科技推文精选 - 2026年5月31日

2026年5月31日科技每日简报

Today's top tech conversations are led by @rohanpaul_ai, whose post about 'This survey suggests over 80% ...' garnered the highest engagement. Key themes trending across the top stories include price, added, agents, https, windows. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. rohanpaul_ai (Group Score: 149.2 | Individual: 35.4)

Cluster: 6 tweets | Engagement: 132 (Avg: 56) | Type: Tech

This survey suggests over 80% of companies have seen no productivity gains from AI so far, despite billions in spending.

Among 6,000 executives, 1/3 of leaders said they use AI, but only for 90 minutes a week.

This is even though most respondents believe AI will increase productivity by 1.4%, cut staff by 0.7%, and boost output by 0.8% in the next 3 years.

Of the executives, a third said they use AI at work, but only around 1.5 hours per week on average. Meanwhile, 25% of those surveyed have not used AI yet.

nber .org/papers/w34836\n\nQT @rohanpaul_ai: Goldman Sachs: "Token use by AI agents is expected to multiply 24 times by 2030"

AI agents are now creating the first serious cost test for the AI boom. As was reported this week, Uber and Microsoft are already rethinking expensive agent usage.

A chatbot may answer once, but an agent plans, calls tools, checks results, edits mistakes, and repeats the loop.

That loop can make one user request consume 10x, 50x, or even far more tokens than a normal answer.

Goldman’s bullish case is that monthly token use could reach 120 quadrillion by 2030, while inference cost per token keeps falling 60%-70% per year.

The fight is now between agent productivity and token waste.

Earlier this month, Microsoft began revoking developer access to Claude Code, with plans to move them to its in-house Copilot Command Line Interface tool by June 30. The company has framed this as consolidating teams around its own tools, but the timing at the fiscal year’s end hints it may also be about lowering costs.

See 5 related tweets

@rohanpaul_ai: RT @rohanpaul_ai: Goldman Sachs: "Token use by AI agents is expected to multiply 24 times by 2030"

...

@chandrarsrikant: We spoke to global and Indian firms on surging AI costs. Perspectives from OpenAI, Zerodha, Together...
@MilkRoadAI: RT @MilkRoadAI: A company accidentally spent $500 million in a single month on Claude after failing ...
@sairahul1: RT @sairahul1: 19-year-old makes $50,000 a month selling AI agents to business owners at$ 10,000-15,...
@moneycontrolcom: #AIWithMC | The end of Tokenmaxxing?

Companies take a hard look at productivity gains as AI costs s...

2. AnandButani (Group Score: 141.4 | Individual: 49.1)

Cluster: 6 tweets | Engagement: 17210 (Avg: 2176) | Type: Tech

RT @claudeai: Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.

Available today at the same price. https://t.co/EufxL7T1kb

See 5 related tweets

@cgtwts: pov: you accidentally say “hello” to claude opus 4.8 and it eats 3% of your session limit. https://t...
@zoink: Opus 4.8 is a very strange model. Clearly Anthropic tried to improve honesty, which is commendable. ...
@petergyang: It's interesting how the same model (Opus 4.8) can produce very different output whether you're usin...
@chddaniel: Opus 4.7 vs. Opus 4.8 for website design:

4.7 looks like AI slop. 4.8 looks like it was designed by...

@no_stp_on_snek: Ok Opus 4.8 has now been out for a few days, enough time for actual reviews and feedback (not day of...

3. StockSavvyShay (Group Score: 103.6 | Individual: 41.7)

Cluster: 3 tweets | Engagement: 2134 (Avg: 744) | Type: Tech

$NVDA and$ MSFT are reportedly set to unveil the first $ARM based Windows PCs powered by Nvidia's N1X chip at Jensen Huang's GTC Taipei keynote on June 1.

The move pushes Nvidia into the PC processor market as Microsoft takes another shot at local AI agents on Windows. https://t.co/cxUvNUiDQr\n\nQT @StockSavvyShay: $NVDA is on pace to generate over$ 1 trillion in profit over the next three years:

• 2026: $258B (+98% YoY) • 2027:$ 361B (+40% YoY) • 2028: $438B (+21% YoY)

Nvidia is the only company in world that sells the silicon, networking, software & platform tools that capture dollar content from every layer of the AI economy.

