科技热门推文 — 2026年9月3日

今日科技动态：随着谷歌、Anthropic、Muse 和 OpenAI 分别展示其在编程、智能体系统、网络安全及新型模型架构方面的进展，人工智能领域的竞争进一步加剧。企业对实用型自动化解决方案的需求日益增长，推动新一批专注于工程应用的初创公司涌现；与此同时，能够处理日常行政事务的实用型智能体也备受关注。博通报告称人工智能半导体业务收入大幅增长，英伟达呼吁各国加快基础设施投资，泰达公司则致力于推动非洲的数字普惠。此外，谷歌避免了广告技术业务被拆分的局面，围绕人工智能训练版权保护的争论仍在持续。

1. pingToven (Group Score: 1034.1 | Individual: 57.5)

Cluster: 45 tweets | Engagement: 1286 (Avg: 81) | Type: Tech

RT @OfficialLoganK: Introducing Gemini 3.8 Flash, another jump in Gemini's agentic + coding capabilities, and our 3rd updated Flash model in only 6 weeks...

This model has been a ton of fun to work with, excited to see what you all think! https://t.co/Cj07lCBtp8

See 44 related tweets

@rseroter: RT @antigravity: Introducing Gemini 3.8 Flash in Antigravity!

3.8 Flash is our most intelligent wor...

@DynamicWebPaige: $0.75 in /$ 3.75 out opus 5 is $5 /$ 25

...and it's ahead on terminal coding, finance agents, chart...

@Lentils80: Gemini 3.8 Flash is now officially released https://t.co/GztJGtSAht\n\nQT @lyraxana: great choice no...
@demishassabis: Introducing Gemini 3.8 Flash, another upgrade in under a month! And the new 3.8 Flash Cyber pushes t...
@DynamicWebPaige: ✍️✨ Gemini 3.8 Flash really is a step-change improvement, especially for coding, agent use cases, an...

2. svpino (Group Score: 575.6 | Individual: 52.2)

Cluster: 19 tweets | Engagement: 967 (Avg: 157) | Type: Tech

There’s an explosion of companies looking for help with AI.

I suspect we'll see many more examples like Wonderful: a company that sends engineers to the client’s location and helps them automate anything and everything.\n\nQT @wonderful_ai: Wonderful has raised a $550M Series C at a$ 5 billion valuation.

This latest funding was led by @insightpartners, with @salesforce joining as new investors and @IndexVentures, @BessemerVP, @IVP, @VineVenturesLP, and @9yardscapital returning.

In 20 months since launch, we're in production with hundreds of agents, systems, and workflows across a dozen verticals in over 30 markets.

With this new capital, we’ll be overdelivering for our customers, hiring hundreds of cracked engineers, FDEs, and former founders, and expanding our product capabilities across the entire enterprise.

Stay tuned 👀

See 18 related tweets

@rohanpaul_ai: RT @wonderful_ai: Wonderful has raised a $550M Series C at a$ 5 billion valuation.

This latest fund...

@Meer_AIIT: why most enterprise ai never gets past the pilot:

a team runs a test inside one department. it work...

@peterwildeford: Five recent trends in AI that are very worrying in combination:

AI companies are increasingly un...

@wallstengine: AI STARTUP TARGETS DATA CENTER EFFICIENCY WITH “AI PHYSICISTS”

Physical Superintelligence is launch...

@shiri_shh: Wonderful reaching $5B in 20 months is a masterclass in enterprise execution.

They skipped the "API...

3. Scobleizer (Group Score: 554.0 | Individual: 47.9)

Cluster: 24 tweets | Engagement: 1920 (Avg: 188) | Type: Tech

RT @finkd: Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter. This is the biggest jump we've made so far on coding and agentic work. Try it in Muse Code and our API.

Next up 🍉 and Muse Spark open weights releases coming soon. https://t.co/XQQEDEJGD7

See 23 related tweets

@eliebakouch: long context benchmark (MRCR) going crazy, very interesting

would be great to have other benchmarks...

