顶级技术推文精选 - 2026-05-19

科技每日简报 2026年5月19日

Today's top tech conversations are led by @ClementDelangue, whose post about 'Very cool to see Cursor doubli...' garnered the highest engagement. Key themes trending across the top stories include claude, https, model, world, dialog. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. ClementDelangue (Group Score: 233.5 | Individual: 36.4)

Cluster: 9 tweets | Engagement: 217 (Avg: 100) | Type: Tech

Very cool to see Cursor doubling down on training great models. In my opinion, ultimately all serious companies in AI will want to train models themselves, based on open-source instead of outsourcing AI to others via APIs!\n\nQT @cursor_ai: Introducing Composer 2.5, our most powerful model yet.

It's more intelligent, better at sustained work on long-running tasks, and more reliable at following complex instructions.

For the next week, we’re doubling the included usage of the model. https://t.co/N87ojcXlOC

See 8 related tweets

@VaibhavSisinty: Cursor genuinely cooked with this one. 🤯

composer 2.5 just dropped, their new ai coding model, and ...

@0xSero: Cursor X Elon is a really good fit.

They have similar mindsets and drive, I would be shocked if th...

@minchoi: Cursor Composer 2.5 is coming for Opus 4.7 and GPT-5.5.

Near-frontier coding performance.

But way ...

@altryne: New Composer from Cursor team! Great to see their ack. to the Kimi base + how much they moved the mo...
@mark_k: Cursor just released Composer 2.5. 🔥

What stands out to me is that this is not just about better be...

2. vasuman (Group Score: 232.7 | Individual: 35.7)

Cluster: 7 tweets | Engagement: 31 (Avg: 153) | Type: Tech

Voice is one of the most overlooked commerce surfaces in any company. The customer dialed, they’re committed for ten minutes, they can't tab away, they can't comparison shop. It's the highest-attention channel a business has, and yet does not get treated like one.

The reason has always been headcount. Every phone call requires a human, and humans cap at ~40 calls a shift.

Voice AI changes the game with consistent scripts, automated responses and routing, and never changes tonality or delivery to customers.

Congrats to Voice AI - customer service over the phone is getting the biggest re-rating of the decade.\n\nQT @polyaivoice: Starting today, we're opening our Agentic Dialog Platform to every enterprise builder.

Our dialog agents have resolved 1 billion+ customer conversations for clients like FedEx, Unicredit, PG&E, Marriott, Foot Locker, and many more.

These aren't easy conversations. They solve problems like:

A patient booking medical transport who needs insurance verified on the spot. A homeowner calling their utility company about a gas leak. A cardholder figuring out why their must-have purchase was declined.

Standard conversational AI was never built for this. It was designed for chat, adapted for voice later. It generates responses, but can't do what dialog requires: hold context under pressure, navigate ambiguity in real time, and actually resolve problems.

So we built a better model.

Our proprietary model Raven was built from the ground up specifically for dialog. Agent harness in the weights, not bolted on through prompts that drift under pressure. And in our platform, you can deploy Raven as your default or bring in GPT-5, Claude, Gemini, whatever model fits your use case or regulatory requirement.

Now that the Agentic Dialog Platform is open, any team can create, test, and deploy dialog agents on the same model and infrastructure the world’s top brands trust on their hardest days. This opens up the pool of builders across your entire enterprise. The person who knows customers best, who runs operations, who owns the customer journey: they're all builders now.

Two ways to build:

Poly Agent Builder: Describe your use case in natural language, and it configures your agent, knowledge base, and conversation flows automatically. Production-ready in ten minutes.

Agent Development Kit (ADK): Developers use this to build dialog agents the same way they build everything else. Use your own IDE, a coding assistant like Claude, version with Git, deploy from your terminal.

Get started now: https://t.co/ifZOy1uEBz

See 6 related tweets

@aakashgupta: PolyAI runs half of Las Vegas, Foot Locker, FedEx, and every Gordon Ramsay restaurant.

Now they're ...

@etnshow: BREAKING: Nikola Mrkšić, Co-Founder and CEO @PolyAIVoice, will be joining the show tomorrow to discu...
@itsolelehmann: I added a voice agent to my product site in 10 minutes

here's how:

Give Poly Agent Builder your...

