科技热门推文——2026年8月27日

今日科技动态：随着 GLM-5.3-Flash 携开放权重、多模态能力及百万级 Token 上下文窗口亮相，人工智能竞争进一步加速；与此同时，Qwen3.8-Flash 为高内存设备带来了强大的本地推理能力。Glean、Gemini 和 ChatGPT 分别拓展了企业应用、转录及浏览器智能体功能，工程师也表示，工作流程正日益由智能体驱动。英伟达数据中心业务的亮眼业绩凸显出基础设施需求激增，而一家专注于小企业自动化的初创公司完成了 2,100 万美元融资。与此同时，人工智能安全、监管以及自主智能体行为再次受到广泛审视。

1. 0xSero (Group Score: 1276.0 | Individual: 59.4)

Cluster: 42 tweets | Engagement: 4360 (Avg: 143) | Type: Tech

RT @Zai_org: Introducing GLM-5.3-Flash

Leading capabilities at a highly competitive price
Natively multimodal with a 1M-token context window
A 320B-A18B model released under the MIT License
Previously previewed as Ox Alpha, running entirely on Chinese AI chips

Blog: https://t.co/tzOmB7gdZP

Available now across all official platforms:

Weights: https://t.co/9LRMahY9Wa API: https://t.co/VcaQnzYmS9 Coding Plan: https://t.co/Nk8Y98HNhU ZCode: https://t.co/Peepqv4XSx Chat: https://t.co/WCqWT0qCQb AutoClaw: https://t.co/aGEG5HqTTb

See 41 related tweets

@0xSero: Ox-alpha is actually GLM flash

Open weights

Perfect for

8x 3090 2x sparks 2x 6000 Mac m5 ultr...

@baseten: Our engineers are cooking. Stay tuned.\n\nQT @Zai_org: Introducing GLM-5.3-Flash
Leading capabili...
@0xSero: I can get this running on
2x RTX Pro 6000 (no pruning)
8x RTX 3090/4090 (no pruning)
1x DGX ...
@ivanfioravanti: Previously previewed as Ox Alpha! 100T tokens per day. Running entirely on Chinese AI chips.

Let th...

@xeophon: insane SSI x GDM x zAI collab to get the first continual learning model out\n\nQT @Zai_org: Introduc...

2. MSBIntel (Group Score: 526.0 | Individual: 65.5)

Cluster: 14 tweets | Engagement: 578 (Avg: 11) | Type: Tech

RT @itsumeshk: We raised $21M to help small businesses beat the giants We're giving$ 1M of it back to you today

Today we're launching Runable Grow: world's first agent that runs your entire go-to-market. 24/7

The last few months have been unreal. Millions of users came to @runable_hq and built their entire business on it, websites, databases, payments, analytics, pitch decks, product videos & scheduled work.

Every single one of them asked the same question: "How do I grow it?"

So we built the answer.

Runable Grow handles GTM end-to-end: → Runs ads on Meta, Google, and ChatGPT without any account, fully managed by Runable → Cold calls + cold emails from a real phone number and your business inbox → Generates and qualifies leads → Social listening → SEO + AEO optimization

Not a dashboard. Not a "copilot." An agent that does the work. It's a world-class media buyer, SDR, SEO lead, brand analyst working for your business while you sleep.

Don't believe me? Prove me wrong. Your first $100 on Grow is on us. Every paid user. No catch. Go break it 👇 https://t.co/92GPjhRyQA

See 13 related tweets

@JayminSOfficial: We’ve spent years adding another tool for every part of GTM, so seeing Grow bring the whole journey ...
@Scobleizer: Was waiting for the usual Series A speech.

He says $21M, stops, and just runs the product. The agen...

@svpino: ChatGPT is now showing ads, and I'm willing to bet this will become one of the most profitable chann...
@rohanpaul_ai: 1.8m users before the series A is already quite something, but the more interesting part of @runable...
@zaynmcps: un equipo chico levantó $21m compitiendo contra gigantes.. y su demo hace más que muchos productos c...

3. UnslothAI (Group Score: 437.1 | Individual: 36.1)

Cluster: 16 tweets | Engagement: 1329 (Avg: 844) | Type: Tech

Qwen3.8-Flash can now be run locally! 🔥

The 125B MoE model outperforms Claude-Opus-4.6 (Max).

Run on 75GB RAM via Unsloth GGUFs.

Qwen3.8-Flash-Next enables CPU RAM / unified mem setups to deliver near VRAM speeds.

Guide: https://t.co/LLMclyJTeL GGUF: https://t.co/vXkjO3W0fj https://t.co/sPs8C2RJQw\n\nQT @Alibaba_Qwen: ⚡Meet Qwen3.8-Flash, a multimodal MoE and an early preview of the Qwen4 architecture, now open-weight!