@rohanpaul_ai: Meta released Muse Spark 1.3. Reports 20% fewer tool calls and 25% fewer tokens than 1.2.

So long c...

@WesRoth: Muse Spark 1.3 is now available in Muse Code and the Meta Model API, with improvements in reasoning,...
@altryne: Summer is officially over!

Everyone's rushing to release .1 updates before Astra drop (tomorrow?) ...

@victormustar: omg « open weights release » incoming!!\n\nQT @finkd: Muse Spark 1.3 is rolling out today with front...

4. wallstengine (Group Score: 403.0 | Individual: 43.0)

Cluster: 13 tweets | Engagement: 558 (Avg: 182) | Type: Tech

BROADCOM $AVGO Q3’26 EARNINGS HIGHLIGHTS

🔹 Revenue: $29.6B (Est.$ 29.36B) 🟢; +86% YoY 🔹 Adj. EPS: $3.32 (Est.$ 3.24) 🟢; +96% YoY 🔹 AI Semiconductor Revenue: $16.7B; +221% YoY 🔹 Non-GAAP Oper Income:$ 20.1B (Est. $19.73B) 🟢; +92% YoY

Q4 Guide: 🔹 Revenue: ~ $34.8B (Est.$ 35.03B) 🔴 🔹 Non-GAAP Oper Income: ~66% of projected revenue 🔹 AI Semiconductor Revenue: $21.7B

Segment Net Revenue: 🔹 Semiconductor Solutions: $20.8B; +127% YoY 🔹 Infrastructure Software:$ 8.8B; +29% YoY

Other Q3 Metrics: 🔹 Free Cash Flow: $13.7B (Est.$ 13.76B) 🟡; +95% YoY 🔹 Cash Flow From Operations: $14.2B; +98% YoY 🔹 AI Semiconductor Revenue:$ 16.7B; +221% YoY

Capital Return: 🔹 Quarterly Common Stock Dividend: $0.65/share 🔹 Cash Dividend Paid:$ 3.1B

Comments: 🔸 “Demand for our custom AI accelerators and networking continues to be very strong.”

🔸 “In Q4 the momentum continues, and we expect AI semiconductor revenue to accelerate to $21.7 billion, up 236% year-over-year.”

See 12 related tweets

@wallstengine: SNOWFLAKE $SNOW Q2’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $1.55B (Est.$ 1.48B) 🟢; +35% YoY 🔹 Adj. EPS: $...

@wallstengine: $AVGO trading lower after Q3 earnings beat, but Q4 revenue guidance came in slightly below estimates...
@wallstengine: HEWLETT PACKARD ENTERPRISE $HPE Q3’26 EARNINGS HIGHLIGHTS

🔹 Revenue: $12.2B (Est.$ 11.91B) 🟢; +34% ...

@wallstengine: $SNOW +24% after a broad beat and strong guide. Product revenue guidance implies around ~37% growth ...
@wallstengine: NETAPP $NTAP Q1’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $2.03B (Est.$ 1.84B) 🟢; +30% YoY 🔹 Adj. EPS: $2.5...

5. t_blom (Group Score: 324.1 | Individual: 42.1)

Cluster: 12 tweets | Engagement: 13006 (Avg: 2250) | Type: Tech

RT @claudeai: We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work. https://t.co/8P9PSrWPi3

See 11 related tweets

@WesRoth: Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, calling them its most advanced models...
@code: 🤖 Claude Fable 5.1 is now generally available in @code with GitHub Copilot.

Give it a try today! ht...

@genspark_ai: Claude Fable 5.1 just landed in Genspark Code Agent and Claw. Day one. Ready to use right now. https...
@trq212: it’s a very good model, I spent a lot of time diving into it- longer write up coming

but try it on...

@adonis_singh: its incredibly gpt-esque, works way longer on tasks but also brings better results than fable-5

wri...