@godofprompt: Call centers are about to get absolutely cooked.

PolyAI just opened the same voice agent platform u...

@heyrobinai: The BPO industry is going to look very different in 24 months

1B+ calls already handled and 95% of ...

3. gdgtify (Group Score: 196.1 | Individual: 47.9)

Cluster: 6 tweets | Engagement: 998 (Avg: 59) | Type: Tech

RT @BostonDynamics: Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability. https://t.co/yXG0dY6Qqf

See 5 related tweets

@jason: We're going to have a massive surplus of extremely strong and dexterous robots to do extremely unnec...
@AndrewCurran_: Great demo, living up to the name. From the blog post:

'Atlas uses reinforcement learning (RL) to l...

@rohanpaul_ai: Boston Dynamics showed Atlas lifting and carrying a 100+ lb mini-fridge, using reinforcement learnin...
@carlosdponx: This is a great example of a whole body task where humanoids have an advantage

Carrying that load w...

@tunguz: So need this.\n\nQT @BostonDynamics: Everyone asks if Atlas can bring them a drink, but this robot c...

4. WSJ (Group Score: 166.2 | Individual: 37.8)

Cluster: 6 tweets | Engagement: 451 (Avg: 102) | Type: Tech

Breaking: A jury sided with Sam Altman and OpenAI in a landmark lawsuit brought by Elon Musk, who alleged that the startup behind the world’s most popular chatbot “stole a charity” when it converted into a for-profit company https://t.co/gTzyhzKSG4

See 5 related tweets

@Reuters: BREAKING: A US jury ruled against Elon Musk in his lawsuit against OpenAI, finding the artificial in...
@TFTC21: A federal jury unanimously ruled against Elon Musk in his lawsuit against Sam Altman and OpenAI.

T...

@moneycontrolcom: BREAKING 🚨 | A California jury sided with OpenAI and CEO Sam Altman in the closely watched legal bat...
@business: A jury rejected Elon Musk’s claims that OpenAI under Sam Altman’s leadership betrayed its mission to...
@WSJTech: A jury sided with Sam Altman and OpenAI in a landmark lawsuit brought by Elon Musk, who alleged that...

5. theinformation (Group Score: 152.2 | Individual: 33.1)

Cluster: 6 tweets | Engagement: 6 (Avg: 15) | Type: Tech

Today on TITV:

-Microsoft executives sound alarm over GitHub | @aaronpholmes

-BlackRock weighs up to $10B investment in SpaceX IPO | @coryweinberg

-Genspark AI COO @sang_wen on his approach to the AI application race

-Anthropic, OpenAI generate 89% of AI startup revenue | @steph_palazzolo\n\nQT @theinformation: The Information | TITV | May 18, 2026 https://t.co/YvQP0rZLBt

See 5 related tweets

@theinformation: Today on TITV:

-Microsoft executives sound alarm over GitHub | @aaronpholmes

-BlackRock weighs up ...

@theinformation: Anthropic and OpenAI now generate 89% of revenue among 34 leading AI startups tracked by The Informa...
@JackKuhr: Cursor (soon to be tucked into SpaceX) is generating the third most ARR at $2.7B, behind Anthropic a...
@exec_sum: Anthropic and OpenAI now capture 89% of all AI startup revenue, per The Information

The data places...

@steph_palazzolo: RT @amir: Anthropic & OpenAI’s share of AI model and application revenues generated by 34 top AI...

6. wallstengine (Group Score: 146.8 | Individual: 31.2)

Cluster: 5 tweets | Engagement: 43 (Avg: 125) | Type: Tech

ODYSSEY LAUNCHES AGORA-1, A MULTI-AGENT WORLD MODEL

Agora-1 lets multiple humans or AI agents interact inside the same generated world in real time.

The research preview simulates a multiplayer GoldenEye-style deathmatch with up to four players.

Odyssey says the model maintains a shared world state across participants, tracks player actions, and streams generated pixels to each player simultaneously.

The important part is the architecture:

simulation and rendering are separated.