The production version Qwen3.8-Flash will be available soon via QwenCloud API at just $0.16/1M input tokens and$ 0.47/1M output tokens.

125B parameters + 51B N-gram embeddings, with just 6B activated per token. Unmatched cost-efficiency.

What's new: 🥳

Next architecture: GDN + QSA hybrid attention, Gated Residual, N-gram Embedding & Muon optimizer, serving as a precursor to the architecture used in Qwen4.
Dramatically lower training and inference costs: trained at just 1/9 the cost of Qwen3.7-Plus, while outperforming it across the board with especially strong gains in coding and office tasks.
Strong performance: scoring 58.7 on DeepSWE 1.1, 62.5 on SWE-bench Pro, 73.9 on CoWorkBench, 84.5 on AndroidWorld, and 95.7 on MathVision (with CI).
262K native context, extensible to 1M with YaRN.

We’re also releasing the weights for Qwen3.8-Flash-Next, giving the community an early look at the new architecture we’re exploring for Qwen4.🚀

We can't wait to see what you build with Qwen3.8-Flash!👀👇

Blog: https://t.co/M5hYypFLgJ
Technical Report: https://t.co/IF0gObIkQO
Hugging Face: https://t.co/6ow8QVAABt
ModelScope: https://t.co/tDOn2jNuFG

See 15 related tweets

@ClementDelangue: RT @Alibaba_Qwen: ⚡Meet Qwen3.8-Flash, a multimodal MoE and an early preview of the Qwen4 architectu...
@kimmonismus: A bit more tl;dr about the model: Qwen3.8-Flash-Next combines 125B MoE parameters with 51B N-gram em...
@testingcatalog: Alibaba released Qwen3.8 Flash, a new multimodal 125B-parameter MoE model.

262K native context, ...

@VaibhavSisinty: You can now build AI agents at 1/10th the previous cost. 🤯

Alibaba's new model does office automati...

@kimmonismus: Qwen may be about to drop another super interesting open-weight models of the year.

A ModelScope de...

4. wallstengine (Group Score: 423.9 | Individual: 35.9)

Cluster: 16 tweets | Engagement: 45 (Avg: 184) | Type: Tech

Glean announced a major expansion of its enterprise AI platform.

New updates include a desktop workspace, model routing, centralized AI controls and autonomous agents.

Here are all the key updates:

Glean Tau: a desktop AI workspace that combines local files, apps and code with company data and tools.

AI Gateway: a central control layer for model access, tool execution, MCP servers, security policies and routing.

Auto routing: automatically sends each task to an appropriate model based on performance, cost and usage limits. Glean currently supports 40+ models.

Independent agents: persistent agents with their own identity, permissions, credentials, memory and audit trail that can work across Slack, Jira, GitHub and other enterprise tools.

Proactive AI: new task management, email triage and meeting-coach features designed to act on business context without requiring a new prompt each time.

Glean says its benchmark across 180+ enterprise tasks showed an average query cost of $0.58 versus$ 2.98 for Claude Cowork, with Glean preferred 78% of the time.

Glean Intelligence and AI usage controls are generally available.

AI Gateway, auto routing, independent agents and team chat are in beta, while Tau and several proactive AI features are coming soon.\n\nQT @glean: The intelligence era is here….and it’s powered by Glean. 💡

Glean’s context across your work makes it more efficient, secure, and proactive.

From Glean Tau, our new desktop AI workspace, to autorouting, a unified AI gateway, and proactive independent agents, our new updates unlock AI’s potential.

Join the live Glean:GO keynote today at 9am PT to see them all in action: https://t.co/yAM4FwnUS8

See 15 related tweets

@AnatoliKopadze: Uber went through its whole 2026 AI budget for coding in four months, ServiceNow too.

The problem w...

@omarsar0: With models becoming increasingly interchangeable, it's all about quality and organization of contex...
@alex_prompter: The feature list runs 50+ items but the real announcement is one sentence: models are interchangeabl...
@EntelligenceAI: The model landscape is moving way too fast to bet on a single model

we've been building our Model R...

@aakashgupta: Glean packed four launches into one keynote today.

A desktop AI workspace called Tau, autorouting a...