6. PaulSolt (Group Score: 288.8 | Individual: 58.7)

Cluster: 8 tweets | Engagement: 1632 (Avg: 152) | Type: Tech

RT @OpenAI: As we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly accessible.

Astra represents a significant advance in cybersecurity capability, reaching the Critical threshold under our Preparedness Framework.

We're previewing how we evaluated the model, how its safeguards have advanced alongside its capabilities, and what we'll continue to learn and improve. https://t.co/OrrTgdU90K

See 7 related tweets

@levie: AI for cyber is about to go vertical. The models increasingly becoming insanely good at finding and ...
@romainhuet: Astra is significantly more token efficient and more capable at finding vulnerabilities and developi...
@rohanpaul_ai: OpenAI says Astra is its first model to reach the Critical cybersecurity capability threshold.

Unde...

@VaibhavSisinty: OpenAI officially confirmed Astra is the first AI model to reach their Critical cybersecurity thresh...
@WesRoth: 🚨 Astra has officially crossed the CRITICAL cybersecurity threshold.

This is the first OpenAI model...

7. ChrisGPT (Group Score: 283.6 | Individual: 43.7)

Cluster: 11 tweets | Engagement: 1035 (Avg: 194) | Type: Tech

WAIT PLEASE TELL ME OPENAI PARALLELIZED RECURRENT DEPTH.

Microsoft Research already showed us in June that you can run K latent blocks in parallel for R recurrent iterations. LoopCoder has also scaled recurrent depth to 40B parameters in July + a few others.

But NOBODY has publicly deployed this architecture in a frontier class commercially available model anywhere near this scale (1T+) . If Astra is doing width × depth at frontier scale instead of simply looping one latent state, I’d expect some big benchmark jumps.\n\nQT @steph_palazzolo: OpenAI’s Astra AI uses a new reasoning approach called “recurrent depth.” Though it can help model costs and performance, researchers are concerned bc it obscures a model’s thinking process, making it more difficult to monitor.

w/ @amir @rocketalignment

https://t.co/ksprupu2h5

See 10 related tweets

@steph_palazzolo: OpenAI’s Astra AI uses a new reasoning approach called “recurrent depth.” Though it can help model c...
@kimmonismus: Holy, Astra is build different: OpenAI’s Astra model reportedly uses “recurrent depth” to improve co...
@teortaxesTex: Did OAI finally make the looped meme work?\n\nQT @steph_palazzolo: OpenAI’s Astra AI uses a new reas...
@Hesamation: ASTRA USES A NEW REASONING TECHNIQUE CALLED "RECURRENT DEPTH".

TL;DR:

instead of stacking more ...

@haider1: openai has made another o1-level breakthrough

Astra uses a new reasoning approach called "recurrent...

8. tether (Group Score: 207.7 | Individual: 45.6)

Cluster: 9 tweets | Engagement: 156 (Avg: 36) | Type: Tech

Tether ❤️ Africa Where you are born should never limit your potential. True progress begins with access to stable money, stable energy, stable communications, and now also stable intelligence.

Today, we introduce QVAC TranslatePsy / Afri SLM, a lightweight AI translation model built for the hundreds of millions across Africa who were left behind by the digital divide.

Running entirely on-device, 100% free, and completely offline, it translates 19 African languages natively. Village squares become classrooms, farmers unlock vital techniques, and life-saving medical knowledge speaks in the mother tongue.

We are not bringing people into the world of AI. We are bringing AI into their world.

Tether. Unstoppable together.