That means Agora-1 can generate different views of the same world from multiple independent player perspectives.

This is less “AI video generator” and more “learned game engine.”\n\nQT @odysseyml: Introducing Agora-1, a multi-agent world model.

Multiple participants—human or AI—can now interact inside the same world simulation, all in real-time.

Try our playable research preview today, with Agora-1 simulating a multiplayer GoldenEye deathmatch! https://t.co/y41LIAd4oz

See 4 related tweets

@alex_prompter: Video games are cooked.

Movies are cooked.

Passive content is cooked.

Odyssey is building interac...

@shiri_shh: this is the ChatGPT moment for video games

Agora-1 generates a live 4-player GoldenEye match with n...

@0x_kaize: 4 players. NO game engine. every frame GENERATED BY AI

Odyssey just dropped Agora-1

I repeat: ...

@rohanpaul_ai: Agora-1, a multi-agent world model from Odyssey just exposed the next bottleneck for world models: k...

7. business (Group Score: 136.4 | Individual: 31.9)

Cluster: 6 tweets | Engagement: 74 (Avg: 76) | Type: Tech

Pope Leo XIV’s launch of his first encyclical focusing on artificial intelligence next week will feature an unexpected guest: Christopher Olah, co-founder of Anthropic. https://t.co/UQMRYvQllK

See 5 related tweets

@Forbes: Pope Leo XIV will present a major Church document on artificial intelligence next week and he will b...
@Forbes: Billionaire Anthropic co-founder Chris Olah will join Pope Leo XIV at the Vatican, where Leo will pr...
@exec_sum: BREAKING: Pope Leo ‌will address the rise of artificial intelligence in his first in-depth text outl...
@Cointelegraph: ⚡️ UPDATE: Pope Leo XIV will meet Anthropic co-founder Christopher Olah next week to discuss AI. htt...
@Polymarket: JUST IN: Pope Leo XIV will meet with Anthropic co-founder Christopher Olah next week to discuss AI....

8. Mayhem4Markets (Group Score: 124.8 | Individual: 30.8)

Cluster: 5 tweets | Engagement: 31 (Avg: 55) | Type: Tech

it's that time again.

rebuilding llama.cpp.

time to unlock the MTP acceleration for Qwen3.6.

this will be going on a secondary lab system.

it's built from spare parts and continues to show its value every single day.

i'll share specs and config notes soon.\n\nQT @ggerganov: llama.cpp adds MTP for the Qwen3.6 family

This is a significant milestone for the local AI ecosystem. The performance jump with these changes is massive and elevates local inference on commodity hardware further.

Special thanks to Aman Gupta for leading this development!

https://t.co/vjaMwEpIaR

See 4 related tweets

@mervenoyann: finally faster Qwen3.6 models with MTP support ⚡️

brb updating my Pi & Hermes setup 🤝\n\nQT @g...

@huggingface: RT @ggerganov: llama.cpp adds MTP for the Qwen3.6 family

This is a significant milestone for the lo...

@victormustar: llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀

Qwen3.6-27B den...

@ivanfioravanti: llama.cpp has MTP for Qwen3.6 now!!! I need to try this asap!\n\nQT @ggerganov: llama.cpp adds MTP f...

9. Parul_Gautam7 (Group Score: 123.7 | Individual: 34.0)

Cluster: 4 tweets | Engagement: 133 (Avg: 124) | Type: Tech

Most AI systems can remember.

Very few can actually get better from experience.

That’s the difference between context… and intelligence.\n\nQT @yyyiiillluuu: Claude Code can now self-improve with this plugin.

Introducing claude-smart — an open-source plugin that helps Claude Code learn from every session.

Memory helps Claude Code remember what happened. claude-smart helps Claude Code improve what it does next.

Example: Claude Code runs npm test without --run, and the command hangs in your repo.

Memory stores: “npm test kept hanging.”

claude-smart learns: “When running tests in this repo, use npm test -- --run because default watch mode hangs.”

claude-smart’s learnings are reusable and actionable, even across different projects. It can also reduce unnecessary planning iterations and token use by 70%+ on similar future tasks.