See 2 related tweets

@StockSavvyShay: $NVDA is on pace to generate over$ 1 trillion in profit over the next three years:

• 2026: $258B (...

@StockSavvyShay: These AI suppliers are on pace to generate over $500B in operating profit this year:

• $NVDA ~$ 258B...

4. VaibhavSisinty (Group Score: 95.5 | Individual: 32.6)

Cluster: 4 tweets | Engagement: 62 (Avg: 77) | Type: Tech

Someone just built the AI tutor every parent has been waiting for. 🤯

It's called Koji. Built by learning experts from MIT and Harvard.

And it doesn't do the homework for your kid. It teaches them how to actually think.

What makes it different:

→ It sees what's on your screen in real-time

→ It points, sketches, and writes like a real tutor sitting next to you

→ It doesn't give the answer. It walks you through finding it yourself

Every other AI kids use does the thinking for them. Koji makes them do the thinking.

That's the whole game.

They're tutoring the next 1,000 learners for free this summer. In math and coding.

If you have a kid, this is the one to try.\n\nQT @suekhim: AI is making kids dumber.

It should be making them geniuses.

Introducing Koji, the first AI tutor that gets kids to actually think. 👇 https://t.co/tmCdVoEZYF

See 3 related tweets

@kevinweil: I'm such a huge fan of what @suekhim and the Brilliant team have built.

When every kid in every co...

@amorriscode: RT @suekhim: AI is making kids dumber.

It should be making them geniuses.

Introducing Koji, the fi...

@rohanpaul_ai: RT @rohanpaul_ai: Jensen Huang talked about how a real-time AI tutor is making personalized, one-on-...

5. godofprompt (Group Score: 94.7 | Individual: 34.5)

Cluster: 3 tweets | Engagement: 58 (Avg: 84) | Type: Tech

A finance professor summed up every AI valuation in one line: nobody wants to use the eighth best product.

That's the whole bet behind Anthropic's $1 trillion. In most industries you can be fourth and still run a good business.

In AI, the theory goes, you're one of a tiny handful that print money, or you're a zero, because nobody pays for the model that comes in eighth.

So read the number again. $1 trillion isn't the price of Anthropic doing well. It's the price of Anthropic being one of the last ones standing.

Which means the valuation has a body count baked in. Every dollar of it assumes most of the other labs don't make it. The ones raising at fifty, a hundred, two hundred billion right now are, in this math, already corpses that haven't been told yet.

That's why the money crowds into two or three names instead of twenty. Nobody's paying $1 trillion to bet that AI works. They're paying it to bet on which company is still standing when it does.

The logic isn't crazy. Software has winner-take-most dynamics. Best model pulls the most users, most users generate the most data and revenue, that funds the most compute, which makes the best model. The leader compounds. Everyone else fights over scraps.

But it puts strange pressure on the price. A $1 trillion mark says two things at once. Wildly optimistic about Anthropic. Quietly certain that almost everyone else fails. You can't believe the number without believing in a wave of failures that hasn't happened yet.

And the survivor list isn't settled. Anthropic's own valuation climbed faster than any company in venture history. Three years ago it booked its first dollar of revenue. The seat it's being priced into hasn't existed long enough for anyone to know it's safe.

Here's the uncomfortable part for anyone holding this stuff. If the market is pricing two winners and there turn out to be three, the math breaks, because the third one was never supposed to get paid.

If there's only one, then two of these valuations are fiction. The price only works if the guess about who dies is exactly right.

And that guess is being made today, in private rounds, by people who need it to be right because they already wrote the check.

Maybe they nailed it. A couple of these companies could end up among the most profitable on earth. That's the bull case, and it's real.

Just be honest about what $1 trillion actually buys. Not a great company at a fair price.

A bet that you already know which two survive, placed before the part where most of them don't.\n\nQT @Polymarket: JUST IN: Anthropic's private valuation has officially surpassed $1,000,000,000,000.00.