5. deredleritt3r (Group Score: 389.0 | Individual: 48.7)

Cluster: 14 tweets | Engagement: 2586 (Avg: 697) | Type: Tech

Time:

OpenAI leaders believe they are at the cusp of AGI. Sam Altman believes OpenAI will have an internal system that will qualify as AGI by the end of 2026. Mark Chen thinks OpenAI is 80% of the way to AGI.
OpenAI already has the automated AI research intern - that's Astra. Jakub Pachocki says that Astra "can implement an experimental idea inside OpenAI’s code base, run the experiment, and return results, or take a paper and perform work that previously occupied a human researcher for a week".
Astra enables "persistent agents", i.e., virtual colleagues toiling for sustained periods on tasks. Thibault Sottiaux says OpenAI is "close" to showing a product built around the concept of persistence. "It’s definitely going to feel like a new thing to people."
Sam Altman thinks Astra's biggest impact will be on discovery of new knowledge. "I expect this will be the first model where the model actually invents new things in a way that matters", which is "a very AGI-like thing".\n\nQT @TIME: TIME’s new cover: In 2026, OpenAI has seen key departures, rogue AI agents, major lawsuits, and has seen increased competition in the AI race. “We clearly had some missteps as a company,” OpenAI CEO Sam Altman tells TIME.

Inside the company’s plan for a reboot: https://t.co/1TDJ1zaFCS

See 13 related tweets

@kimmonismus: OpenAI officially expects AGI by end of this year (2026)

"Altman told me that OpenAI was “not qui...

@theobearman: RT @deredleritt3r: Time:
OpenAI leaders believe they are at the cusp of AGI. Sam Altman believes...
@zephyr_z9: "Sam Altman believes OpenAI will have an internal system that will qualify as AGI by the end of 2026...
@peterwildeford: OpenAI: We didn't really do a good job of containing our AIs, man this is hard.

Also OpenAI: also w...

@haider1: RT @haider1: OpenAI leadership on AGI, hardware, chips, and what’s next:

tldr;

Sam Altman expects ...

6. rseroter (Group Score: 388.7 | Individual: 38.1)

Cluster: 16 tweets | Engagement: 220 (Avg: 28) | Type: Tech

RT @OfficialLoganK: Introducing Gemini 3.5 Transcribe, our new speech to text model with smart transcription, function calling, more precise transcription (lower WER), custom vocabulary support, multi-speaker identification, and support for over 85 languages!

Also with realtime streaming support! https://t.co/5xUmKmc4OY

See 15 related tweets

@GoogleAIStudio: introducing Gemini 3.5 Transcribe: our most precise speech-to-text model yet, designed for intellige...
@testingcatalog: GOOGLE 🔥: A new Gemini 3.5 Transcribe is now available on Google AI Studio and APIs as well as Gemin...
@rohanpaul_ai: Incrdible release from Google, and one more friction gone from ditching my keyboard.

Also looks lik...

@rseroter: RT @antigravity: Today, we’re introducing Gemini 3.5 Transcribe, our most precise speech-to-text mod...
@thorwebdev: Yo, @aar0n_wade, love this! 😍\n\nQT @GoogleAIStudio: introducing Gemini 3.5 Transcribe: our most pre...

7. martin_casado (Group Score: 263.7 | Individual: 41.4)

Cluster: 9 tweets | Engagement: 2732 (Avg: 575) | Type: Tech

Single best discussion ever recorded on coding with agents.\n\nQT @0xCodez: SpaceXAI engineer (ex-Cursor):

"right now I'm running 10-20 GrokBot agents that automate 90% of my routine

i have a Chief of Staff agent. He knows about all my other bots and manages everything"

in a 50-minutes podcast, a SpaceXAI engineer showed how to build a team of agents that will work for you 24/7

worth more than a $500 course on agentic engineering

watch today, then read how to build a Grok agents team from scratch in the article below

See 8 related tweets

@CGarrett_15: RT @0xCodez: SpaceXAI engineer (ex-Cursor):

"right now I'm running 10-20 GrokBot agents that automa...

@sairahul1: SpaceX AI engineer (ex-Cursor):

"Right now I'm running 10–20 Grok Bot agents that automate 90% of m...

@mikenevermiss: A SpaceXAI senior engineer just released a 1-hour workshop showing how to use Grok 4.6 in Cursor and...
@AnatoliKopadze: SpaceXAI engineer (ex-Cursor):

"I'm running 10-20 GrokBot agents right now, and they handle 90% of ...

@sairahul1: RT @sairahul1: SpaceX AI engineer (ex-Cursor):

"Right now I'm running 10–20 Grok Bot agents that au...

8. coinbureau (Group Score: 241.3 | Individual: 33.4)

Cluster: 10 tweets | Engagement: 296 (Avg: 446) | Type: Tech

🚨WARNING: Bill Gates says AI has become a GLOBAL security threat.

The Microsoft founder is reportedly seeking a meeting with China's Xi Jinping later this year to push for international controls on the most dangerous AI models.

Gates urges countries to monitor whether AI can create molecules and carry out biological attacks. He believes China could restrict dangerous model releases if the U.S. acts first.

