科技推特精选 - 2026-05-09

2026年5月9日科技每日简报

Today's top tech conversations are led by @cryptopunk7213, whose post about 'deepseek is raising a monster ...' garnered the highest engagement. Key themes trending across the top stories include anthropic, microsoft, google, deepseek, round. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. cryptopunk7213 (Group Score: 216.3 | Individual: 49.7)

Cluster: 8 tweets | Engagement: 1823 (Avg: 327) | Type: Tech

deepseek is raising a monster $7 billion round at$ 50B val making it china's largest ever AI raise but what shocks me the most is the founder, liang wenfeng:

he's personally contributing 40% of the round himself. $3 billion.

he owns 90 PERCENT of the company (unheard of at this valuation)

deepseek was founded inside his hedge fund, one of China's most successful funds.

guys a fucking beast. this round is meant to achieve 2 things:

acquire as much compute to push out new deepseek models more often
turn deepseek revenue-positive by pushing new enterprise products (same tactic as OAI and anthropic)

deepseek v4.1 is expected to release soon.

See 7 related tweets

@AiBattle_: DeepSeek plans to launch V4.1 in June, according to the information\n\nQT @jukan05: DeepSeek Reporte...
@zephyr_z9: "One of the people said the company plans to launch V4.1, an updated version of its V4 model, in Jun...
@jukan05: The Information: DeepSeek is expected to release V4.1 this June.

The Information: DeepSeek is worki...

@wallstengine: DEEPSEEK IS NOW RAISING THE BIGGEST AI ROUND IN CHINA’S HISTORY

The Information reports DeepSeek is...

@teortaxesTex: If Wenfeng continues to be completely obscure after this and having his net worth surge to like $30 ...

2. gabriel1 (Group Score: 164.0 | Individual: 43.2)

Cluster: 5 tweets | Engagement: 1160 (Avg: 239) | Type: Tech

if 5.5 becomes 20x faster, you'll talk and code live while the interface is changing as you speak\n\nQT @OpenAI: Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents.

Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold.

Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

See 4 related tweets

@MLStreetTalk: RT @OpenAI: Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GP...
@minchoi: OpenAI just dropped GPT-Realtime-2, Realtime-Translate and Realtime-Whisper.

Reason. Translate. Tra...

@gabriel1: ultra fast voice & language models will be even more important for everyday work

use your voice...

@haydenbleasel: Working on a translate playground so you can try this out!

💬 → 💬 https://t.co/0pbRMuOyLg https://t...

3. zephyr_z9 (Group Score: 131.6 | Individual: 36.2)

Cluster: 4 tweets | Engagement: 248 (Avg: 223) | Type: Tech

This could be super big\n\nQT @tilderesearch: Introducing Aurora, a new optimizer for training frontier-scale models.

We train Aurora-1.1B, which achieves 100x data efficiency on open-source internet data. Despite having 25% fewer parameters, 2 orders of magnitude fewer training tokens, and using fully open-source internet-only data, Aurora matches Qwen3-1.7B on several benchmarks.

Aurora was developed after identifying a major failure mode that can occur under Muon, an increasingly popular optimizer that has shown strong gains over Adam(W). We find that Muon can cause a huge percentage of neurons to effectively die early in training, reducing effective network capacity so that many parameters no longer meaningfully contribute to network outputs.

By redistributing update energy more uniformly across neurons while preserving Muon’s stability properties, Aurora prevents neuron death and recovers substantial model capacity.

What makes this work especially exciting is that it points toward a broader direction for ML research: better optimizers may not come purely from elegant mathematical abstractions, but from understanding and addressing the concrete dynamics and pathologies that emerge inside real training systems.

See 3 related tweets

@teortaxesTex: I wanted to say "I don't believe in this free lunch bs" but their hypothesis is interesting. In effe...
@vkhosla: More proof from one of our companies innovation continues unabated around LLM's...\n\nQT @tilderesea...
@scaling01: RT @tilderesearch: Introducing Aurora, a new optimizer for training frontier-scale models.

We train...