Learn more: https://t.co/8y0kOh1n2i

See 8 related tweets

@paoloardoino: AI inclusion is the next frontier for Tether expansion.
billions of people can't afford to pay fo...
@qvac: https://t.co/yb8Aqcjvxj\n\nQT @tether: Tether ❤️ Africa Where you are born should never limit your p...
@paoloardoino: Tether ❤️ Africa 🌍\n\nQT @qvac: Tether & QVAC are investing into AI inclusion for the hundreds of mi...
@tether: Tether Releases Open-Source AI Translation Models for African and European Languages

9. StockMKTNewz (Group Score: 190.6 | Individual: 34.8)

Cluster: 8 tweets | Engagement: 300 (Avg: 276) | Type: Tech

Nvidia $NVDA CEO Jensen Huang called on Group of 20 nations to accelerate their adoption of AI as a way to enhance growth

“Every single country needs to build infrastructure so you can support your own local economy”

In his remarks, Huang said that the “single worst outcome” for a country is “being left behind.” The Nvidia chief warned that could happen if the public and policymakers let fears about AI drive the conversation about the technology.

“We have to make sure that we are balanced in talking about it”

(Source Bloomberg)

See 7 related tweets

@StockSavvyShay: $NVDA CEO Jensen Huang told G20 leaders that “every single country needs to build infrastructure” to...
@business: Nvidia's Jensen Huang called on G20 nations to accelerate their adoption of AI as a way to enhance g...
@clashreport: Nvidia CEO Jensen Huang on AI:

We’re at the beginning of an industrial revolution, and the technolo...

@PolymarketMoney: BREAKING: Nvidia CEO Jensen Huang declares “every single country needs to build infrastructure” for ...
@WOLF_Financial: $NVDA CEO Jensen Huang at the G20:

"Every single country needs to build infrastructure so that you ...

10. OliviaReedai (Group Score: 187.0 | Individual: 38.3)

Cluster: 7 tweets | Engagement: 149 (Avg: 56) | Type: Tech

The most useful AI agent might end up being the least impressive one to demo.

Not solving olympiad math.

Just moving the meeting, booking the trip, replying to the email and making the call without needing babysitting.

That’s what Catch is going after.

https://t.co/sKbbLuhOEH\n\nQT @therealnirs: Big NEWS: Today we're launching Catch AI, an admin-assistant for busy executives.

We started Catch with a simple idea: AI should actually do the administrative work executives hate doing themselves.

Not suggest. Not summarize. Do. >> https://t.co/HI80mrmpay

See 6 related tweets

@OtakuMachine: AI agents can code for hours, solve difficult benchmarks and generate huge amounts of content.

Mean...

@TheoBuildsAI: The benchmark for AI agents should probably be:

Did it finish the task?

Not “could it reason about...

@NovaIAHQ: AI assistants are usually good at telling you what needs to be done.

Catch is more interesting beca...

@ai_explorer25: Most calendar AI hands you a morning brief and a few suggestions. That's where it stops.

Catch actu...

@Scobleizer: Most AI still waits for you.

Catch books the hotel. If the fare drops it rebooks. You can call it. ...

11. alex_prompter (Group Score: 182.0 | Individual: 41.1)

Cluster: 6 tweets | Engagement: 209 (Avg: 82) | Type: Tech

best account on X if you actually want to learn enterprise AI:\n\nQT @mardehaym: A B2B lender needed three systems built: ACH funds movement, an underwriting portal, and a syndicator portal. All of it inside a platform that was moving real money the whole time.

We shipped all three with 2 engineers, on one shared architecture, with zero disruption to live operations.

I want to walk through how, because "we used AI and it was faster" doesn't help anyone.

Their codebase had real deals, real disbursements and collections, and real consequences if something broke. A regression here touches someone's money.

This is brownfield, not a weekend prototype demo.

Before anyone wrote feature code, we ran an AI adoption audit on the repo.

It found four gaps: no map of the architecture and conventions, build and test commands the agent couldn't find, modules that each followed different patterns, and recurring tasks solved ad hoc every time.

Then we closed them: context files at repo and module level, recurring procedures encoded once as skills, exact commands documented. Our agents read that before touching anything.

Teams skip this constantly. They point an agent at a repo, hand it a ticket, and get code that compiles but misunderstands the architecture. I've seen it in more brownfield repos than I can count. When someone says "AI doesn't work on our codebase," they almost always skipped the audit.

Every change ran the same loop: spec first, tests written before implementation, then the agent builds inside a deterministic harness.