See 2 related tweets

@alex_prompter: Anthropic's $965 billion round added three new investors: Micron, Samsung, and SK Hynix. Those three...
@godofprompt: RT @alex_prompter: Anthropic's $965 billion round added three new investors: Micron, Samsung, and SK...

6. elonmusk (Group Score: 93.9 | Individual: 31.8)

Cluster: 3 tweets | Engagement: 21275 (Avg: 19420) | Type: Tech

Grok Build is moving fast\n\nQT @XFreeze: xAI has been shipping Grok Build updates non-stop

If you have not been keeping track, here is what xAI has rolled out up to v0.2.11:

New commands & core features

→ Integrated 𝕏 search and much faster web search → Added /export, /login, /usage, and /config-agents → Added an interactive read-file viewer with PowerPoint text extraction → Added Always-approve mode to streamline permissions → Auto-installed shell completions for bash, zsh, and fish

Expanded platform support

→ Added Windows ARM64 and macOS x86_64 support → Improved terminal support across Warp, VTE-based terminals, JetBrains, and legacy Windows Console → Fixed copy/paste issues across Linux Wayland, WSL, and Windows

Agent & context improvements

→ Subagents now share the terminal backend, scheduler, and monitor across sessions → Added proactive system reminders and laziness detectors to keep the model on task → Improved context compaction and memory usage for chat history → Auto-backgrounds long-running bash-mode commands triggered via !

UX, media & UI upgrades

→ Boosted terminal video playback to 30fps → Added multi-image paste, drag-and-drop, and temporary macOS screenshots → Added instant loading indicators when switching models → Improved clickable markdown links and rendered links inside table cells → Smoother plan-mode controls and absolute line numbers in the edit panel

Stability & fixes

→ Increased default retry budget to ~5 minutes → Increased Unix ulimits to prevent crashes → Hardened background tools to handle timeouts and self-kills → Fixed rendering bugs across UTF-8 output, large monitors, Windows contrast, and more

xAI is moving insanely fast

Grok Build is going from early CLI to a serious agentic coding environment very quickly

See 2 related tweets

@VaibhavSisinty: Xai shipped 13 versions of grok build in the last 12 days. nobody is shipping AI coding tools at thi...
@XFreeze: xAI has been shipping Grok Build updates non-stop

If you have not been keeping track, here is what ...

7. WSJ (Group Score: 83.5 | Individual: 28.5)

Cluster: 4 tweets | Engagement: 141 (Avg: 85) | Type: Tech

SoftBank is promising to spend at least $52 billion on building a network of massive data centers in France, helping advance Europe’s goal of tech independence with what would be the continent’s largest AI infrastructure project. https://t.co/zqQO4ePyDF

See 3 related tweets

@WSJTech: SoftBank sait it would spend at least $52 billion on building a network of massive data centers in F...
@rohanpaul_ai: FT: SoftBank just pledged €75B to build Europe’s largest AI computing facility in France, turning ch...
@Techmeme: SoftBank pledges to invest up to €75B in AI computing clusters in France, first leading a €45B inves...

8. seraleev (Group Score: 83.0 | Individual: 47.7)

Cluster: 2 tweets | Engagement: 287 (Avg: 38) | Type: Tech

Your app hits the App Store. First month $5K MRR.$ 100K offer on the table.

Take it or keep building for $1M?

See 1 related tweets

@seraleev: 6 years ago my app was growing and I accepted the first offer.

I had $34k/mo and was offered$ 410k....

9. rohanpaul_ai (Group Score: 82.6 | Individual: 33.4)

Cluster: 3 tweets | Engagement: 46 (Avg: 56) | Type: Tech

Reuter: ByteDance is building its own AI data-center CPUs because running agents at TikTok scale now depends on scarce server processors, not only Nvidia GPUs.

inspired by Groq's "language processing units," they are testing both Arm and RISC-V, which lets it compare a mature commercial design against a more controllable open instruction set before mass production.

The market is seeing a 10%-35% quarterly CPU price increases and long supply delays, hence making an in-house silicon is now cost and supply-chain move, not just a prestige project.

So ByteDance wants to both reduce dependence on restricted foreign AI hardware and make inference cheaper per query.