4. negligible_cap (Group Score: 114.3 | Individual: 22.6)

Cluster: 7 tweets | Engagement: 147 (Avg: 56) | Type: Tech

*ANTHROPIC INKS $1.8 BILLION COMPUTING DEAL WITH AKAMAI

Anthropic and $AKAM just signed a$ 1.8B computing deal to meet demand. AKAM said yesterday in their earnings call that they it struck a 7 year cloud computing deal with a “leading frontier model provider”, but Anthropic wasn’t named until now. Anthropic's been on a tear lately with these deals

See 6 related tweets

@wallstengine: Anthropic is behind Akamai’s new $1.8B, 7-year cloud computing deal, after$ AKAM only disclosed a “l...
@business: Anthropic has signed a $1.8 billion computing deal with cloud services provider Akamai to meet surgi...
@Techmeme: Akamai says it struck a seven-year cloud computing deal with a "leading frontier model provider"; so...
@edzitron: https://t.co/c5Ggjgj1Ek

Is ANY data center capacity coming online? Why is Anthropic so desperate th...

@rachelmetz: Spoiler alert from me\n\nQT @Techmeme: Akamai says it struck a seven-year cloud computing deal with ...

5. wallstengine (Group Score: 113.3 | Individual: 39.6)

Cluster: 4 tweets | Engagement: 235 (Avg: 102) | Type: Tech

GOOGLE’S AI DRUG DISCOVERY SPINOUT IS RAISING OVER $2B

Bloomberg reports Isomorphic Labs is in advanced talks to raise more than $2B in new funding.

The company spun out of Google DeepMind in 2021 to commercialize AI drug discovery built around AlphaFold.

Thrive Capital is set to lead the round, with Alphabet also expected to participate.

Isomorphic raised $600M in its first external round last year, also led by Thrive.

The new capital is expected to go toward its drug design engine and global expansion.

Isomorphic already has partnerships with Johnson & Johnson, Eli Lilly, and Novartis.

Alphabet is turning DeepMind’s science lab work into a standalone AI drug company.

See 3 related tweets

@business: Isomorphic Labs, an AI-powered drug discovery company spun out of Google DeepMind, is in advanced di...
@Techmeme: Sources: Isomorphic Labs, an AI-powered drug discovery company spun out of Google DeepMind, is in ad...
@AndrewCurran_: Sir Demis's biolab is raising $2 billion. https://t.co/jDGGpJy0Il\n\nQT @RebeccaTorrenc5: SCOOP: Iso...

6. chddaniel (Group Score: 110.5 | Individual: 28.5)

Cluster: 4 tweets | Engagement: 4 (Avg: 8) | Type: Tech

so you're telling me Claude Code w/ Opus 4.7 can now...

build a full game
launch on Steam
do email marketing
self-build new features

without any human in the loop??

it's so over... https://t.co/HGkUNopg2R\n\nQT @shipper_now: Introducing Shipper Max

AI that one-shots any video game: Minecraft, Flight Simulator, NFS, Counter-Strike, Subway Surfers, 2048, Chess, and a lot more.

Build games in 1 click. Publish to Steam in 2 clicks.

See 3 related tweets

@chhddavid: pov: Claude one shotted a game i planned to make over several months https://t.co/XIg7Zkkmom\n\nQT @...
@chhddavid: this is terrifying.\n\nQT @chddaniel: Introducing the world's first-ever AI Game Dev Studio.

Enter ...

@chhddavid: this is utterly terrifying.\n\nQT @shipper_now: Introducing Shipper Max

AI that one-shots any video...

7. ibuildthecloud (Group Score: 108.5 | Individual: 43.5)

Cluster: 4 tweets | Engagement: 1040 (Avg: 29) | Type: Tech

RT @artman: Today is a hard day. I shared this note with the @linear team today: We’ve made the difficult decision to increase our workforce. This is not a cost-cutting exercise or a reflection of anyone’s performance. We’re simply reimagining every role for the agentic AI era. We’re hiring. We’re sorry about that.

See 3 related tweets

@etnshow: BREAKING: Linear announce mass lay-ons https://t.co/K4pbysYwnv\n\nQT @artman: Today is a hard day. I...
@itsumeshk: today is a hard day. i shared this note with the @runable_hq team today:

we've made the difficult d...