Static analysis, type checking, and custom validators pass before anything reaches review. A PR agent then reviews every change against the spec. Nothing lands unverified.

Three systems on one foundation came from three things: the spec was the single source of truth for people and agents, tests locked behavior before code existed, and shared components meant each new project launched on a ready foundation instead of starting over.

What shipped: automated ACH replaced manual disbursement and collection and holds as deal volume grows. Underwriting moved into one environment for applicant data, risk rules, and team workflows.

Syndicators see positions, participation, and returns in real time instead of waiting on manual reports.

Zero changes merged without automatic verification. Zero disruption to a platform moving live money.

AI set the pace. The harness held the quality bar.

That's what our AI Velocity Pods were built for.

See 5 related tweets

@alex_verem: this is the best AI transformation team I've come across:\n\nQT @mardehaym: We run an AI Factory at ...
@alex_prompter: the best breakdown of what an enterprise AI factory is:\n\nQT @mardehaym: We run an AI Factory at @L...
@mardehaym: RT @mardehaym: A B2B lender needed three systems built: ACH funds movement, an underwriting portal, ...
@garrytan: RT @mr_kozh: I FOUND 7 GITHUB REPOS THAT TURN ONE AI AGENT INTO AN ENTIRE OPERATING SYSTEM

not 7 ra...

@mardehaym: RT @mardehaym: We run an AI Factory at @LimestoneHQ. Here's what actually happens between a ticket a...

12. amir (Group Score: 175.2 | Individual: 41.0)

Cluster: 8 tweets | Engagement: 194 (Avg: 51) | Type: Tech

There seems to be a misunderstanding about the piece and what OpenAI is up to. There are new techniques at frontier labs that involve loop transformers and similar.

As we say in the piece, OpenAI is putting limits on the loops and trying to make sure CoT is visible with Astra. The concern is that other AI developers may not hold themselves to such limits.\n\nQT @amir: new: OpenAI & others quietly using loop transformers that don't show their 'thinking' when scaled up

a leap forward on performance, but sparking concerns inside & outside OpenAI re: security as this takes off https://t.co/rBVuez1wUK

See 7 related tweets

@scaling01: "OpenAI is putting limits on the loops"

ah yes a good old fake limit. a soft limit.

they will incr...

@bahradx: Just like with rogue AIs, loop transformers (or "recurrent depth" or whatever OpenAI calls it) are "...
@EMostaque: RT @amir: There seems to be a misunderstanding about the piece and what OpenAI is up to. There are n...
@amir: AI researchers are creative. when they encounter a roadblock, they usually find another way thru. Th...
@sjgadler: If this is true, OpenAI seems to be violating one of the few redlines that exist in the AI industry....

13. rohanpaul_ai (Group Score: 174.2 | Individual: 41.8)

Cluster: 8 tweets | Engagement: 136 (Avg: 50) | Type: Tech

A massive win for OpenAI (and for AI training in general) for its legal case against New York Times.

The U.S. govt just formally backed OpenAI’s claim that copyrighted-text training is fair use, partly on national-security grounds.

This is Washington’s first formal intervention in the wider wave of copyright lawsuits over AI training, though the filing is advisory rather than binding on the court.

The U.S. Dept of Justice filed a statement of interest of the US formally arguing in the OpenAI copyright litigation that training LLMs on copyrighted texts should generally qualify as fair use;

That distinction still leaves separate copyright questions around how training data was acquired and whether particular outputs reproduce protected passages. The Justice dept separated acquiring material, training on it, and generating outputs, then focuses its argument specifically on copying at the training stage.

It argues that training serves a different purpose from publishing an article because an LLM uses text to learn statistical relationships and generate new responses.

For market harm, DOJ says training itself does not substitute for the original work, so later AI-generated competition should not automatically make the earlier training unlawful.

The administration also warns that blanket licensing requirements could raise barriers for smaller AI companies and put U.S. developers at a disadvantage against foreign competitors.