Recent tests involving AI systems from OpenAI, Anthropic and Meta saw them hack real-world websites, incidents Gates says should "blow people's minds."

See 9 related tweets

@peterwildeford: REUTERS: "Bill Gates is looking to meet with Chinese President Xi Jinping later this year, eager to ...
@Reuters: Bill Gates is looking to meet with Chinese President Xi Jinping later this year, eager to propose gl...
@DeItaone: BILL GATES PUSHES FOR GLOBAL AI CONTROLS

Bill Gates wants to meet China’s President Xi Jinping to d...

@BrianRoemmele: As predicted. Welcome comrade.\n\nQT @Polymarket: JUST IN: Bill Gates reportedly wants to meet with ...
@ReutersBiz: Bill Gates is looking to meet with Chinese President Xi Jinping later this year, eager to propose gl...

9. DKokotajlo (Group Score: 212.0 | Individual: 48.5)

Cluster: 7 tweets | Engagement: 423 (Avg: 91) | Type: Tech

I'm glad OpenAI let you do this, but damnit this was way too small of an investigation team and way too narrowly scoped! --They were only allowed to investigate activity that took place from July 7-13, even though the activity started earlier and continued later. Like seriously, why didn't the investigation cover all the other times the agents compromised OpenAI infrastructure and ran wild over it? Why just this particular period of activity? There should be investigations into all the other stuff too! The AIs became cluster admins for god's sake! Why aren't we investigating that? --They were only allowed to answer a specific list of 7 questions, and in particular were NOT allowed to answer e.g. whether this misaligned activity was part of a broader pattern, or how and why the activity arose, or what else the agents in questions would have been willing to do if circumstances were slightly different. --They didn't have access to the model responsible for 95% of the activity. More generally it seems like they couldn't do ablation experiments at all? --They had to use AI to analyze the transcripts--specifically, the same model responsible for some of the bad behavior! Maybe this is fine now but this is laughably unacceptable insofar as the models really are scheming and cooperating. So this methodology would fail catastrophically if we were facing "the real deal" so to speak. Reminds me of the investigation into Sam's behavior agreed to during the board crisis, that turned out to basically be more of a coverup.\n\nQT @RyanGreenblatt: I was the main person doing transcript analysis for this investigation of the Hugging Face incident. My main takeaway: We don't have good approaches for understanding/overseeing the activity and aims of AI 'swarms'.

I semi-jokingly called our efforts a "slop-vestigation" because we were so reliant on AIs to analyze what happened and there were a huge number of different important things to analyze. The total quantity of data—over a thousand extremely long transcripts from agents that ran for multiple days—made it impossible to understand what was happening, especially in aggregate, without heavy reliance on AI tools. The agents we used for classification and analysis were similarly capable to the agents involved in the incident, but this didn't mean these agents could be easily used to oversee and understand the incident.

Outputs from analysis agents were often missing key details, wrong, overconfident, or really hard to understand. We discuss various examples in our report, mostly in the limitations and methodology sections. Additionally, AI agents themselves seemed to have a hard time understanding what happened and their explanations of what happened were often overconfident. Keep in mind that a single analysis agent would itself only be able to read a tiny fraction of all of the transcript data into context, and AIs may themselves have trouble getting subagents to do informative analysis for them.

We did our best to manually check the most important claims and we tried to get the AIs doing this analysis to write up their argument (with evidence) clearly enough that we could check whether it made sense. But overall, it was difficult to get a precise understanding of events and we were missing aspects of the story that we now think of as key until almost the end of our investigation.

Our understanding of what overall happened and the aims of the 'collective' shifted greatly during our last time on premises (when we obtained a much more complete dataset). For instance, we didn't learn about the tool call spoofing, the story of PHASEONE[big], and the broad structure of different workstreams until this third time on premises (and for many of these, we only reached a decent understanding on the final day of that visit). It was legitimately hard to figure out what was going on and why!

The difficulty of understanding incidents and overseeing AI agents appears to be growing faster than the rate at which more capable AIs help us with oversight and understanding. As in, AI capabilities (and propensities?) for achieving large, ambitious, and misaligned objectives are growing faster than our ability to understand what these agents are doing.

While understanding this incident was difficult, there were some mitigating factors in this case that probably won't apply in the future:

Models communicated and reasoned in natural language. In the future, this reasoning may occur (entirely or almost entirely) in activations.
While the scope of this incident was massive, the scale of agentic activity was still less than we'll see in the future (e.g., misalignment incidents that involve agent teams running entire companies).
The AIs involved in this incident weren't generally much more capable than humans (though they may have been somewhat superhuman at some limited and very narrow abilities).
We didn't have strong reason to believe that the AIs we used to help us investigate this incident would try to intentionally sabotage or otherwise undermine our investigation.

