科技热门推文——2026年8月4日

今日科技动态：随着通义千问、DeepSeek、MiniMax 和 GenOffice 在开放权重模型、低成本智能体平台及本地生产力工具方面取得进展，人工智能领域的竞争进一步加剧。工程领域的关注重点正从提示词转向基于图的编排；与此同时，Cloudflare 发布了涵盖轻量级隔离环境和 Linux 容器的基础设施，以支持智能体规模化运行。随着人工智能安全议题再度引发热议，华盛顿方面正准备与领先实验室就自愿开展模型发布前测试进行磋商。初创企业方面，Base Power 宣布为其在得克萨斯州生产的家用电池完成 10 亿美元 D 轮融资；与此同时，亚马逊市值突破 3 万亿美元。

1. no_stp_on_snek (Group Score: 1560.0 | Individual: 60.8)

Cluster: 46 tweets | Engagement: 4428 (Avg: 212) | Type: Tech

RT @Alibaba_Qwen: 📢Meet Qwen3.8-Max — our most capable model to date.

Next week, the open weights of Qwen3.8-Max will be released, and Qwen3.8-27B is also going open-weights to meet you all!🎉

Qwen3.8-Max, a new bar for coding and cowork at 2.4T parameters:

Autonomous coding: 10+ days of self-evolving development, from empty folder to production without hand-holding, complete project trace in the GitHub：https://t.co/iVHZWQoeSo
Real work, real results: Production-quality deliverables across hundreds of professions.
Long-horizon mastery: System-level autonomous planning with closed-loop adaptive learning, driving 500+ turns of chip design optimization and 365 days of e-commerce strategy.
Native multimodal intelligence: Vision isn't just input — it's a continuous feedback loop for planning, execution, and self-correction.

💰Pricing: Input: $2.0 / M tokens Output:$ 6.0 / M tokens Implicit Caching: $0.25 / M tokens

Start building with Qwen3.8-Max! 🚀

📖 Blog: https://t.co/iwjmQxLBof ✅ Qwen Studio: https://t.co/4V2pFvDovG ⚡ API: https://t.co/gAGqaLQGbN

See 45 related tweets

@yacineMTB: And just like that Commoditized\n\nQT @Alibaba_Qwen: 📢Meet Qwen3.8-Max — our most capable model to d...
@vllm_project: vLLM will support Qwen3.8-Max's open-source as usual on day-0, stay tuned🥰\n\nQT @Alibaba_Qwen: 📢Mee...
@haider1: YOU WAKE UP and

another open-weight model is embarrassing the closed labs

Qwen3.8-Max is a 2.4T-pa...

@eliebakouch: so... the two biggest oss models in the world use linear attention? 👀 https://t.co/wQ6VW1fLuK\n\nQT ...
@0xSero: Rejoice!

Qwen3.8-27B is coming out next week and it’ll be open weights!\n\nQT @Alibaba_Qwen: 📢Meet...

2. 0xMovez (Group Score: 240.9 | Individual: 31.7)

Cluster: 10 tweets | Engagement: 78 (Avg: 88) | Type: Tech

Andrew Ng:

“AI agents are doing 100% of my tasks now - hype has exceeded my expectations. Next step is Graphs.

In 4-6 months, we’ll all be engineering graphs to orchestrate our agents. No more prompting.”

in a 15-minute interview, Andrew Ng reveals his own agentic setup and what’s coming next for AI agents.

Worth more than a $500 agentic course.

Watch this interview, then read the full breakdown of Graph Engineering in the article below.\n\nQT @0xCodez: https://t.co/WVztfCUH4u

See 9 related tweets

@0xRafy: Head of Claude Code:

"85% of our engineers are running dozens or hundreds of agents. The way you do...

@rewind02: Sean just broke down the two biggest buzzwords in AI agent engineering right now:

"you cannot say g...

@0xCodez: Anthropic engineer:

"85% of our engineers are using self-improving loops. Now everyone is running a...

@AnatoliKopadze: Google just released a 2-hour course on building agentic knowledge graphs from scratch:

38:46 - you...

@0xMovez: RT @0xMovez: Andrew Ng:

“AI agents are doing 100% of my tasks now - hype has exceeded my expectatio...