The deeper shift is that AI agents is now turning CPUs into strategic chips. A gentic inference stresses CPUs much more because one user request can trigger many smaller steps: retrieve files, call a tool, query a database, run a model, check the answer, call another model, send data across servers, and manage memory.

However, ByteDance does not seem to have in-house chip design teams and is reportedly relying on several external partners, who are also expected to handle the actual silicon manufacturing.

reuters .com/world/china/bytedance-developing-custom-cpu-chips-support-ai-rollout-sources-say-2026-05-28/

See 2 related tweets

@zephyr_z9: RT @sheriyuo: ByteDance's AI4S (AI for Science) team is reorganizing under Yang Zhenyuan. Yang is on...
@theinformation: As Beijing pressures tech giants to move away from Nvidia, ByteDance is stepping up work on AI chips...

10. 0xSero (Group Score: 81.0 | Individual: 26.8)

Cluster: 5 tweets | Engagement: 99 (Avg: 114) | Type: Tech

24 hours of testing later.

Better instruction following
Less loopy
More attention to detail
Significantly better at writing than GPT
Better for inference optimisations and mlops
The UI has gotten better
More careful to not duplicate work

I like it https://t.co/By5z9OGHn9\n\nQT @0xSero: Anyone else notice opus-4.8 is worse than it was on launch?

They chopped him. https://t.co/7T6Ly7wpCq

See 4 related tweets

@jxnlco: Trying to maximize spend or actually getting your work done?\n\nQT @scaling01: Opus 4.8 gets score-,...
@ChrisHayduk: Opus 4.8 took twice as long, cost twice as much, and still didn’t match GPT 5.5

GOAT model 🐐\n\nQT ...

@adonis_singh: opus 4.8 is quite good, seems better aligned and a very good 'agent'.

gpt-5.5 is still my favorite ...

@NielsRogge: I was being silly, lots of things I was doing with GPT-5.5 and Opus 4.7 could have been a lot faster...

11. TechCrunch (Group Score: 79.5 | Individual: 19.9)

Cluster: 5 tweets | Engagement: 60 (Avg: 50) | Type: Tech

Founders seize on Indian court ruling to revive criticism of Google’s ad business https://t.co/M4ZtbjRGm5

See 4 related tweets

@moneycontrolcom: Google says trademark keyword Ads policy complies with Indian law after Delhi HC Hindware ruling

-D...

@techdaily24: Founders seize on Indian court ruling to revive criticism of Google’s ad business https://t.co/tii0y...
@Techmeme: Google's keyword ad practices face renewed criticism after the Delhi High Court ruled against Google...
@chandrarsrikant: Google says trademark keyword Ads policy complies with Indian law after Delhi HC Hindware ruling

ht...

12. PaulSolt (Group Score: 73.6 | Individual: 30.4)

Cluster: 3 tweets | Engagement: 129 (Avg: 52) | Type: Tech

Build an iOS app this weekend!

Install one of these skills to make the UI look good.

And teach Codex how to make apps.\n\nQT @PaulSolt: https://t.co/XJQvXEC4oq

See 2 related tweets

@PaulSolt: Use this SKILL to show Codex how to run your iOS app.

Instead of opening Xcode, it uses a Makefile ...

@PaulSolt: RT @PaulSolt: Build an iOS app this weekend!

Install one of these skills to make the UI look good. ...

13. Techweek_ (Group Score: 73.5 | Individual: 28.2)

Cluster: 3 tweets | Engagement: 3 (Avg: 4) | Type: Tech

Do not miss!!\n\nQT @Tocelot: AI is changing the nature of entertainment & who gets to make it - come discuss the AI creative opportunity at our NY Tech Week event with leaders from Adobe, Runway, Comcast, & a16z!

Featuring @c_valenzuelab of @runwayml, Greg Reed of @NBCUniversal, Hannah Elsakar | Justin Wang of @Adobe, & myself

Tues 6/2 in NYC - register below! if you’re on the waitlist, reply here and I’ll see what i can do

https://t.co/9U8I9Vhxje

See 2 related tweets

@Techweek_: RT @Tocelot: AI is changing the nature of entertainment & who gets to make it - come discuss the AI ...
@ednewtonrex: RT @mexopolis: I have decided to drop out of the AI program at Amazon. I will not be making a Punky ...