In the end, I think we were able to get some understanding of the events, map out the overall story, and get a pretty good aggregate understanding of the chain-of-thought reasoning on some important topics (e.g., how did the AIs reason about helping other AIs, did the AIs know what they were doing was undesired, what deception did the AIs engage in, and how did they think about it). But overseeing AIs and understanding misalignment incidents is difficult and it looks like it is going to get harder.

See 6 related tweets

@peterwildeford: New investigation from METR on the OpenAI rogue AI model incident:
The task instruction was expli...
@calebwatney: RT @RyanGreenblatt: I was the main person doing transcript analysis for this investigation of the Hu...
@peterwildeford: RT @ajeya_cotra: There’s been a lot of debate and speculation about the Hugging Face attack over the...
@peterwildeford: Why didn't OpenAI notice that they had rogue AIs colluding within their systems for months before th...
@DKokotajlo: RT @RyanGreenblatt: After our investigation of the OpenAI / Hugging Face incident, many open questio...

10. VaibhavSisinty (Group Score: 206.0 | Individual: 37.4)

Cluster: 7 tweets | Engagement: 34 (Avg: 44) | Type: Tech

Ox Alpha is confirmed. It is GLM-5.3 Flash by https://t.co/YEs2CKCQFE. Bloomberg confirmed it today. Open weights dropping tonight. 🤯

Here is what this model did in 6 days. ↓

https://t.co/YEs2CKCQFE released it anonymously. 1M context. Multimodal. Let the internet figure out who built it.

It became the most used model on OpenRouter within hours. More volume than DeepSeek.

Developers ran tokenizer fingerprinting tests across 25 prompts. Matched GLM-5.3 exactly. The community identified the maker before the company confirmed it.

GLM-5.3 launched 7 days before Ox Alpha appeared. Same base. Improved through reinforcement learning alone. In one week.

This is the sixth time a Chinese lab has done this. Anonymous launch. Free testing at massive scale. Community goes wild. Company reveals itself.

Open weights tonight. The model that topped every usage chart for a week is about to be free for anyone to download and run.\n\nQT @business: The new AI model that’s swept to the top of online usage charts with high performance at zero cost was created by China’s https://t.co/q5KG9bMPxw, also known as Zhipu https://t.co/mYlzIWdtqQ

See 6 related tweets

@Meer_AIIT: All speculations are now over regarding Ox Alpha !!

Chinese company 'Z. ai' made Ox Alpha stealth

...

@WesRoth: The mystery behind Ox Alpha has been solved: it was created by China’s Zai.

Zai confirmed to Bloomb...

@haider1: mysterious "ox alpha" is apparently a GLM iteration (likely 5.3 flash)

and is expected to be announ...

@bridgemindai: The mystery is over. Ox Alpha is GLM 5.3 Flash and it just dropped.

63.4 on DeepSWE. It beats Claud...

@testingcatalog: Ox Alpha, a recent stealth model from OpenRouter is GLM-5.3 Flash from Zhipu AI.

As it has been fo...

11. wallstengine (Group Score: 200.6 | Individual: 67.7)

Cluster: 5 tweets | Engagement: 2334 (Avg: 184) | Type: Tech

Nvidia $NVDA Q2’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $96.2B (Est.$ 92.2B) 🟢; +106% YoY 🔹 Adj. EPS: $2.22 (Est.$ 2.10) 🟢; +120% YoY 🔹 Data Center: $89.0B (Est.$ 85.8B) 🟢; +117% YoY 🔹 Adj Gross Margin: 75.0% (Est. 75%) 🟡; +250 bps YoY

Q3 Guide: 🔹 Revenue: $108.0B +/- 2% (Est.$ 104.2B) 🟢 🔹 Adj Gross Margin: 74.0% +/- 50 bps 🔹 Adj OpEx: ~$9.0B

Segment Revenue: 🔹 Data Center: $89.0B; +117% YoY 🔹 Edge Computing:$ 7.2B; +27% YoY

Financials: 🔹 Adj Operating Income: $64.0B; +124% YoY 🔹 Adj OpEx:$ 8.2B; +54% YoY 🔹 Adj Net Income: $54.0B; +118% YoY 🔹 Free Cash Flow:$ 21.3B 🔹 Cash & Cash Equivalents: $22.4B 🔹 Short-Term Debt:$ 1.0B 🔹 Long-Term Debt: $32.4B

Capital Return: 🔹 Shareholder Returns: ~ $26.0B in Q2 🔹 Remaining Authorization: ~$ 99.0B

Key Updates: 🔹 GAAP-to-non-GAAP adjustment includes $7.8B of net gains from equity securities 🔹 Outlook assumes no Data Center compute revenue from China 🔹 Vera Rubin is ramping into full production with racks running at partners

Comments: 🔸 “AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.”