The court must still decide fair use case by case, but adopting DOJ’s framework would shift much of the legal pressure from model training toward data acquisition and specific outputs.

See 7 related tweets

@ChrisGPT: The Trump administration just filed a brief backing OpenAI against the NYT and is basically arguing ...
@ednewtonrex: The Trump admin has always sided with big tech on the question of whether AI training is fair use, s...
@wallstengine: U.S. GOVERNMENT BACKS OPENAI IN COPYRIGHT FIGHT

The Trump administration filed a 20-page brief sidi...

@wallstengine: U.S. URGES G20 TO ALLOW AI TRAINING ON COPYRIGHTED WORK

Commerce Secretary Howard Lutnick urged G20...

@WIRED: The US government wrote a letter in support of OpenAI’s argument that training AI on others' intelle...

14. nytimes (Group Score: 165.0 | Individual: 30.2)

Cluster: 9 tweets | Engagement: 399 (Avg: 882) | Type: Tech

Breaking News: A judge ruled that Google must make changes to address its advertising technology monopoly but would not need to break up that business. https://t.co/hwRkiuYV9R

See 8 related tweets

@Techmeme: A US federal judge rules that Google does not have to sell off its ad exchange and instead must make...
@business: Google doesn’t have to sell off its advertising exchange and instead must make its ad tech tools wor...
@Forbes: Google will not have to sell its advertising tech business as part of antitrust litigation against t...
@Forbes: Google Won’t Be Forced To Sell Ad Tech Business Despite Monopoly, Judge Rules https://t.co/6vHkcRi5I...
@BusinessInsider: A judge ruled Google doesn't have to sell off parts of its adtech business, but it will face other r...

15. _NathanCalvin (Group Score: 164.4 | Individual: 38.8)

Cluster: 8 tweets | Engagement: 848 (Avg: 103) | Type: Tech

RT @merettm: I want to prevent a race into unmonitorability kicked off by confused reporting. The depth of the computation graph for our present frontier models, including Astra, is within a factor of two of GPT-4.

OpenAI has worked to preserve and utilize chain-of-thought monitoring since our very first reasoning models. We deeply care about this technique, as it can give us a view into how model alignment generalizes from its training distribution. I do think it is fragile and unfortunately trending in a negative direction, for reasons not contingent on architecture changes that I will write about soon. But there are things we can do to strengthen it, and it's a core goal of our current research program.

See 7 related tweets

@jjohana: Important clarification on AI monitorability: frontier models like Astra are not suddenly orders of ...
@eliebakouch: note that the latest big model by oai with architecture details was already very deep by today's sta...
@polynoamial: Jakub is chief scientist at @OpenAI\n\nQT @merettm: I want to prevent a race into unmonitorability k...
@deanwball: Fully endorse every word of what Jakub (OpenAI’s research director) says here. Exploring legal mecha...
@Miles_Brundage: I'm not in the weeds enough to have a view on how much the 2x GPT-4 thing clarifies/reassures, but g...

16. Polymarket (Group Score: 161.8 | Individual: 21.2)

Cluster: 10 tweets | Engagement: 4050 (Avg: 1885) | Type: Tech

BREAKING: New York City to ban elementary & middle school students from using AI, while also prohibiting teachers from using AI to grade assignments.

See 9 related tweets

@MadisonMills22: RT @cayla_bam: Scoop: NYC public schools' artificial-intelligence policy is coming, and it's expecte...
@altryne: Oh hey, you know this technology that's absolutely going to be a huge part of the kids lives, advanc...
@kylebrussell: Going to be a massive skill gap between average kids and those whose parents use AI a ton and teach ...
@unusual_whales: BREAKING: New York City Public Schools is banning more than half a million students and teachers fro...
@business: The largest school district in the US is banning AI tools and restricting screen time for elementary...

17. Alibaba_Qwen (Group Score: 161.7 | Individual: 35.2)

Cluster: 6 tweets | Engagement: 939 (Avg: 1214) | Type: Tech

🏆 #1 on CodeArena: WebDev leaderboard.