Runs locally. 100% open source. No data is shared.

Install: npx claude-smart install

With Codex: npx claude-smart install --host codex

GitHub: https://t.co/D4F0v5FnGk

See 3 related tweets

@svpino: This is kind of interesting: it just not memory, but "smart memory"?

Assuming this works, this coul...

@Origin_AI_01: AI coding tools are evolving from assistants into systems that accumulate operational experience ove...
@Origin_AI_01: RT @yyyiiillluuu: Claude Code can now self-improve with this plugin.

Introducing claude-smart — an ...

10. Parul_Gautam7 (Group Score: 123.6 | Individual: 32.2)

Cluster: 5 tweets | Engagement: 150 (Avg: 124) | Type: Tech

We’re moving from “using AI tools” to building AI teams.\n\nQT @lobehub: Meet your Chief Agent Operator in LobeHub now.

You don't manage agents. Your CAO does.

LobeHub hires agents from a 273K-skill marketplace, runs them in the cloud 24/7, and reports through IM apps you already live in.

You run the strategy. We run the agents. https://t.co/FmscryTirl

See 4 related tweets

@sukh_saroy: Everyone selling you a smarter chatbot

LobeHub just shipped a manager for the chatbots

A Chief Age...

@rohanpaul_ai: Managing AI agents by hand is yesterday’s job. hire a Chief Agent Operator and move on.

LobeHub lau...

@svpino: Beyond all of the marketing speak, this creates an orchestration layer for agents.

Basically, this ...

@heynavtoor: you don't manage AI agents. you hire a Chief Agent Operator and let it manage them for you.

LobeHub...

11. markgurman (Group Score: 116.4 | Individual: 31.2)

Cluster: 6 tweets | Engagement: 751 (Avg: 978) | Type: Tech

Long planned, the new AI system in the Shortcuts app will allow the use of natural language to create and install a Shortcut.\n\nQT @markgurman: NEW: Apple plans several new AI features across iOS 27, looking to better compete with Android. That includes new AI writing tools like a Grammar Checker, AI-created Wallpapers and new Shortcuts app with AI-based shortcut creation. https://t.co/kn4khH4NJN

See 5 related tweets

@markgurman: The new AI writing tools include a Grammar Checker that works similarly to Grammarly. It joins the o...
@markgurman: NEW: Apple plans several new AI features across iOS 27, looking to better compete with Android. That...
@tunguz: Such advance. So much cutting edge.\n\nQT @Polymarket: JUST IN: Apple is reportedly planning AI-powe...
@Polymarket: JUST IN: Apple is reportedly planning AI-powered writing help, shortcut creation, & wallpaper ge...
@business: Apple is planning new AI features in iOS 27 for writing, app shortcuts and custom wallpapers in an e...

12. eng_khairallah1 (Group Score: 115.0 | Individual: 35.3)

Cluster: 4 tweets | Engagement: 29 (Avg: 66) | Type: Tech

🚨 this new startup just showed where advertising is heading.

1 prompt. ad agents generating thousands of variations. 5 minutes.

watch the demo. bookmark it now.

most people are still designing ads one by one. manually. proudly.

the workflow already moved on without them.

worth more than most "creative strategy" courses people are still selling.\n\nQT @Creatify_AI: Introducing Creatify Agent.

The world’s first multi-agent orchestrator that makes viral ads without AI slop.

Drop an ad idea below, and we’ll build it for you. https://t.co/WcI2hk85FV

See 3 related tweets

@heynavtoor: > make 5 ads by hand

pick the one that "feels right" run it for 2 weeks it flops start over ...

@shiri_shh: Sir…they’re spinning up entire ad campaigns in minutes https://t.co/QLtO8piBJa\n\nQT @Creatify_AI: I...
@eng_khairallah1: RT @eng_khairallah1: 🚨 this new startup just showed where advertising is heading.

1 prompt. ad agen...