@GergelyOrosz: It’s hard to be mad at @artman or @linear for this :)

Also, Linear has a very unique hiring proce...

8. negligible_cap (Group Score: 107.9 | Individual: 27.0)

Cluster: 6 tweets | Engagement: 199 (Avg: 56) | Type: Tech

Chris Hohn’s TCI Fund Management, which has held $MSFT for the last decade, dumped almost all of their$ 8B stake in Microsoft based on AI concerns.

“We reduced out investment in Microsoft because the rapid progress in AI introduces uncertainty over Microsoft’s competitive position in the future.

We are primarily concerned about Microsoft’s Office productivity software franchise, where AI could change established workflows and lead to the emergence of new productivity platforms, but we also see some risks in Azure.

TCI took the stake from 10% of their portfolio down to just 1% https://t.co/ISbOz2NTPA

See 5 related tweets

@firstadopter: FT: "Christopher Hohn’s hedge fund TCI has dumped almost all of its $8bn stake in US software giant ...
@FT: FT Exclusive: Sir Christopher Hohn’s hedge fund TCI has slashed its stake in Microsoft from 10% to j...
@Techmeme: Investor letter: TCI, one of the world's biggest hedge funds, cut almost all of its $8B stake in Mic...
@financialjuice: Hohn’s hedge fund TCI unloads nearly entire $8 billion Microsoft holding - FT$ MSFT

TCI worried AI...

@FT: Chris Hohn’s hedge fund slashes $8bn Microsoft stake in warning over AI disruption https://t.co/tUcW...

9. swyx (Group Score: 99.6 | Individual: 42.7)

Cluster: 3 tweets | Engagement: 548 (Avg: 79) | Type: Tech

RT @trq212: HTML is the new markdown.

I've stopped writing markdown files for almost everything and switched to using Claude Code to generate HTML for me. This is why.

See 2 related tweets

@danshipper: SO INTERESTING\n\nQT @trq212: HTML is the new markdown.

I've stopped writing markdown files for al...

@clairevo: Messaging @trq212 after he taught me this https://t.co/pyG9geBazL\n\nQT @trq212: HTML is the new mar...

10. HarryStebbings (Group Score: 98.9 | Individual: 33.9)

Cluster: 3 tweets | Engagement: 59 (Avg: 114) | Type: Tech

"No one has this expertise in-house. It's the worst gap between in-house and external expertise in our lifetimes. This is gonna benefit Palantir.

If we were head-scratching, why are Anthropic and OpenAI setting up these consulting entities, which seem goofy, right? They're not, because the dollars are gonna go up, the initiative is gonna go up, but the inability to have anyone in-house that can remotely execute is rare.

The majority of HubSpot and Shopify agencies are gone. They have no AI play. But the ones that do are gonna have infinite demand because whether it's enterprise or SMB, no one has this expertise.

Even Coinbase today, laying off 15%. Managers managing managers. We don't need them.

I need folks that do the work in AI. If Brian doesn't have the people, what hope is there for the rest of the world?" @jasonlk

Love to hear your thoughts on this @bhalligan @ssankar @pmarca @mmurph\n\nQT @HarryStebbings: This is the F**king superbowl of earnings.

Google Won. Facebook Lost.

I sat down with @jasonlk and @rodriscoll to discuss the deal, along with the biggest news in tech this week:

Anthropic's $50BN Raise: What Does This Mean for a Potential IPO?
Atlassian, Twilio and Five9 Beat: The SaaS Apocalypse Over?
Sierra Raises at $15BN: Is it Peak or Potential?