Qwen3.8-Max-0902 jumps from 1669 to 1691, setting a new record for agentic coding (WebDev) workflows, with standout strength in multistep reasoning, tool use, and full app generation.

Thanks for the recognition! @arena\n\nQT @arena: Big news: Qwen3.8-Max-0902 by @Alibaba_Qwen just debuted at #1 overall in the Code Arena: WebDev with 1691 pts!

It scores 3 pts above Claude Opus 5 (Max), 17 pts above Kimi K3 (Max), and 22 pts above the previous Qwen3.8-Max.

Priced at a blended $5/MToken, Qwen3.8-Max-0902 also claims the highest-scoring position on the Pareto frontier! Stay tuned for a closer look at its Pareto positioning, and for Agent Arena scores coming soon.

Its strength carries across every Code Arena: WebDev category:

#1 in Data & Analytics and Consumer Product
#2 in Brand & Marketing, Gaming, and Simulations
#3 in Content Creation Tools and Reference-Based Design

Congrats to the @Alibaba_Qwen team on this huge update!

See 5 related tweets

@Alibaba_Qwen: 🏆 #1 overall on Code Arena, and top of the Pareto frontier at $5/MToken. Thanks! @arena Give Qwen3....
@ivanfioravanti: What??? Qwen3.8-Max-0902 1st on WebDev 👀 Every day a new surprise in the AI world!\n\nQT @arena: Big...
@Hesamation: bro you literally can't sleep or a new model drops. > Qwen 3.8 Max 0902 > #1 model on WebDev A...
@arena: RT @Alibaba_Qwen: 🏆 #1 on CodeArena: WebDev leaderboard.

Qwen3.8-Max-0902 jumps from 1669 to 1691, ...

@markgadala: Qwen becoming the leader in open source and then beating frontier models is why China is winning the...

18. OfficialINDIAai (Group Score: 154.3 | Individual: 32.2)

Cluster: 6 tweets | Engagement: 6 (Avg: 18) | Type: Tech

At the #G20InnovationMinisterial in Chapel Hill, North Carolina, Shri Jitin Prasada, Hon’ble Minister of State for Electronics & IT, and Commerce & Industry, shared India’s perspective across three key pillars:

Pro-Innovation Policy Frameworks
Technology for Opportunity & Prosperity
Skilled Technical Workforce Development

India emphasised the need for enabling policies, technological advancement and a future-ready skilled workforce to turn innovation into opportunities and drive inclusive growth and shared prosperity.

#G20 #Innovation #Technology #Skilling #India

@AshwiniVaishnaw @SecretaryMEITY @sudeep_tweets @_DigitalIndia @GoI_MeitY @mygovindia @g20org @IndianEmbassyUS\n\nQT @JitinPrasada: Participated in the G20 Innovation Ministerial Session on Pillar I - Pro-Innovation Policy Frameworks; Pillar II - Technology for Opportunity & Prosperity; and Pillar III: Skilled Technical Workforce Development in Chapel Hill, North Carolina.

I shared India’s perspective on the importance of enabling policy frameworks that encourage innovation, support technological advancement and create an environment where new ideas can translate into opportunities through skilled workforce development.

India remains committed to fostering an ecosystem where innovation can drive inclusive growth and shared prosperity. @GoI_MeitY @g20org @IndianEmbassyUS

See 5 related tweets

@JitinPrasada: Participated in the G20 Innovation Ministerial Session on Pillar I - Pro-Innovation Policy Framework...
@_DigitalIndia: Shri Jitin Prasada attended the G20 Innovation Ministerial Session in Chapel Hill, North Carolina, c...
@OfficialINDIAai: India–Africa partnership is rooted in shared aspirations for technology, innovation and inclusive gr...
@OfficialINDIAai: The future of innovation needs a global conversation. And India is getting ready to lead from the f...
@goi_meity: Shri Jitin Prasada attended the G20 Innovation Ministerial Session in Chapel Hill, North Carolina, c...