13. secureainow (Group Score: 107.8 | Individual: 55.0)

Cluster: 3 tweets | Engagement: 3960 (Avg: 170) | Type: Tech

RT @FearedBuck: College graduates were pissed after their school used AI to announce graduates’ names and missed hundreds of names https://t.co/dwz6xFIWiv

See 2 related tweets

@Hesamation: this is the same academia that used bs AI detectors to falsely accuse students of cheating and prohi...
@chatgpt21: People BOO the new AI system for failure is like booing schools in 1999 for having internet problems...

14. sairahul1 (Group Score: 101.3 | Individual: 34.1)

Cluster: 4 tweets | Engagement: 111 (Avg: 47) | Type: Tech

🚨 You’re probably using Claude wrong.

Anthropic just released a 24-minute workshop showing how to prompt Claude properly.

Taught by the actual team behind Claude.

100% free. No registration.

Honestly better than a lot of paid AI courses.

The first 8 minutes alone are worth it.

Bookmark this.\n\nQT @sairahul1: https://t.co/SNvxZeAemb

See 3 related tweets

@Origin_AI_01: Most people are still guessing how to prompt Claude.

Meanwhile, Anthropic quietly released a 24-min...

@sairahul1: RT @eng_khairallah1: 🚨 Anthropic just showed a 24-minute workshop on how to actually do prompts for ...
@sairahul1: RT @sairahul1: 🚨 You’re probably using Claude wrong.

Anthropic just released a 24-minute workshop s...

15. rohanpaul_ai (Group Score: 97.6 | Individual: 33.2)

Cluster: 4 tweets | Engagement: 84 (Avg: 111) | Type: Tech

Anthropic just acquired Stainless to make Claude agents better at reaching real software systems through cleaner SDKs, CLIs, and MCP servers.

Shows again that better connectivity infrastructure will matter as much as better model intelligence once companies start asking agents to perform actual work.

Stainless has generated every official Anthropic SDK since the early Claude API days, which means Anthropic is buying a toolchain that already sits inside its developer platform.

See 3 related tweets

@StockSavvyShay: Anthropic acquired Stainless which is the company behind its official Claude SDKs.

The deal shows t...

@MTSlive: SITUATION DETECTED: Anthropic has acquired Stainless, a leader in SDK and MCP server tooling that ha...
@AnthropicAI: Anthropic is acquiring @stainlessapi, an SDK and MCP server platform that has powered every Anthropi...

16. NathanpmYoung (Group Score: 93.9 | Individual: 52.7)

Cluster: 4 tweets | Engagement: 1950 (Avg: 131) | Type: Tech

RT @dianemontagna: JUST IN: Vatican announces that Pope Leo XIV’s first encyclical — titled Magnifica Humanitas, on the safeguarding of the human person in the age of AI — will be presented at 11:30am on Monday, May 25, in the Vaticanʼs Synod Hall, in the presence of the Holy Father.

Speakers at the presentation will include:

Cardinal Víctor Manuel Fernández, Prefect of the Dicastery for the Doctrine of the Faith;

Cardinal Michael Czerny, Prefect of the Dicastery for the Service of Integral Human Development;

Professor Anna Rowlands, Political Theology, including Catholic Social Teaching, and theological ethics of human migration, Department of Theology and Religion, Durham University, United Kingdom;

Christopher Olah, co-founder of Anthropic (USA) and head of interpretability research for artificial intelligence;

Dr. Leocadie Lushombo, Political Theology and Catholic Social Thought, Jesuit School of Theology / Santa Clara University, California.

Concluding remarks will be delivered by thel Secretary of State, Cardinal Pietro Parolin.

The presentation will also include an address by Pope Leo XIV.

Magnifica Humanitas was signed and dated on May 15, the 135th anniversary of the promulgation of Pope Leo XIII’s Encyclical Letter Rerum Novarum.

See 3 related tweets

@AndrewCurran_: Update on Pope Leo XIV's AI encyclical: 𝘔𝘢𝘨𝘯𝘪𝘧𝘪𝘤𝘢 𝘏𝘶𝘮𝘢𝘯𝘪𝘵𝘢𝘴. It will be presented on Monday, May 25t...
@Hadas_Gold: RT @AndrewCurran_: Update on Pope Leo XIV's AI encyclical: 𝘔𝘢𝘨𝘯𝘪𝘧𝘪𝘤𝘢 𝘏𝘶𝘮𝘢𝘯𝘪𝘵𝘢𝘴. It will be presented...
@ianmiles: RT @breeadail: VATICAN CITY—The Holy See has just announced that the much-anticipated first Encyclic...