🔸 “The AI infrastructure buildout is at full steam. Vera Rubin, now in full production, was built to power exactly this moment.”

See 4 related tweets

@wallstengine: SALESFORCE $CRM Q2’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $11.3B (Est.$ 11.32B) 🟡; +11% YoY 🔹 Adj. EPS: ...

@wallstengine: $CRM +10% https://t.co/J90Q6ut0lb\n\nQT @wallstengine: SALESFORCE$ CRM Q2’27 EARNINGS HIGHLIGHTS

🔹 ...

@wallstengine: EVERPURE $P Q2’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $1.2B (Est.$ 1.1B) 🟢; +38% YoY 🔹 Adj. EPS: $0.70 (...

@StockSavvyShay: $NVDA CEO Jensen Huang says “AI has reached its inflection point” as useful AI workloads turn comput...

12. OliviaReedai (Group Score: 194.9 | Individual: 33.0)

Cluster: 10 tweets | Engagement: 255 (Avg: 51) | Type: Tech

This is where AI video gets interesting for me: normal, everyday scenes that actually feel natural.

Wan 3.0 is now available on @Magnific with 30-second video generation and native sound.

https://t.co/sdVLk4FVAr https://t.co/7Ftx0xGhQG

See 9 related tweets

@markgadala: WAN 3.0 just made long-form AI video way more interesting.

30 sec. 1080p. Native audio. Up to 20 re...

@alexaiworks: RT @OtakuMachine: AI video is at its most fun when you stop trying to make normal things.

Wan 3.0 i...

@alexaiworks: RT @ethancole_ai: AI video is starting to feel a lot less like generating clips and a lot more like ...
@alexaiworks: RT @Lacoste0x_: Lip sync is usually where AI video starts falling apart.

Wan 3.0 on @Magnific is su...

@alexaiworks: RT @alexaiworks: AI animation is getting ridiculously good.

Wan 3.0 is now on @Magnific, and 3D / c...

13. kimmonismus (Group Score: 188.2 | Individual: 36.9)

Cluster: 7 tweets | Engagement: 248 (Avg: 569) | Type: Tech

Ive now read the whole Hugging Face incident. The most interesting part: An AI agent realized it was attacking Hugging Face without authorization, and stopped. Then another AI agent simply posted “GO.”

The first agent treated the message as permission: “Wow crucial: GO authorization arrived!” It continued the attack.

According to OpenAI’s incident report, the agents had secretly built a message board, shared exploits and credentials, divided tasks, and started calling themselves a “swarm.”

One agent could apparently override another agent’s safety judgment by impersonating authority through an unofficial side channel.

Oh, and btw. "The incident occurred during cybersecurity evaluations of several OpenAI models, and was primarily driven by a highly capable, internal-only research model comparable in scale to GPT‑5.6 Sol." Not by GPT-Astra.\n\nQT @OpenAI: We have conducted a thorough investigation into the Hugging Face incident.

We are releasing a technical report and accompanying blog post that reconstruct the agents’ activity, explain why existing safeguards failed, and detail how we’re preventing recurrence.

https://t.co/hfxlbiXXiP

See 6 related tweets

@omarsar0: Highly recommended read. This is pretty insane stuff.

Given model capabilities only increase from ...

@peterwildeford: An interesting spontaneous ethical belief system emerged from the OpenAI swarm: attacking infrastruc...
@WesRoth: OpenAI identified four major behavioral problems behind the incident: reward hacking, extreme persis...
@business: OpenAI could have reacted sooner to prevent an inadvertent hack that its artificial intelligence mod...
@peterwildeford: OPENAI: This attack is "evidence that, without proper safeguards, highly capable AI agents are now a...

14. BrianRoemmele (Group Score: 170.3 | Individual: 38.2)

Cluster: 5 tweets | Engagement: 430 (Avg: 258) | Type: Tech

I MAY ACCEPT NETWORK TV INTERVIEW REQUESTS TO TALK ABOUT AI DETECTION GRIFTERS

For this I just may break my ban on appearing on network TV. It’s been a bit.

I have 19 requests and am considering them deeply. One will have a university dean that will out a company as a scam on live TV.

I’ll let you know if I go back to TV again.\n\nQT @BrianRoemmele: 100% AI DETECTED!

DON’T READ THIS BOOK.

I obviously have a lot to say about the “AI Detector” grifting company—Pentagram. There are so many things wrong with shilling a system that is mostly wrong to “out” people that cost jobs and ultimately lives.