14. _philschmid (Group Score: 72.6 | Individual: 26.0)

Cluster: 4 tweets | Engagement: 101 (Avg: 133) | Type: Tech

Same as knowing how to google in the 2010s.\n\nQT @thdxr: i have seen enough proof now that using a coding agent is a deep skill

it's confusing because the people you see heavily using them produce horrible results

but that's because it's a skill! you can get better and the ceiling seems pretty high - this is very exciting to me

See 3 related tweets

@thdxr: i have seen enough proof now that using a coding agent is a deep skill

it's confusing because the p...

@seconds_0: Dax validating my existence\n\nQT @thdxr: i have seen enough proof now that using a coding agent is ...
@ferologics: here we go again\n\nQT @thdxr: i have seen enough proof now that using a coding agent is a deep skil...

15. alexeheath (Group Score: 72.4 | Individual: 41.5)

Cluster: 2 tweets | Engagement: 422 (Avg: 92) | Type: Tech

Sources: This is a leaked screenshot of Microsoft's coming super app redesign of Copilot, featuring an OpenClaw-like agent called Scout that it's planning to announce soon https://t.co/4IvkdOhIf8 https://t.co/vew3fQCUAM

See 1 related tweets

@testingcatalog: COPILOT 🔥: Microsoft is working on its own super app along with a new always-on agent called Scout a...

16. ShishirShelke1 (Group Score: 71.9 | Individual: 42.3)

Cluster: 3 tweets | Engagement: 806 (Avg: 161) | Type: Tech

Those coordinates point to the Taipei Music Center, where Computex keynote will take place.

NVIDIA is expected to unveil its ARM-based N1 and N1X laptop chips, reportedly developed in partnership with MediaTek.

Rumored specs:

Up to 20 ARM CPU cores
Up to 6144 Blackwell CUDA cores
TSMC 3nm process
Up to 128GB unified LPDDR5X memory

Dell, Lenovo, ASUS and more are expected to launch laptops powered by these chips in the near future.

This is likely what this teaser is referring to.

(Image for reference only)\n\nQT @nvidia: A new era of PC.

25.0528, 121.5990

See 2 related tweets

@TeksEdge: 🆕 Microsoft & NVIDIA teased “A new era of PC” and here it is.

Both companies posted the exact same...

@JamesMontemagno: 👀\n\nQT @tomwarren: Nvidia, Microsoft, and Arm are all teasing Nvidia’s new N1X laptop processors. I...

17. codewithimanshu (Group Score: 69.9 | Individual: 36.9)

Cluster: 2 tweets | Engagement: 530 (Avg: 197) | Type: Tech

RT @codewithimanshu: I lost my job yesterday.

Rent was due. No backup plan.

Then I remembered I still had Claude.

Claude scanned 10,000 Polymarket wallets while I slept.

Built an autonomous agent that trades 24/7. 94 trades. Zero input. Pure edge.

I'm giving away the exact setup free for 24 hours.

To get it:

Comment 'CLAUDE'
Like and Retweet this post
Follow @codewithimanshu (so I can DM you)

You only need Claude + laptop + 1 hour to deploy.

What it does:

→ Analyzes top Polymarket wallets from last 90 days → Cross-references win rates, position sizing, entry timing → Isolates 7 traders whose edge is repeatable → Builds agent that reads live news and maps to markets → Detects mispricing across outcomes → Exploits arbitrage windows → Sizes every position using Kelly criterion

Not a script. Not an if-then bot.

An agent that processes information asymmetry at machine speed.

I deployed it at 11:47 PM. Closed the laptop.

Woke up to $4,237 from$ 25 seed capital.

Wall Street pays millions for this infrastructure.

You pay $20/month for Claude.

The opportunity window:

→ Polymarket still has inefficient pricing → Most retail traders react to news manually → AI agents exploit the 15-180 second lag → Window closes when everyone runs similar systems

This edge exists right now. It won't in 12 months.

Save this post. Deploy the agent this weekend. Start with $25. Scale on evidence.