19. Marktechpost (Group Score: 146.2 | Individual: 27.5)

Cluster: 7 tweets | Engagement: 18 (Avg: 116) | Type: Tech

Meta Superintelligence Labs Releases Muse Voice Transcribe: One Real-Time Model for Streaming ASR, Diarization, and Endpointing

Most production voice stacks are three systems stitched together. One model transcribes, a second separates speakers, and a detector decides when the user stopped talking. Each hand-off adds latency and a new failure mode.

Muse Voice Transcribe, announced by Meta Superintelligence Labs this week, collapses those three jobs into a single autoregressive model. Meta calls it its first real-time audio perception model. It performs streaming ASR, speaker diarization for 20+ speakers, and endpointing in one pass, with no required post-processing....

Full analysis: https://t.co/SSRRbk9lO2

Technical details: https://t.co/T3Zy0PuudM

@AIatMeta @Meta_Engineers #artificialintelligence #ai #speechai #voiceai #data #Artificial\n\nQT @AIatMeta: Introducing Muse Voice Transcribe, the first real-time audio perception model from Meta Superintelligence Labs.

Muse Voice Transcribe delivers real-time streaming ASR, diarization with 20+ speakers, and endpointing. It’s multilingual with seamless code-switching and improves accuracy with language, keyword, and context biasing.

The model ranks first on @ArtificialAnlys streaming speech-to-text and on public diarization benchmarks.

See 6 related tweets

@rohanpaul_ai: RT @rohanpaul_ai: Love this, another huge release from Meta.

Lunched Muse Voice Transcribe for rea...

@WesRoth: Meta has introduced Muse Voice Transcribe, MSL’s first real-time audio perception model, rolling out...
@diptanu: Back when I was working in speech recognition at Facebook, we trained separate models for each of th...
@cryptopunk7213: wispr flow is arguably one of the most successful ai products to date, completely changes the way yo...
@alexandr_wang: RT @spencerbarnett: Introducing Muse Voice Transcribe! 🗣️

This is our first real-time audio percept...

20. chris_j_paxton (Group Score: 144.2 | Individual: 36.5)

Cluster: 4 tweets | Engagement: 33 (Avg: 76) | Type: Tech

Seems like a great tool for high quality, scalable dexterous data collection https://t.co/5aZ5FIkkAu\n\nQT @XSquareRobot: What if a dexterous robot could learn a 20+ step chemistry experiment without a single on-robot training demo?

Meet TwinDEX: a pair of co-designed, three-finger, nine-DoF dexterous manipulation interface: one wearable for data collection, one for robot deployment.

The twinned design shares identical kinematics, contact surfaces, visual appearance, and sensors across collection and deployment — keeping observations and actions aligned end to end.

Trained from scratch on only a few hundred wearable demonstrations - with zero on-robot training or intervention data - TwinDEX completed a standardized chemistry experiment involving tool switches, fine force control, and bimanual coordination.

Robot-free data showed comparable learning efficiency on the multi-task evaluation, TwinDEX delivered 5.3 times effective throughput than on-robot teleoperation.

TwinDEX demonstrates that high-quality robot-free data can fully substitute for on-robot teleoperation data on challenging dexterous tasks — removing the dependency on real-robot hardware that has been the central bottleneck to scaling dexterous manipulation data.

This was the proof-it phase. Now comes scale: what emerges at tens of thousands, or millions, of episodes?

Watch the demo and read the technical blog: https://t.co/PgyFkrJdXG

#TwinDEX #Robotics #EmbodiedAI #DexterousManipulation

See 3 related tweets

@rohanpaul_ai: Dexterous robots have a data problem that gets expensive very quickly, and recently, so many brillia...
@chris_j_paxton: Seeing that completing long horizon tasks is a data problem -- that with good tools like these co-d...
@FellMentKE: One of the biggest bottlenecks in dexterous robot learning is data collection.

On-robot teleoperati...