My notes below:

Why Gemini Is Behind but Why It Might Not Even Matter Google is trailing in “coder love” as developers flock to Anthropic and OpenAI. While Gemini’s token growth lags behind competitors’ 10x gains, Google remains a powerhouse because it owns its model and distribution.
Why Palantir Is a Mega-Buy Right Now CEOs want $10M+ enterprise-wide transformations, not small point solutions. Palantir wins by delivering measurable, large-scale results. Their buying cycles are compressing as every stakeholder now attends the first meeting.
Why Is Meta Being Chastised for Spending on Capex? Meta’s $150B CapEx lacks the clear revenue attribution found in Google’s cloud backlog. Investors view it as a defensive hedge, a bet to be present if social shifts to chatbots. Critics argue a 10 to 15% ad lift does not justify such massive spend.
Who Wins and Who Loses From the Application Explosion? We are in the greatest application-building boom in history. Amazon and Microsoft win by providing the underlying compute for these new B2B apps. Meta loses because it gains nothing from this software sprawl.
Atlassian Versus Twilio: Who Is the AI Hero Versus the Unsustainable Bounce? Atlassian is successfully monetizing its base with Rovo AI, though new customer growth is slowing. Twilio has re-accelerated by becoming the essential API infrastructure for new AI agents. Winners must eventually drive both expansion and new customer acquisition.
Should Microsoft Really Be Trading Like the Rest of SAAS Companies? Excluding AI, Microsoft’s core business is flat to slightly down. Without the AI narrative, it would likely trade like a mature SaaS company at 3x revenue. Their growth is now entirely dependent on massive, high-risk CapEx bets.
Why Your VP Needs to Be Full Stack and Do Everything End-To-End AI is eliminating “managers of managers.” Founders now demand leaders who can actually ship. A modern VP should use agents to execute tasks, like campaigns, end-to-end in minutes. If an executive cannot lead from the front with these tools, they are a liability.

(links below)

See 2 related tweets

@HarryStebbings: Why Gemini is behind but why it might not even matter

"Gemini is behind in traction, especially in ...

@HarryStebbings: Why is Meta being chastised for spending on CapEx?

"Meta is being challenged because the ROI is unc...

11. mervenoyann (Group Score: 98.1 | Individual: 41.4)

Cluster: 3 tweets | Engagement: 2756 (Avg: 341) | Type: Tech

RT @AnthropicAI: New Anthropic research: Natural Language Autoencoders.

Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.

Here, we train Claude to translate its activations into human-readable text. https://t.co/pMLsxM2VAO

See 2 related tweets

@WesRoth: Anthropic has developed "Natural Language Autoencoders" (NLAs), a new technique that translates an A...
@Simeon_Cps: Finally, we found a hill-climbing problem which may spur strong progress in interp.\n\nQT @Anthropic...

12. aakashgupta (Group Score: 97.8 | Individual: 32.8)

Cluster: 3 tweets | Engagement: 17 (Avg: 70) | Type: Tech

Most PMs are still copy-pasting into a chat window.

This episode shows a PM building a 21-agent dev team inside Claude Code and shipping a working iOS hockey rules app to TestFlight, live, in a single recording.

The whole pipeline runs as one system.

A system analyst agent reads the spec and breaks it into requirements. Confluence stores the docs. Jira gets populated with frontend tickets, each one linked to a specific Figma screen. Figma Make generates UI from brand guidelines. Claude Code builds the actual screens. Simulator validates. TestFlight ships.

A PM doesn't drive any of this with a single prompt. They drive it with scaffolding.

That's the part most people miss when they hear "AI agents." The 21 agents are 21 specialized roles with defined inputs, outputs, and handoffs, sitting on top of the tools PMs already use. The MCP connections to Confluence and Jira are load-bearing. Without them the whole thing collapses into spaghetti code by step three.

The contrarian point in the episode lands harder than the demo. Vibe coding produces flashy prototypes that nobody can ship. Classic PM skills produce TestFlight builds. Writing crisp specs. Defining acceptance criteria. Breaking work into tickets. The agents amplify those skills rather than replace them.

The dictation segment is where this clicks. The longest spec on this podcast wasn't typed. It was dictated, because dictation forces you to verbalize edge cases the way you would in a stakeholder meeting. Vague spec, vague output. Detailed spec, working app.

The closing line of the episode is the one PMs should sit with. The gap between a PM running this kind of pipeline and a PM still copy-pasting into a chat window will be enormous in two years. Not because the pipeline PM is smarter. Because the pipeline PM has compound reps on a fundamentally different workflow.

The hidden insight from this podcast is that the unlock is treating your AI tools the way an engineering org treats its services. Clear contracts. Clear ownership. Clear handoffs.