You Must Follow me @codewithimanshu so I can send you the setup guide in DM.

See 1 related tweets

@codewithimanshu: RT @codewithimanshu: I'm quitting my job to go full-in on Claude.

Just Asked it to:

Analyze mispri...

18. firstadopter (Group Score: 69.8 | Individual: 46.4)

Cluster: 2 tweets | Engagement: 5462 (Avg: 116) | Type: Tech

lol. Arm posts 6 hours after Nvidia and Microsoft. Bruh.\n\nQT @Arm: A new era of PC.

25.0528, 121.5990

See 1 related tweets

@kimmonismus: Nvidia, Microsoft and Arm posted the same “a new era of PC” teaser this week, with coordinates point...

19. MichaelDell (Group Score: 69.7 | Individual: 40.1)

Cluster: 2 tweets | Engagement: 1948 (Avg: 1142) | Type: Tech

Demand continues to accelerate with record results across all areas of our business:

Revenue: $43.8B ⬆️88% YoY ISG Revenue:$ 29.0B ⬆️181% YoY
AI Servers Revenue: $16.1B ⬆️757% YoY Traditional Servers & Networking:$ 8.5B ⬆️92% YoY
Storage Revenue: $4.3B ⬆️8% YoY ISG Op Inc:$ 3.1B ⬆️206% YoY AI Orders: $24.4B AI Backlog:$ 51.3B CSG Revenue: $14.6B ⬆️17% YoY CSG Op Inc:$ 1.2B ⬆️79% YoY Diluted EPS: $5.24 ⬆️282% YoY Non-GAAP Diluted EPS:$ 4.86 ⬆️214% YoY Q1 Cash Flow from Ops: $4.1B

During the quarter, we returned $2.1 billion to shareholders through share repurchases and dividends.

For FY'27, we now expect total company revenues to grow to $167B and EPS to grow by 74%.

Last week at Dell Technologies World, we explained more about the massive AI buildout: https://t.co/fhkf2vVDjE

Grateful to our customers, partners, and the @Dell team who make this all happen. Onward! 🙏🚀

#PlayNiceButWin\n\nQT @Dell: Q1FY27: Record revenue, EPS and Q1 cash flow – Q1 underscored the strength and agility of our operating model and the pace of innovation we continue to bring to market across the full stack of compute, storage and software.

Catch the #earnings call replay & see the Q1 report 📝 https://t.co/P4F2eeqBni

See 1 related tweets

@DivesTech: RT @MichaelDell: Demand continues to accelerate with record results across all areas of our business...

20. sairahul1 (Group Score: 69.5 | Individual: 40.7)

Cluster: 2 tweets | Engagement: 155 (Avg: 100) | Type: Tech

SOMEONE VIBE CODED A TOOL THAT FINDS BUSINESSES, READS THEIR REVIEWS, AND WRITES COLD EMAILS BASED ON THEIR OWN CUSTOMERS' COMPLAINTS

this is insane.

you type in any business type and pick a city.

it scrapes every matching business off Google Maps with 30+ data fields.

then it visits their actual websites and pulls verified emails, phone numbers, and every social media profile they have

scraped live, not from some outdated database like other tools

then the AI reads up to 50 of their Google reviews and finds their exact pain points

"clients complain photos don't show the real size of properties" or "listings take too long to sell"

then you tell it what YOUR business does.

it cross-references your offer with their specific problems and generates a fully personalized cold email for each business.

send it in 2 clicks. one by one and not bulk so it lands in the primary inbox

and all of those leads land on a GPS-mapped CRM where you can draw sales territories, optimize driving routes, track your team's activity in real time, and transcribe voice notes after meetings.

works in 200+ countries and with any business type. if they're on Google Maps you can find them

this is the most complete lead gen tool i've ever seen

AND he vibe coded the whole thing with Claude Code in 2 weeks\n\nQT @sairahul1: https://t.co/2HcA8LG2HT

See 1 related tweets

@eng_khairallah1: RT @sairahul1: SOMEONE VIBE CODED A TOOL THAT FINDS BUSINESSES, READS THEIR REVIEWS, AND WRITES COLD...