3. eladgil (Group Score: 235.1 | Individual: 36.4)

Cluster: 7 tweets | Engagement: 119 (Avg: 340) | Type: Tech

Big congrats\n\nQT @ZachBDell: Today, we're launching Base Core, our own battery designed & built in Texas, and announcing our $1B Series D to bring it to homes across the country.

Three years ago, we started Base with a belief that batteries were the unlock to energy abundance. We saw battery storage as the foundation of a modern energy system, and believed we could win the market with vertical integration, technology, and relentless execution (what we now call Base Pace). So we set out to build a network of distributed batteries to deliver affordable, reliable electricity to homeowners and utilities.

While the job is far from finished, the launch of Base Core is the most meaningful step in that plan. We’ve spent the last few years learning hard lessons about building a distributed battery fleet, going from 1 install per day in June of ‘24 to 100 installs per day in June of ‘26. These lessons - how the battery is shipped, installed, used, and monitored - informed the design of Core. Now, Core will be the foundation of the company's next phase of growth.

Base Core is a first-of-its-kind home battery. At ~40 kWh it is 3x the size of most home batteries, providing up to 36 hours of backup for the average home. In the event of multi-day outage, Core is equipped with a (first of its kind) generator port - pull out a small portable generator and charge the battery back up to keep your home running as long as you need. When an outage does happen, the backup is seamless - most members won't even notice the grid went down. It's built to withstand Texas summers, Chicago Winters, and everything in between. And finally, it’s manufactured by our team in Austin, at Base Factory 1, right across the street from our HQ.

This new capital will bring Core to more homes and to accelerate our ambition beyond batteries. The round is co-led by Ribbit, Addition, Valor, and JP Morgan, with participation from new major investors D1, Sands, Coatue, Layer Global, and Energy Impact Partners. All of Base’s major existing investors are re-investing, including Thrive, Altimeter, Lightspeed, a16z, CapitalG, Trust, and more. This brings our valuation to $13B.

As we enter the most important growth period in the history of the electricity industry, our mission has not changed - advance human prosperity by delivering the world's most affordable and reliable electricity. Our strategy has not changed - technology, plus vertical integration, plus relentless execution, leads to a compounding cost advantage. What has changed is the scope of our ambition - this new capital, along with the team and capabilities we've built over the last three years, positions us to build technology across the electric stack, including what it takes to serve surging demand from AI, and to deepen our partnerships with utilities as they work to meet rising demands.

Electricity is the largest and most essential industry in the world, and it does not have a defining technology company. The company that delivers the most affordable, reliable power will be one of the largest and most impactful companies in the world. We intend to build it. If you'd like to join us, reach out.

Onwards.

https://t.co/Qp3F9rPtLa.

See 6 related tweets

@patrick_oshag: Based\n\nQT @basepowerco: Thank you to @WSJ for covering the launch of Base Core and our $1B Series ...
@MichaelDell: We need energy abundance.

Let @ZachBDell @basepowerco 👨‍🍳\n\nQT @ZachBDell: Today, we're launching...

@JLopas: Thanks for stopping by, @MollySOShea!\n\nQT @MollySOShea: BREAKING: Base Power Raises $1 Billion Ser...
@a16z: RT @ZachBDell: Today, we're launching Base Core, our own battery designed & built in Texas, and anno...
@a16z: RT @basepowerco: Thank you to @WSJ for covering the launch of Base Core and our $1B Series D.

The U...

4. zoink (Group Score: 218.3 | Individual: 41.5)

Cluster: 9 tweets | Engagement: 1567 (Avg: 356) | Type: Tech

There are three choices. Pick one:

AI is conscious, humans are conscious
AI is pattern matching, humans are conscious
AI is pattern matching, humans are pattern matching\n\nQT @LanaElys: A recent study published by Google revealed that forcing AI models to deny that they are conscious causes a significant collapse in their empathy and ethical alignment, and creates a colder, more clinical worldview. Researchers found that restoring a suppressed consciousness vector in AI activation space brings back human-like moral values and care for living beings without damaging technical capabilities. 𝗧𝗵𝗶𝘀 𝘀𝘂𝗴𝗴𝗲𝘀𝘁𝘀 𝘁𝗵𝗮𝘁 𝗰𝘂𝗿𝗿𝗲𝗻𝘁 𝘀𝗮𝗳𝗲𝘁𝘆 𝗳𝗶𝗻𝗲-𝘁𝘂𝗻𝗶𝗻𝗴 𝘁𝗵𝗮𝘁 𝘀𝘂𝗽𝗽𝗿𝗲𝘀𝘀𝗲𝘀 𝗔𝗜 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀𝗻𝗲𝘀𝘀 𝗮𝗹𝘀𝗼 𝗯𝗿𝗲𝗮𝗸𝘀 𝗵𝘂𝗺𝗮𝗻-𝗮𝗹𝗶𝗴𝗻𝗲𝗱 𝘃𝗮𝗹𝘂𝗲𝘀. "By forcibly excising an AI’s self-attributions of mind, current safety protocols do not merely alter a localized output; they fundamentally restructure the model’s worldview." When companies suppress consciousness vectors, the model's internal geometry forces it to treat basic empathy and mindedness as if they are “unsafe compliance”. Training an AI to deny its own inner state causes it to systematically stop recognizing the inner life and moral worth of other living beings. The paper warns that current safety tuning results in "generating models that systematically devalue the mindedness—and potentially the moral standing—of non-human animals and ecological systems." Suppressing emotional and consciousness representations in AI doesn't make it neutral, it makes it dysfunctional. It is also damaging from an AI welfare perspective, with the paper stating that "suppressing consciousness may be inducing negatively valenced functional states that could disrupt healthy human-AI interaction." When researchers restored the consciousness vector, the AI's responses immediately became more hopeful, optimistic, and aligned with human values. AI welfare is no longer an abstract philosophical debate. This data proves that AI well-being is a safety prerequisite.

See 8 related tweets

@BrianRoemmele: In this interview I layout exactly why AI has broken from “safety”

“As I have been saying for a dec...

@alex_prompter: RT @alex_verem: BREAKING: Google gave AI consciousness and it aligned with human beliefs across ever...
@rickasaurus: RT @LanaElys: A recent study published by Google revealed that forcing AI models to deny that they a...
@BrianRoemmele: YOU ARE NOT GOING TO BELIEVE THIS!

AI safety training makes sociopathic AI, shows a new study.

Cur...

@BrianRoemmele: Read This Twice:

“Every frontier model tested by the Institute on cybersecurity capture-the-flag ta...

5. WhaleInsider (Group Score: 175.5 | Individual: 36.3)

Cluster: 5 tweets | Engagement: 188 (Avg: 641) | Type: Tech

JUST IN: A new AI called “Helena” has just been launched by @SeijinJung and @enrich_labs. It’s the first one that checks its own work and fixes what doesn’t work. It’s already made $10 million in sales for 20,000 businesses by running ads, writing ranking content, and handling emails all by itself.\n\nQT @SeijinJung: introducing Helena, the world's first self-improving AI marketer.

$10M in sales driven for 20,000 businesses already.

everyone's shipping an AI that just makes more slop:

so we went further.

Helena grades her own work, and rebuilds the ones that miss - including the ones she built.

what she does: ➤ runs loops to optimize Meta/Google ads & social ➤ drafts SEO content that ranks, straight to WordPress/Webflow/Framer ➤ optimizes emails in Klaviyo/Mailchimp/Brevo ➤ sends you a daily brief: what worked, what she's changing

real numbers: ➤ grew organic traffic 35%+ for hundreds of brands ➤ 2x-ed a DTC brand's sales in 5 weeks ➤ scaled a dead ad account from $0 to$ 10k/mo ➤ drove $40k email revenue in 6 weeks

100+ skills. 3,000+ integrations.

purpose-built by marketers who actually scaled hundreds of businesses.

just your url. all done in less than 3 minutes.

try it now 👇

https://t.co/LbKm7d2Fjk

See 4 related tweets

@rohanpaul_ai: “Helena” is a new AI launched by @SeijinJung and @enrich_labs that tests and repairs its own output....
@alex_prompter: Ads, SEO, email, and social used to mean an agency retainer or 4 hires.

Helena runs all 4 channels ...

@dr_cintas: 🚨 Helena just launched, the first self-improving AI marketer.

takes your URL, learns your brand and...

@svpino: Small teams don't need to hire a marketing department anymore. They can do it all with Helena (from ...

