absorbed.

The Lone Genius is the person sitting alone at a computer, amplified 1000x by AI. One person with taste, vision, and relentless focus who can now do what used to take a team of 50.

The Manager is the person who becomes CEO of their own "firm" where most of the employees are AI agents. They define the goals. They decide what matters. They coordinate. The AI does the execution.

The Marxists will hear "two jobs" and panic. "What about everyone else?!" But here's what they're missing: AI doesn't shrink these two categories. It explodes them open. More people get to be geniuses. More people get to be managers. The barrier to entry for both just collapsed.

What actually gets eliminated? David Graeber called them "bullshit jobs." Graeber was no libertarian! He inspired Occupy Wall Street.

His words: "Huge swaths of people spend their entire working lives performing tasks they secretly believe don't really need to be performed. The moral and spiritual damage that comes from this situation is profound. It is a scar across our collective soul."

Graeber said bullshit jobs are "a form of spiritual violence directed at the essence of what it means to be a human being." They induce "hopelessness, depression, and self-loathing."

This is who the left should be fighting for. Not to preserve those jobs. To liberate people from them and give them better ones.

The dirty secret of the modern economy: millions of people sit in roles so pointless that even they can't justify their existence. Compliance layers. Reporting layers. Coordination layers. Meeting-about-the-meeting layers. They know it's meaningless. It eats them alive.

AI eats those layers. Good. That's a jailbreak.

What I love about Bob's framework is where it points. The Lone Genius used to require a PhD, a lab, institutional backing. Now a 19-year-old with taste and Codex can ship what took a research team a year. The genius bottleneck was never talent. It was access.

The Manager used to mean you needed to hire 50 people, raise money, build an org chart. Now you can orchestrate a fleet of AI agents from your laptop. The management bottleneck was never skill. It was capital.

AI doesn't concentrate genius and management into fewer hands. It distributes them into more hands. The working class kid in West Virginia. The single mom in Ohio. The 55-year-old who got laid off and now builds software for the first time. Those are some of Bob's future geniuses and managers.

The best founders I see at YC are already living this. They toggle between both modes in the same day. Morning: lone genius, creative insight, the thing nobody else sees. Afternoon: manager, spinning up agents, steering, shipping.

The cycle time between genius and manager IS the new productivity metric.

So when someone tells you AI means "only two jobs and everyone else starves," quote Graeber to them, they’ll get it.

Graeber knew the real violence was making people do meaningless work and pretending it was dignity. AI ends that. More genius. More agency. Fewer spiritual prisons.

See 1 related tweets

@hammer_mt: Actually the opposite will happen because we already automated away the need for these jobs and yet ...

15. seraleev (Group Score: 74.6 | Individual: 37.7)

Cluster: 2 tweets | Engagement: 79 (Avg: 40) | Type: Tech

Fast track to $10k/month

ICON SCREENSHOTS ONBOARDING PAYWALL STEP-BY-STEP UX https://t.co/mF9uwl0aI1\n\nQT @seraleev: Let’s run through the basics. Your mobile app has a sales funnel:

ICON. The tip of the iceberg. It drives your entire unit economics. A single icon change can cut your install cost 7x. Expensive installs? Start experimenting with the icon.
SCREENSHOTS. The second thing users look at. The first 3 are what matter. Put your best copywriting into those first 3 headlines.
Icon and screenshots are solid but still no money? Go deeper. ONBOARDING. You hooked the user, but they forgot why they came and dropped off mid-flow. Even worse – no onboarding at all.
MONEY. Paywall after onboarding. Hard paywall, soft paywall, product mix, red button or blue – do whatever you want. Just follow the guidelines.
STEP-BY-STEP UX. Hold the user’s hand. Help them solve their problem. Then you get another shot at the paywall.

That’s it. Start with 1 and 2. Once those are dialed in, move down the funnel. Study competitors in your niche. Ask yourself: why did they do it this way? That question alone helps you understand the logic and beat it.

Bookmark. Apply. Good luck.

See 1 related tweets

@seraleev: Let’s run through the basics. Your mobile app has a sales funnel:

ICON. The tip of the iceberg. ...

16. javinpaul (Group Score: 71.3 | Individual: 12.4)

Cluster: 6 tweets | Engagement: 1 (Avg: 2) | Type: Tech

I Joined 30+ Generative AI Courses on Udemy: Here Are My Top 7 Recommendations for 2026 https://t.co/SEGcDYoy4T #ArtificialIntelligence #courses

See 5 related tweets

@javinpaul: I Joined 30+ Generative AI Courses on Udemy: Here Are My Top 7 Recommendations for 2026 https://t.co...
@javinpaul: I Joined 30+ Python Courses on Udemy: Here Are My BEST 5 Recommendations for 2026 https://t.co/CDv22...
@javinpaul: I Tried 20+ Agentic AI Courses on Educative: Here Are My Top 5 Recommendations for 2026 https://t.co...
@javinpaul: I Tried 30+ Coding with AI Courses on Udemy: Here Are My Top 7 Recommendations for 2026 https://t.co...
@javinpaul: I Tried Generative AI for Software Development Certificate on Coursera and Its Awesome https://t.co/...