17. VaibhavSisinty (Group Score: 92.2 | Individual: 19.0)

Cluster: 6 tweets | Engagement: 54 (Avg: 55) | Type: Tech

okay this one's hard to ignore. 😨

Microsoft's AI chief just said most office jobs get automated in 12 to 18 months. Lawyers, accountants, marketers, project managers.

Everyone's arguing about whether he's right.

He's not the one you should be listening to anyway. The people winning the next 18 months aren't debating the timeline. They're already learning to run the AI instead of competing with it.\n\nQT @Polymarket: NEW: Microsoft AI chief predicts AI will automate most white-collar work within 18 months.

See 5 related tweets

@rohanpaul_ai: Microsoft’s AI chief is warning that AI may automate most computer-based professional tasks within 1...
@WesRoth: Mustafa Suleyman has warned that most professional computer-based work could be automated by AI with...
@Cointelegraph: 🚨 LATEST: Microsoft AI CEO Mustafa Suleyman predicts all white-collar work will be fully automated b...
@moneycontrolcom: Microsoft AI chief believes AI will automate all white-collar jobs in 18 months

18. FirstSquawk (Group Score: 91.7 | Individual: 25.2)

Cluster: 5 tweets | Engagement: 370 (Avg: 72) | Type: Tech

ELON MUSK SAYS THAT IN 10 YEARS, 90% OF ALL DISTANCE DRIVEN WILL BE THROUGH AI IN SELF-DRIVING CARS

See 4 related tweets

@MarioNawfal: 🇺🇸 Elon: in 10 years, 90% of all distance driven will be driven by AI, and owning your own car will ...
@niccruzpatane: Elon Musk says that in 10 years, 90% of the distance driven will be driven by AI in self-driving car...
@XFreeze: Elon Musk on Tesla Full Self-Driving progress:

"The Tesla Full Self-Driving software is really just...

@elonmusk: RT @SawyerMerritt: Elon Musk in new interview: "10 years from now, probably 90% of all distance driv...

19. wallstengine (Group Score: 90.4 | Individual: 28.9)

Cluster: 4 tweets | Engagement: 124 (Avg: 125) | Type: Tech

$META’S MAY 20 LAYOFFS ARE ALSO AN AI REORG

Internal document says Meta will conduct layoffs in three batches globally, each at 4 a.m. local time.

North America employees were told to work from home Wednesday.

Leaders are also expected to announce org changes.

Meta plans to move 7,000 people to new AI workflow initiatives and eliminate managerial roles.

Reuters previously reported the May 20 wave would cut about 10% of Meta’s workforce.

See 3 related tweets

@Techmeme: Internal memo: Meta is reassigning 7,000 workers to four new units focused on building AI tools, two...
@HedgieMarkets: 🦔Meta detailed its May 20 layoff plans in an internal memo today, confirming 10% of its global workf...
@Reuters: Exclusive: Meta is planning to carry out its layoffs of 10% of its employees in three batches global...

20. pierceboggan (Group Score: 87.6 | Individual: 23.0)

Cluster: 5 tweets | Engagement: 78 (Avg: 32) | Type: Tech

Remote control is now generally available in GitHub Copilot CLI and @code :)\n\nQT @github: Start work on your computer, continue your local session anywhere. 📲

Remote control for GitHub Copilot CLI and @code sessions is now generally available. https://t.co/wwSEBd5lqL https://t.co/Yc5R6tBfBl

See 4 related tweets

@code: 🚀 Remote control for GitHub Copilot CLI and @code sessions is now generally available.

Monitor prog...

@github: Start work on your computer, continue your local session anywhere. 📲

Remote control for GitHub Copi...

@burkeholland: Y'all can stop walking around with your laptops cracked open now\n\nQT @github: Start work on your c...
@JamesMontemagno: RT @code: 🚀 Remote control for GitHub Copilot CLI and @code sessions is now generally available.

Mo...