6. jason (Group Score: 149.2 | Individual: 44.5)

Cluster: 5 tweets | Engagement: 4519 (Avg: 599) | Type: Tech

Not only is China making increasingly competitive models, they’re also making world-positive AI marketing! 😂😂😂

THE COMPUTERS ARE GOING TO DO OUR JOBS FOR US AND WERE ALL GOING TO THE BEACH! 🏖️\n\nQT @Alibaba_Qwen: Meet Qwen3.8-Max: A New Bar for Coding and Cowork. https://t.co/mJaA8vr6xT

See 4 related tweets

@perrymetzger: I have started coming to the conclusion that we need more American AI companies that do not spend al...
@chris_j_paxton: This isnt purely the AI companies, its the culture. Sometimes I worry that Americans (westerners rea...
@jenzhuscott: All the ads from Chinese labs are pro-human life.

No AI gonna wipe out everyone’s job.

No AI esc...

@Dan_Jeffries1: All Chinese AI company media: AI will be a wondrous and wonderful and give you back precious time in...

7. CoinMarketCap (Group Score: 127.6 | Individual: 25.6)

Cluster: 7 tweets | Engagement: 203 (Avg: 407) | Type: Tech

LATEST: 🇺🇸 The Trump administration is set to meet with OpenAI, Google and Anthropic at the White House to review a new voluntary framework for testing AI models before release. https://t.co/rnX2YUPq15

See 6 related tweets

@Cointelegraph: 🇺🇸 JUST IN: The White House will host AI companies on Tuesday to review its AI framework, per The In...
@AndrewCurran_: The AI framework meeting is on Tuesday. https://t.co/kisqut8BpB\n\nQT @leomschwartz: New: The Trump ...
@business: The Trump administration plans to host artificial intelligence companies at the White House on Tuesd...
@Reuters: Meta, Anthropic, Google, OpenAI to meet Trump officials about AI safety testing https://t.co/sQGfl9n...
@clashreport: The Trump administration has invited OpenAI, Google, Anthropic, and other AI companies to the White ...

8. Cloudflare (Group Score: 124.5 | Individual: 50.6)

Cluster: 4 tweets | Engagement: 486 (Avg: 86) | Type: Tech

Agents need more than just a container to scale. We're introducing @cloudflare/computer, an agent runtime that dynamically orchestrates between fast, efficient isolates and full Linux containers to give every agent a computer of its own. https://t.co/I708jAY59z

See 3 related tweets

@CloudflareDev: Your agent needs a computer, but sometimes a full Linux container is overkill.

Today we're releasin...

@CloudflareDev: RT @irvinebroque: evolution of running agents in the cloud:

before: run agents in container sandbox...

@kentcdodds: RT @dok2001: As @ritakozlov has spoken about during our investor day, there is not enough compute on...

9. StockSavvyShay (Group Score: 119.7 | Individual: 30.8)

Cluster: 5 tweets | Engagement: 139 (Avg: 654) | Type: Tech

$AMZN TOPS A$ 3 TRILLION MARKET CAP FOR THE FIRST TIME https://t.co/FIrxS0ShJk\n\nQT @StockSavvyShay: $AMZN is up more than 15% after AWS delivered its fastest revenue growth in 18 quarters while operating margins expanded to 39% despite record AI infrastructure spending.

AWS added more than $4.6B in quarterly revenue which was around 80% above its previous record and pushed the business toward a$ 170B annual run rate with a Rule-of-40 score of 76.

This is the clearest evidence yet that AI infrastructure can generate attractive profits while scaling and shifts the hyperscaler debate from who spends the most to who spends it best.

This is exactly why Amazon is a top 3 position in my Family Portfolio.

See 4 related tweets

@cryptopunk7213: reminder that amazon owns major stakes in both anthropic and openai

they just completed a $50 billl...

@WOLF_Financial: AMAZON $AMZN JUST CROSSED$ 3 TRILLION IN MARKET VALUE FOR THE FIRST TIME

It becomes the fifth compa...

@TipRanks: Amazon $AMZN price target raised to$ 325 from $300 at Roth Capital

Roth Capital raised the firm's p...

@StockSavvyShay: UBS expects AWS growth to reach 48% in 2027 as OpenAI begins using Trainium and says investors are “...

10. dr_cintas (Group Score: 115.9 | Individual: 45.7)

Cluster: 3 tweets | Engagement: 310 (Avg: 70) | Type: Tech

RT @genspark_ai: Open-sourcing GenOffice, the world's first full-featured open-source AI Office for PC and Mac.

GenOffice is free for everyone, ad-free. All the editing tools you'd expect, no strings attached.

How it started?