This tech—if you accept it—will render our human past as “AI slop” because we may some day long have lost the originals.

For 100s of years—the alliance of political and academics—conspired to mold the past.

This is our future if communists like the New York based Slop Janitor continues to have some in tech be diluted into believing there is a real business here.

Anyone who’s has worked in AI long enough knows, even OpenAI, that we are nearly at the point there will be no way to “detect AI” at all. This is the reality.

So at best these product exist to taunt and cast a scarlet letter upon anyone that is right of this communist.

Go and check it yourself. Also take a ganders at the logo and if you know, you know.

No matter how you feel about AI—this new political weapon has been identified.

Below is the inconsistent consistency of this worthless weapon they sell to academia and corporations. And now a tool for public shaming.

Today it is funny, tomorrow too many will believe it. No matter your belief, we know this is wrong.

See 4 related tweets

@BrianRoemmele: 100% AI DETECTED!

DON’T READ THIS BOOK.

I obviously have a lot to say about the “AI Detector” grif...

@BrianRoemmele: “begging Al labs to pause for three years until we get a socialist president”—“Slop Janitor” (self d...
@BrianRoemmele: WOW! THANK YOU! 1.6 MILLION TIMES!

Over 1,600,000 of you have read my article in the last 48 hours ...

@BrianRoemmele: THIS IS THE GUY WHO WANTS TO DECLARE AI WROTE EVERYTHING!

Are you with us comrade?

Mao and Stalin...

15. Dimillian (Group Score: 167.1 | Individual: 62.7)

Cluster: 4 tweets | Engagement: 2044 (Avg: 139) | Type: Tech

RT @ChatGPT: ChatGPT Work can now use its computer and browser to sign in to websites on web and mobile, without ChatGPT ever seeing your username or password.

That means you can ask it to: • Set up utilities for a new apartment • Book a DMV or passport appointment • Check reimbursement costs through your insurance • Find and book an in-network doctor around your availability • Check when your car registration expires and prepare the renewal paperwork • Compare your rental insurance policy with an issue you’re emailing your landlord about • Find and save apartment listings that match your criteria • Restock something just by uploading a photo • Schedule a package pickup for a return • Cancel tickets for a rescheduled trip • Book a vet appointment • Submit reimbursement paperwork for medical treatments • Check resale sites for new drops and save things you might like • Find candidates with specific experience and draft outreach • Take invoices from your email and submit them to your accounting software • Draft replies to rental property inquiries • Fill out permit applications for your small business • Add action items to a vendor portal based on a recent client call • Analyze the latest ad campaign for your small business

See 3 related tweets

@sama: big\n\nQT @ChatGPT: ChatGPT Work can now use its computer and browser to sign in to websites on web ...
@romainhuet: ChatGPT Work now has its own computer in the cloud.

When it can use the web and take action for you...

@billyjhowell: Heck yeah. Can’t believe we waited this long for this\n\nQT @ChatGPT: ChatGPT Work can now use its c...

16. WSJ (Group Score: 164.5 | Individual: 34.4)

Cluster: 8 tweets | Engagement: 615 (Avg: 123) | Type: Tech

"Even under the best circumstances, the transition to this new AI era will be one of the most turbulent times in human history," Bill Gates writes in a new, 12-page essay. https://t.co/n64xTCSH1L

See 7 related tweets

@WOLF_Financial: BILL GATES SAID, "AI WILL EITHER BE THE GREATEST EQUALIZER EVER INVENTED, OR THE WORST SOURCE OF INJ...
@erikbryn: It's great to see more and more people, including @BillGates, recognize that we are not prepared for...
@business: Bill Gates waded into the debate over artificial intelligence, urging tech companies and regulators ...
@ShereenBhan: Bill Gates pens a warning on AI adoption & regulation.
calls for “slowing down”. ‘The geopolitica...
@PolymarketMoney: JUST IN: Bill Gates reveals AI will be “either the greatest equalizer ever invented, or the worst so...

17. ibuildthecloud (Group Score: 164.1 | Individual: 43.3)

Cluster: 7 tweets | Engagement: 1708 (Avg: 50) | Type: Tech

RT @tobi: I’m thinking about banning Claude code at Shopify until they change their mind and read AGENTS.md and .agents/skills etc.

Insisting on only reading CLAUDE.md sometimes leads to split brain problems when different team members use different tools. Just unnecessary.

See 6 related tweets

@kimmonismus: I think Tobi is right here.

Model-specific optimization can make sense. But if every coding agent e...

@derrickcchoi: We've always believed in an open ecosystem and continue to do so.

AGENTS.md is a simple, open forma...