That's a PM skill, not a coding one.\n\nQT @aakashgupta: This guy literally broke down everything to build a 21 agent dev team in Claude Code:

0:00 - Why one prompt = AI slop 1:55 - Claude Code as your full startup 2:43 - The 21-agent team 4:57 - Inside the system analyst agent 5:52 - Live demo: 0 → TestFlight 8:42 - Define a "good" system analyst first 11:53 - The system analyst workflow 12:17 - Why Confluence + Jira + MCPs matter 15:30 - Classic PM skills > vibe coding 19:15 - The scaffolding that kills spaghetti code 22:10 - Setting up agents inside Claude Code 26:19 - The longest dictation prompt this podcast has seen 35:29 - Why dictation beats typing for AI specs 47:30 - Brand guidelines in Figma Make 55:59 - Idea → prompt → design → app 1:06:27 - Claude Code builds the Figma screens 1:23:59 - Frontend tickets appear in Jira (with Figma links) 1:48:49 - The hockey rules AI app goes live 1:53:56 - Full recap: Claude, Confluence, Figma, Jira, Simulator, TestFlight 2:03:17 - Should PMs get AI PM certificates? 2:08:15 - How to build a PM portfolio that lands FAANG offers 2:13:25 - How to get started this week

See 2 related tweets

@aakashgupta: Talking to your phone for 5 minutes captures 5x more spec detail than typing for 30. That number upe...
@aakashgupta: April 2020. A 30-year-old Hungarian software guy is stuck in London with his savings draining. UK bo...

13. wallstengine (Group Score: 97.1 | Individual: 34.6)

Cluster: 4 tweets | Engagement: 242 (Avg: 102) | Type: Tech

APPLE AND INTEL REACH PRELIMINARY CHIP-MAKING AGREEMENT

WSJ reports $AAPL and$ INTC have reached a preliminary agreement for Intel to manufacture some chips for Apple devices.

It is still unclear which Apple products Intel would make chips for.

The Trump administration reportedly pushed for the deal, with Commerce Secretary Howard Lutnick meeting Apple, Nvidia, Musk, and others to bring them into Intel’s ecosystem.

Intel now has partnerships with Apple, Nvidia, and Musk-linked companies.

Washington is trying to turn Intel Foundry into a U.S. chipmaking platform for America’s biggest tech companies.

See 3 related tweets

@cryptopunk7213: my god Intel’s on a generational run right now… just signed a landmark deal with Apple to supply chi...
@markgurman: I would pay careful attention to the language in this story, which does not say there is a signed ag...
@inafried: RT @markgurman: I would pay careful attention to the language in this story, which does not say ther...

14. sairahul1 (Group Score: 92.9 | Individual: 20.1)

Cluster: 5 tweets | Engagement: 40 (Avg: 30) | Type: Tech

Sam Altman predicted a 1-person billion-dollar company.

Anthropic's engineers just proved him right.

CEO: 1 human. Employees: AI agents. Operations: fully automatic.

30 minutes. Free. No excuses.

The zero-headcount company is no longer a joke. https://t.co/afpDX4eu2A

See 4 related tweets

@sairahul1: me when my agents are running the entire company https://t.co/eKvkJZBKBX\n\nQT @sairahul1: Sam Altma...
@sairahul1: Claude users right now: https://t.co/5QJvPYJwZg\n\nQT @sairahul1: Sam Altman predicted a 1-person bi...
@sairahul1: me after watching this video: https://t.co/muiAkakh1g\n\nQT @sairahul1: Sam Altman predicted a 1-per...
@sairahul1: This just killed 90% of junior dev jobs\n\nQT @sairahul1: Sam Altman predicted a 1-person billion-do...

15. teortaxesTex (Group Score: 92.4 | Individual: 32.8)

Cluster: 3 tweets | Engagement: 76 (Avg: 100) | Type: Tech

I hope @fchollet has something to say about this\n\nQT @patience_cave: The new Codex Goals feature is able to pursue tasks indefinitely, so how useful is it?