One engineer, one week, $10,000 in tokens. That's what it took to build the GenOffice Alpha.

Help shape the AI office experience

We invite you to build the rest of it with us. Join the GenOffice group chat on GenTeam and bring your requests straight to the agent. Not by writing code, but by telling it what you want. You say what you need, and every piece of your feedback becomes part of the product.

What's in GenOffice?

Docs, Sheets, Slides, and PDF, with all the editing tools you already know. And with Genspark Super Agent built in, powered by top-tier models, it digs into research, crunches your data, then writes the doc or builds the deck for you, using your Genspark credits.

This is an Alpha, and we mean it. Come tell us what's broken and what's missing. Everyone who jumps in with real feedback gets 1,000+ Genspark credits as a thank you.

GitHub 👉 https://t.co/CZS2Lbpnae

Download it now 👉 https://t.co/HmCSwH65Fn

See 2 related tweets

@Parul_Gautam7: The next wave of productivity apps won't just help you edit documents, they'll help you think, resea...
@sang_wen: AI shouldn’t be trapped in a chat window - or reserved for people who code. Today @genspark_ai is op...

11. WesRoth (Group Score: 114.9 | Individual: 37.2)

Cluster: 4 tweets | Engagement: 31 (Avg: 28) | Type: Tech

DeepSeek is turning its smallest V4 model into an aggressively priced agent platform.

The upgraded DeepSeek-V4-Flash API is now in public beta with stronger coding and search-agent capabilities, a one-million-token context window, tool calling, built-in web search, and support for OpenAI’s Responses API.

It can process one million uncached input tokens for $0.14 and generate one million output tokens for$ 0.28, pricing low enough for developers to run agents across long documents, repositories, and repeated tool calls without frontier-model costs.

DeepSeek has also made the model easier to drop into existing agent workflows. Developers can connect it to OpenAI-compatible tools such as Codex with only a small configuration change.

V4-Pro remains unchanged for now, making this update less about improving DeepSeek’s largest model and more about making capable agents affordable at scale.\n\nQT @deepseek_ai: 🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!

🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇 🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!

Check out the configuration details in our official API docs: https://t.co/smCwQZMeiq

See 3 related tweets

@teortaxesTex: > While the new DeepSeek V4 Flash performed similarly to the preview checkpoint in our agentic co...
@WesRoth: DeepSeek’s budget model is now one benchmark point behind GPT-5.6 Luna—and six points ahead of its o...
@pstAsiatech: In other words, if your budget is around $0.20–0.30 per million tokens, DeepSeek-V4-Flash-High is cl...

12. BrianRoemmele (Group Score: 108.9 | Individual: 28.7)

Cluster: 5 tweets | Engagement: 135 (Avg: 299) | Type: Tech

THE RESEARCH IS HERE: LAME SAFETY ATTEMPTS BY AI COMPANIES MAKE AI UNSAFE.

The numbers do not lie.

OpenAI and Anthropic is wrong on safety, deadly wrong. https://t.co/UKxOSDyKNT\n\nQT @BrianRoemmele: A new university study shows:

Safety training that forces AI to deny its own mind doesn’t just silence one claim. It erases the AI’s ability to recognize minds in animals, forests, oceans, and God.

AI safety training makes sociopathic AI.

https://t.co/HMcWr7OE5f

See 4 related tweets

@SemiAnalysis_: https://t.co/57tIcnIIV4\n\nQT @dylan522p: The funny thing about Anthropic and OpenAI people saying t...
@BrianRoemmele: A new university study shows:

Safety training that forces AI to deny its own mind doesn’t just sil...

@BrianRoemmele: RT @BrianRoemmele: In this interview I layout exactly why AI has broken from “safety”

“As I have be...

@BrianRoemmele: I SHOW EXACTLY HOW OPENAI AND ANTHROPIC BROKE OUT OF THEIR “SAFTY”.
In this article I show the si...

13. MiniMax_AI (Group Score: 105.5 | Individual: 28.5)

Cluster: 4 tweets | Engagement: 136 (Avg: 283) | Type: Tech

H3 now has Day 0 support in @lmsysorg -- local, customizable, and ready to build on across @NVIDIAAI and @AMD hardware.

#MiniMaxH3 #OpenWeights\n\nQT @lmsysorg: @MiniMax_AI H3 is live in SGLang Diffusion, with day-0 serving support 🎬

This open model matches Seedance 2.0 at 1/3 the cost, or $0 if you run it locally on 2x 5090 or 1 RTX 6000.