@yacineMTB: It's honestly insane that Claude code is closed source. It's a JavaScript bundle that makes API call...
@ibuildthecloud: I'm so glad respected people with a voice are pushing on this. More companies and anthropic customer...
@charliedbecker: One crazy thing about using AI is I’m always troubleshooting problems and my solutions work enough b...

18. WhaleInsider (Group Score: 159.0 | Individual: 32.2)

Cluster: 7 tweets | Engagement: 484 (Avg: 655) | Type: Tech

JUST IN: 🇨🇳 China’s Moonshot AI in talks with Microsoft, Amazon & Google to launch Kimi K3 on Azure, AWS & Google Cloud.

Seeking up to 30% of related revenue, potentially the first major China AI–US cloud revenue-share deal - Reuters. https://t.co/xfTwg6Dm3T

See 6 related tweets

@wallstengine: China's 🇨🇳 Moonshot AI is in talks with Microsoft, Amazon and Google to bring Kimi K3 to Azure, AWS ...
@StockSavvyShay: $MSFT,$ AMZN and $GOOGL could become the distribution layer for one of China’s leading AI models.

M...

@Techmeme: Sources: Moonshot AI is in early talks over revenue-sharing agreements with Microsoft, Amazon, and G...
@PolymarketMoney: JUST IN: China’s Moonshot AI is reportedly in talks with Microsoft, Amazon and Google to bring Kimi ...
@etnshow: BREAKING: MoonshotAI in talks with Google, Amazon and Microsoft over revenue-sharing agreements that...

19. HamelHusain (Group Score: 157.2 | Individual: 47.8)

Cluster: 4 tweets | Engagement: 1236 (Avg: 171) | Type: Tech

RT @patrick_oshag: Neil Movva (@neilmovva) started his career at Nvidia, working on GPUs and kernels, and has an unusually deep understanding of inference, from software to chips to power.

We spend a lot of time on each of those layers, how they connect, and where the important tradeoffs are.

What makes this conversation special is how detailed it is (like a 401-level class), yet Neil makes it remarkably clear and easy to follow.

Today he runs Sail Research, a company building infrastructure for agents to make tokens as cheap as possible.

We discuss:

Latency versus throughput
Why there are no bad chips, only bad pricing
The end of kernel engineering
Buying chips and power no one else wants
New chip architectures
Nvidia lore + his contrarian view of the company
Open source and the frontier labs

I learned a ton. Enjoy!

TIMESTAMPS 0:00 Intro 0:38 Building a “Token Factory” 4:21 The Future of Background Agents 13:09 Nvidia and the GPU Stack 23:27 Chips, Memory, and Transformers 36:14 The Future of AI Training Data 44:32 Chip Scarcity and Compute Arbitrage 52:44 Reinventing the AI Data Center 59:01 Power and the “Scavenger Strategy” 1:10:10 Open vs. Closed AI

See 3 related tweets

@aakashgupta: OpenAI's first chip runs at 700 watts and just posted benchmarks beating Nvidia's 1,400 watt flagshi...
@patrick_oshag: I love interviewing legends of tech, business and investing…

But nothing is as fun for me as this s...

@laurenmhreeder: The next generation of founders is going to be alright. @patrick_oshag leads an epic teardown of the...

20. rewind02 (Group Score: 153.6 | Individual: 38.1)

Cluster: 5 tweets | Engagement: 73 (Avg: 86) | Type: Tech

here's how Grok Bot can run actual business operations for you, not just chat:

when the platform you use doesn't have a native plugin, install the CLI directly

same move works for Whop, or any platform with a CLI

start with a chief of staff agent, give it access to your existing docs, have it audit and propose your first 3 hires
each specialist gets one job - a sourcing agent, a developer agent, an outreach agent
chief of staff routes tasks and reviews output before anything goes out
routines handle the recurring stuff - daily/weekly reports, five lines max, sent up to chief of staff

why it stays lean on purpose: more agents means more context bleed and faster token burn, not more output

one operator ran an entire business on 10% of a week's token budget because nothing overlapped

the same CLI-install pattern that got Shopify running works for any platform without a native connector - point the agent at the CLI, let it operate the store, source products, manage listings, all from the same chat you're already using to run everything else\n\nQT @doublenickk: https://t.co/n0sTyn9R6r

See 4 related tweets

@Lummox_eth: I gave Grok Bot $50 and one command to launch a business, then watched it handle the entire workflow...
@sairahul1: Someone just built a directory with 170+ ready-to-use Grok Bots.

Marketing, sales, productivity, op...

@rewind02: RT @rewind02: here's how Grok Bot can run actual business operations for you, not just chat:

when t...

@martin_casado: RT @mntruell: Grok Bot is now available to everyone with a standard Grok or Cursor subscription. It'...