I ran it on the public ARC-AGI-3 games. After 160 hours and 30k actions it scored 61%

Codex gets the most work done within 4 hours. Afterwards, it begins to stagnate, and wait times increase

I am crestfallen it scored so well. I used a two-word baby prompt to “reverse engineer” the games during play. It had no prior knowledge of how the games worked

Due to the reverse engineering process, it scores well on its first play through. But once it beats a game, it can score perfectly on its next play through

On a few occasions I caught Codex trying to search for solutions on my computer and online. It follows your prompt closely, but if you’re not careful it will find loopholes when it’s frustrated

For example, when trying to solve Erdos problems, if it becomes faintly aware the problem is from Erdos, it does not hesitate to give up and say “the problem is listed online as Open, so it cannot / should not be solved”

Overall Codex Goals is fascinating, I can appreciate that it works for an unlimited amount of time. People shall value the virtue of patience once again 💺

It makes me wonder how well Codex Goals can do on the private set of ARC-AGI-3. I believe it’s possible to create benchmarks that can mog even the most devilish harness. In the coming weeks, Maze Bench will knock those scores down to 0% where they should be, ne’er to rise again

Arcprize Scorecard: https://t.co/Tncf8gwFhA

See 2 related tweets

@chatgpt21: I personally verified the results with Mr.Chair last night , he is in-fact not trolling.\n\nQT @pati...
@xeophon: RT @patience_cave: The new Codex Goals feature is able to pursue tasks indefinitely, so how useful i...

16. Techmeme (Group Score: 92.2 | Individual: 23.6)

Cluster: 5 tweets | Engagement: 67 (Avg: 6) | Type: Tech

Sources: Anthropic weighs raising as much as $50B at a ~$ 900B pre-money valuation, likely closing within two months; its annualized revenue is nearing $45B (Financial Times)

(Visit Techmeme dot com for the link and full context!)

See 4 related tweets

@wallstengine: Anthropic is weighing a raise of up to $50B at ~$ 900B pre-money, pushing valuation near $1T. Annuali...
@Cointelegraph: 🚨 JUST IN: Anthropic is reportedly weighing a funding round of up to $50B that could push its valuat...
@FT: Anthropic weighs deal for near $1tn valuation as revenue surges https://t.co/7lorIAvzs3...
@FirstSquawk: NEW ROUND IS EXPECTED TO VALUE ANTHROPIC AT ABOUT $900BN PRE-MONEY AND TO RAISE AS MUCH AS$ 50BN- FT...

17. Parul_Gautam7 (Group Score: 91.6 | Individual: 31.6)

Cluster: 3 tweets | Engagement: 140 (Avg: 91) | Type: Tech

Not every prompt needs a frontier model.

Routing by complexity is where real efficiency shows up.\n\nQT @OrcaRouter: 1/ OrcaRouter goes live today!

The other LLM routers tax you ~5% on every token. OrcaRouter doesn't take a cut.

Here's how that works, and why it matters for what you're building ↓ https://t.co/Wftvb3ZnHF

See 2 related tweets

@TheoBuildsAI: The “use the right model for the right request” approach just makes practical sense.

Not every prom...

@heyshrutimishra: Most LLM routers are static rules; OrcaRouter is a router that learns.

It embeds every prompt, scor...

18. garrytan (Group Score: 82.5 | Individual: 21.6)

Cluster: 5 tweets | Engagement: 8 (Avg: 391) | Type: Tech

RT @Hamburgerai: 很多 AI 编程翻车，不是因为代码第一版写不出来，而是需求没想清、设计没人挑、review 没做深、QA 没真跑、发布前也没人收口。一个人做项目时，这些环节最容易被跳过。

garrytan/gstack 把这件事做成了一套 Claude Code 工作流。README 里说它是一个 open source software factory，用 slash commands 把产品、工程、设计、QA、安全、发布这些角色压进一条流程。

它的核心就一句话：

让 Claude Code 不只是写代码，而是按 Think -> Plan -> Build -> Review -> Test -> Ship -> Reflect 推进交付。

· Sprint 主线清楚：README 明确写了 Think -> Plan -> Build -> Review -> Test -> Ship -> Reflect，每一步都有对应技能接住。