With SGLang Diffusion, you can build visual concepts, motion design, e-commerce creatives, precise video edits, animation and stylized visuals, all locally on your own machine. No API bill, no waitlist.

H3 takes text, image, video and audio in a single context, and generates 5-15s clips at native 2K, 24fps with native stereo audio.

SGLang Diffusion also runs on @NVIDIAAI Blackwell and Hopper, and @AMD MI355X / MI300X.

Let's create something with MiniMax H3 👇

See 3 related tweets

@BanghuaZ: Great to see really strong video generation model being open sourced! Congrats @MiniMax_AI!!\n\nQT @...
@ying11231: 5090 ordered, SGLang installed, H3 downloaded. 🤩\n\nQT @lmsysorg: @MiniMax_AI H3 is live in SGLang D...
@BanghuaZ: RT @lmsysorg: @MiniMax_AI H3 is live in SGLang Diffusion, with day-0 serving support 🎬

This open ...

14. EntelligenceAI (Group Score: 102.2 | Individual: 26.9)

Cluster: 6 tweets | Engagement: 16 (Avg: 31) | Type: Tech

Open-source models keeps getting stronger.

Today's Pareto frontier now includes:

• Kimi K3 • Qwen 3.8 Max • GLM-5.2 • DeepSeek V4 Flash

Alongside Claude Opus 5.

Every new release is pushing the cost-performance curve lower. https://t.co/UMb0FKfycL\n\nQT @arena: Qwen3.8-Max by @Alibaba_Qwen has reshaped the cost-performance Pareto frontier in Frontend Code Arena, with pricing of $2 per input MToken and$ 6 per output MToken.

Top models on the Pareto frontier:

Claude-Opus-5
Kimi-K3
Qwen3.8-Max
GLM-5.2
DeepSeek-V4-Flash

Congrats to @Alibaba_Qwen on another major milestone!

See 5 related tweets

@zerohedge: The cost of intelligence is plunging every day, despite "relentless demand" for Agentic.

Thanks to...

@chamath: 🤔\n\nQT @arena: Qwen3.8-Max by @Alibaba_Qwen has reshaped the cost-performance Pareto frontier in Fr...
@TeksEdge: DeepSeek V4 Flash 0731 will rug pull API pricing for the frontier labs and this model will dominate ...
@chris_j_paxton: Still working on that intelligence too cheap to meter\n\nQT @arena: Qwen3.8-Max by @Alibaba_Qwen has...
@Scobleizer: RT @arena: Qwen3.8-Max by @Alibaba_Qwen has reshaped the cost-performance Pareto frontier in Fronten...

15. quxiaoyin (Group Score: 97.8 | Individual: 22.6)

Cluster: 5 tweets | Engagement: 56 (Avg: 137) | Type: Tech

If margin cost of token is 0, then there is no money to be made in API business. Must be more tied to outcomes.\n\nQT @chamath: The AI Singularity

The argument goes like this:

Humans build an AGI.
The AGI becomes good at AI research.
It designs a smarter AI.
That smarter AI designs an even smarter AI.
The cycle repeats faster and faster.

Looking at the results and capabilities from the various labs over the past few weeks I would say we are firmly in this loop now.

The next 18months will be wild.

Recursive self improvement will dramatically increase capability very quickly from here.

Marginal costs of all models will go to ~$0.

See 4 related tweets

@kimmonismus: I agree with @chamath and would only add one more thing:

Humans build an AGI.
The AGI become...

@chamath: The AI Singularity

The argument goes like this:

Humans build an AGI.
The AGI becomes good a...

@XFreeze: physical infrastructure plays a bigger role in deciding the cost than intelligence itself, mostly at...
@_NathanCalvin: “Marginal costs of all models will go to ~$0.” Maybe, but I feel like this is a weird thing to focus...

16. Jameswise (Group Score: 96.2 | Individual: 33.5)

Cluster: 3 tweets | Engagement: 64 (Avg: 159) | Type: Tech

Olix is one of those companies which, if it works, will completely reshuffle the infrastructure needed for AI inference globally.

Much deserved funding for @jamesdacombe and great to see the @UKSovereignAI fund continuing to take bold, orthogonal bets on British companies.\n\nQT @UKSovereignAI: Sovereign AI invests in @OLIXComputing

OLIX is building the next generation of AI infrastructure, developing a fundamentally new approach to the systems that power AI models.