17. mitsuhiko (Group Score: 69.2 | Individual: 27.3)

Cluster: 4 tweets | Engagement: 39 (Avg: 101) | Type: Tech

In the composer setup (and go fwiw) it’s rather distributed. I wonder if GitHub will want to start doing something more dramatic to affected repositories.\n\nQT @SocketSecurity: 🚨 Supply chain attack on the Laravel Lang organization:

700+ historical versions across multiple community-maintained Laravel Lang packages were compromised with an RCE backdoor, including:

laravel-lang/lang laravel-lang/http-statuses laravel-lang/attributes Laravel-Lang/actions

The payload targets cloud creds, CI/CD secrets, Kubernetes tokens, Vault, browser data, password managers, SSH keys, and more.

See 3 related tweets

@levelsio: I'm gonna get annoying so don't block me but another reason I don't use frameworks or dependencies

...

@Kyrannio: Here we go again: Laravel Lang it’s your turn\n\nQT @SocketSecurity: 🚨 Supply chain attack on the La...
@matteocollina: RT @AhmadNassri: "700+" is a magic number today!

Earlier we've observed 700+ GitHub Repositories ...

18. Hamburgerai (Group Score: 69.0 | Individual: 26.9)

Cluster: 3 tweets | Engagement: 3 (Avg: 3) | Type: Tech

🚀 强烈推荐大家去看看这个免费神课 —— zero2claude！作者是 Wix 的 VP Itay Shmool（@IShmool），他在 Wix 负责 Payments、Premium 和 Domains 等核心业务，同时也是 AI Agent Swarms 的实践者。

他一个人用 Claude Code 就完整搭建并运营了这个课程平台，目前已经有 17,000+ 学生在学习，支持 7 种语言，每天有几百人活跃，完全靠口碑传播，没有花一分钱营销。课程从「零终端经验」开始，系统性地教：

•软件基础 •Claude Code 核心用法 •进阶技巧

没有 shortcuts，没有假设前提，真正帮你从零建立硬核能力。平台本身也是他用 Claude Code 一手打造的，Lighthouse 满分（性能96、Accessibility/最佳实践/SEO 全部100），日请求量高达 640万，错误率极低，证明 Claude Code 完全能支撑生产级、可扩展的产品。

最打动我的一点是他的理念：最好的推广 AI 工具的方式，不是简化工具，而是升级人。给鱼不如教渔，真正让人掌握底层能力。免费、无付费墙，完全是给社区的贡献。 🔗 课程地址：https://t.co/TfRT9wAeNi 这正是当前 AI 时代最需要的「授人以渔」型内容，无论是产品经理、想转行的、还是已经在用 AI 但想系统提升的，都非常值得去学。\n\nQT @IShmool: I built "zero2claude", a free course that takes people from zero terminal experience to shipping with Claude Code.

The curriculum goes from absolute zero → software basics → Claude Code fundamentals → advanced usage. No shortcuts, no assumptions.

17,000+ students. 7 languages. ~500 active hourly. No marketing. No ads. Pure word of mouth.

The entire platform? Built and operated by one person with Claude Code.

Lighthouse audit: ✅ Performance : 96 ✅ Accessibility : 100 ✅ Best Practices : 100 ✅ SEO : 100

Production stats: ~6.4M requests/day. 74 req/sec sustained. <0.003% error rate.

Claude Code doesn't just write code. It builds production-grade, scalable products.

The best way to grow Claude Code adoption isn't to simplify the tool. It's to level up the people. Give fishing rods, not fish.

Free. No paywall. My contribution to the community. Link in the comments 👇

See 2 related tweets

@Hamburgerai: 一旦你装了这个，你会发现 Claude Code 完全像变了个东西。

Anthropic 最近悄悄放出了官方插件 claude-code-setup，它基本上把 Claude Code 从“能用”直...

@Hamburgerai: 最近在玩 Claude Code 的朋友，一定要看 @dani_avila7 这篇关于 Subagents（子代理）的长文。

核心问题大家应该都遇到过：
随着会话越来越长，你让主代理去 grep...

19. MTSlive (Group Score: 68.8 | Individual: 35.4)

Cluster: 2 tweets | Engagement: 127 (Avg: 120) | Type: Tech

SITUATION DETECTED: More than half of American businesses now have a paid AI transaction in a given month, up from 6.6% in January 2023. https://t.co/yHaXcNt0H5\n\nQT @MTSlive: We built a live job board of all the frontier labs hiring right now.

Plus a way to see news, funding rounds, compare model releases and more.

Monitor the frontier at: https://t.co/PxDc5ImbPx https://t.co/QM93PufWB0

See 1 related tweets

@MTSlive: AI adoption by industry, over time https://t.co/g5hDnzs56I\n\nQT @MTSlive: We built a live job board...

20. Shipper_now (Group Score: 67.0 | Individual: 37.0)

Cluster: 3 tweets | Engagement: 386 (Avg: 32) | Type: Tech

RT @shipper_now: BREAKING: Today, we closed the chapter on vibe coding.

I just witnessed my Mac erase a $31B company in 186 seconds.

This is beyond insane. https://t.co/Jtokd39QLS

See 2 related tweets

@chhddavid: many startups just died today…\n\nQT @shipper_now: BREAKING: Today, we closed the chapter on vibe co...
@chddaniel: this thing is actually terrifying...\n\nQT @shipper_now: BREAKING: Today, we closed the chapter on v...