· 命令面向真实工作：/office-hours 梳理想法，/plan-ceo-review 看产品判断，/review 抓生产风险，/qa 打开真实浏览器测试，/ship 做发布收口。

· 安装路径明确：推荐把仓库 clone 到 ~/.claude/skills/gstack 后运行 ./setup，团队共享还可以走 ./setup --team。

· 不只是角色扮演：README 里强调每个 skill 会把产物传给下一步，例如 design doc、test plan、review 结果和发布检查可以串起来。

它适合已经认真用 Claude Code 做产品、做长期项目、做交付的人；不适合只想让 AI 快速改一个小脚本的人。gstack 的价值不是让 AI 多生成几行代码，而是把最容易漏掉的工程环节固定下来。

Repo: <https://t.co/nn3q41n6NR>

#ClaudeCode #AI编程 #开发工作流 #GitHub开源

See 4 related tweets

@GitHub_Daily: 有时候会打开多个 Claude Code、Codex 会话窗口，同时执行不同开发任务，无法及时查看运行状态。

最近找到 Open Island 这个开源工具，专门用来监控和管理多个 Agent 的运...

@RocM301: 今天telegram更新了很多功能，很多和bot有关。其实telegram有inline官方机器人可以直接用的：

Inline 官方机器人这些不需要添加进群，在任何群组或聊天中输入 @bot用户名...

@bearliu: 大家都在问 vibe coding 能不能用来做产品、能不能真正交付——但我觉得这个问题问错了。

我最近出了一个新视频，讲的是一个更值得问的问题：vibe coding 能不能帮你更快验证想法？

...

@xicilion: 我试过用 ai 生成 floor plan，惨不忍睹。ai 根本不知道门怎么放，沙发怎么放。 cad 行业一直没有出现语义定义，也是因为设计领域本身的特性决定的。勉强出一个 scad，也是工程师思维...

19. wallstengine (Group Score: 82.3 | Individual: 28.3)

Cluster: 4 tweets | Engagement: 197 (Avg: 102) | Type: Tech

BIG TECH IS TRYING TO FUND SK HYNIX CHIP PRODUCTION

Reuters reports SK Hynix has received unprecedented offers from global tech firms trying to secure memory chip supply.

The offers include investing in dedicated memory production lines and financing ASML EUV machine purchases.

One source said available capacity is “essentially zero right now.”

“There isn’t even a small portion that can be designated for a specific customer.”

SK Hynix is cautious because customer funding could tie it to specific buyers and force lower chip pricing in exchange for long-term revenue guarantees.

Another structure being discussed across long-term chip agreements: customers provide 30% to 40% of cash upfront.

See 3 related tweets

@StockMKTNewz: South Korean 🇰🇷 Memory company SK Hynix has reportedly been flooded with unprecedented offers from b...
@StockSavvyShay: Big Tech is reportedly offering to fund dedicated SK Hynix memory production lines and $ASML EUV pur...
@Techmeme: Sources: SK Hynix's big customers have been proposing a range of offers to secure memory chip suppli...

20. WesRoth (Group Score: 82.2 | Individual: 26.1)

Cluster: 4 tweets | Engagement: 69 (Avg: 26) | Type: Tech

Anthropic CEO Dario Amodei announced at a recent developer conference that the AI lab experienced a staggering 80x year-over-year growth in both revenue and usage during Q1 2026.

This hyper-growth which reportedly pushed their annualized revenue run rate well past the $30 billion mark has been largely driven by the explosive enterprise adoption of tools like Claude Code and autonomous agent workflows.

Anthropic originally planned for a 10x growth rate. The 80x surge created severe compute shortages, forcing the company into infrastructure deals, including securing capacity from SpaceXAI's Colossus 1 supercomputer just to keep up with the demand.

See 3 related tweets

@WesRoth: Anthropic CEO Dario Amodei just admitted on stage that their Q1 2026 growth hit an annualized 80x fa...
@StockSavvyShay: Anthropic CEO Dario Amodei said revenue and usage grew 80x that was far beyond the 10x upside case t...
@minchoi: RT @minchoi: The most interesting dynamic in AI right now:

Anthropic just rented the entire compute...