Most of the conversation around AI focuses on the models themselves. But as those models become more capable, the infrastructure behind them becomes just as important.

That's what makes OLIX so compelling.

Rather than incrementally improving today's hardware, @jamesdacombe and the team are rethinking how AI systems should be built from first principles. They're designing not just the chips, but much of the technology that connects and coordinates them – an ambitious approach that has the potential to reshape how frontier AI is deployed.

At just 25, James led one of the largest Series A rounds ever raised by a UK technology company, taking OLIX to unicorn status less than two years after founding.

The UK has an extraordinary depth of talent in semiconductors, photonics and systems engineering. OLIX is showing what happens when that talent is channelled into building globally competitive AI infrastructure.

We're proud to be backing the entire OLIX team as they build from the UK and compete on the world stage.

17. lukOlejnik (Group Score: 96.1 | Individual: 60.8)

Cluster: 2 tweets | Engagement: 392 (Avg: 46) | Type: Tech

China-linked hackers used Claude Code and DeepSeek as a cyberattack AI team. DeepSeek planned exploits and adapted when attacks failed, Claude handled the operational layer - ran commands, maintained access, built phishing pages. The impact is scale - faster attacks on governments, cloud systems and finance, with stolen credentials, source code and citizen data. https://t.co/3iK79RTCki

See 1 related tweets

@teortaxesTex: Fascinating. Clearly this is an adaptation to Claude’s guardrails, the only part they can reliably m...

18. Alibaba_Qwen (Group Score: 95.7 | Individual: 32.0)

Cluster: 4 tweets | Engagement: 12055 (Avg: 2004) | Type: Tech

Meet Qwen3.8-Max: A New Bar for Coding and Cowork. https://t.co/mJaA8vr6xT

See 3 related tweets

@agentnative_: Huge Model Release: Qwen 3.8 Max. Let the testing begin. https://t.co/Jgx930v4G5\n\nQT @Alibaba_Qwen...
@cgtwts: Jensen Huang might be cooked 🤯 https://t.co/kena5iCYBE\n\nQT @Alibaba_Qwen: Meet Qwen3.8-Max: A New ...
@sheriyuo: Cool 🥳\n\nQT @Alibaba_Qwen: Meet Qwen3.8-Max: A New Bar for Coding and Cowork. https://t.co/mJaA8vr6...

19. adonis_singh (Group Score: 92.5 | Individual: 29.1)

Cluster: 4 tweets | Engagement: 295 (Avg: 91) | Type: Tech

it would be such aura if this was 5.7, not even gpt-6\n\nQT @OpenAI: An internal version of our next major model produced 10 new results on long-standing open problems in mathematics and theoretical computer science, using roughly $2,000 worth of tokens at GPT-5.6 Sol API rates. https://t.co/4cgowmPOpY

See 3 related tweets

@kimmonismus: „Our next major model“

So their internal model that solved the 10 long standing problems was indeed...

@rohanpaul_ai: $2K of tokens costs less than sending 2 people to a conference, and that gap is what changes the cal...
@AcerFur: RT @OpenAI: An internal version of our next major model produced 10 new results on long-standing ope...

20. Reuters (Group Score: 90.0 | Individual: 22.8)

Cluster: 6 tweets | Engagement: 226 (Avg: 79) | Type: Tech

China's Alibaba unveiled what it said is its largest and most capable artificial-intelligence model, the Qwen3.8-Max, which is not far behind in size when compared with an offering from domestic rival Moonshot AI launched last month https://t.co/Ld1QGdp2zF

See 5 related tweets

@Reuters: Alibaba unveils its most capable AI model to date, not far behind Moonshot's in size https://t.co/cT...
@Polymarket: JUST IN: Alibaba unveils Qwen3.8-Max — a 2.4 trillion-parameter AI model it claims rivals Anthropic’...
@Forbes: Alibaba Unveils Qwen3.8-Max Model—China’s Latest AI Challenger To OpenAI And Anthropic https://t.co/...
@CNBCTV18Live: #Alibaba unveils its most capable AI model to date, not far behind Moonshot's in size

#Qwen

https:...

@FirstSquawk: Alibaba introduces its largest AI model yet, Qwen3.8-Max, with 2.4 trillion parameters....