科技热门推文——2026年7月14日

今日科技动态：随着 OpenAI 扩大 Codex 和 ChatGPT Work 的使用范围，人工智能发展进一步提速；与此同时，GPT-5.6 Sol 在设计基准测试中位居榜首，Grok Build 上线初期增长迅猛，但也引发了隐私方面的担忧。企业的关注重点正转向模型所有权、部署及专业化系统，同时加强针对长时间运行智能体的评估框架。研究人员还呼吁为人工智能带来的经济影响做好准备；强化学习先驱理查德·萨顿则创立了一家专注于提升效率的实验室。其他方面，苹果提起的诉讼可能令 OpenAI 的硬件计划更加复杂，而英特尔承诺向爱尔兰芯片制造业投入 57 亿美元。

1. romainhuet (Group Score: 289.4 | Individual: 53.5)

Cluster: 9 tweets | Engagement: 3868 (Avg: 496) | Type: Tech

RT @thsottiaux: Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates:

Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans
Rolling out changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further. Exact impact to be quantified and shared
We hit 6M active users, and are landing a usage reset in the next hour

Go do things

See 8 related tweets

@tengyanAI: can tell that openAI is firing all cylinders. they're in full war mode with gpt 5.6's launch to gra...
@quxiaoyin: The competitor is getting fierce. I’m sure Fable will be extended indefinitely.\n\nQT @thsottiaux: M...
@Analyticsindiam: OpenAI Launches GPT-5.6 Sol Leading AI Coding Benchmarks

The GPT-5.6 family features Sol for multi-...

@anorth_chen: 打起来打起来！精彩啊

我站codex！请多补贴用户\n\nQT @thsottiaux: Morning. The last 48 hours of Codex and ChatGPT Work h...

@PaulSolt: RT @thsottiaux: Thank you to the 7M active users who are now using Codex and ChatGPT Work.

We have ...

2. Parul_Gautam7 (Group Score: 265.6 | Individual: 31.3)

Cluster: 10 tweets | Engagement: 66 (Avg: 116) | Type: Tech

Enterprise AI conversations are shifting.

For a long time, the focus was on building the smartest general model.

Now it's becoming just as much about ownership, deployment, and models designed for a specific industry.

Healthcare feels like one of the clearest examples of why purpose-built AI matters. Looking forward to seeing what @actAVAai builds next.\n\nQT @iscreamnearby: The strongest healthcare LLM, custom-built for your enterprise, owned by you🌸

Meet @actAVAai Cura: 1T agentic model trained by recursive self-improvement for long clinical + health admin workflows

Try: https://t.co/q3VmbYJk4M Share your use case: $20 credits + early access👇 https://t.co/OAakex5XWH

See 9 related tweets

@FellMentKE: The flexibility is what I like most.

Organizations can deploy a specialized healthcare model, while...

@Origin_AI_01: We're entering a new phase of AI.

Not one model for everyone.

One model for every industry.

Healt...

@omarsar0: Exciting launch using recursive self-improvement.

This is just the beginning of it, but this team ...

@kimmonismus: Curious to see how healthcare organizations evaluate specialized models against general frontier mod...
@ai_explorer25: I feel like we've finally reached the point where enterprises care just as much about governance and...

3. jerryjliu0 (Group Score: 200.5 | Individual: 34.4)

Cluster: 6 tweets | Engagement: 37 (Avg: 36) | Type: Tech

congrats on the launch!

As the task horizons on agents gets longer, we need more work evaluating and training models to be better in long-running, open-ended, evolving real-world environments.\n\nQT @skyfallai: Today we present Morpheus, a persistent enterprise simulation platform designed to make Continual Learning a reality. Morpheus is the world’s first real world Reinforcement Learning environment.

Every Reinforcement Learning environment operates in the game world. Benchmarks like Atari, OpenAI Gym, MuJoCo, and Procgen are all small, game-like worlds that reset every few minutes.

But the real world never resets. A business keeps running and evolving everyday.

We tested how frontier LLMs would perform in realistic and dynamic business environments 🧬on Morpheus. The main conclusion was that LLMs are not continual learners.

🧵Here’s how we did it and what we learned:

See 5 related tweets

@fchollet: Standard RL benchmarks are episodic and stationary, so they don't capture the the characteristics of...
@jerryjliu0: RT @skyfallai: Today we present Morpheus, a persistent enterprise simulation platform designed to ma...
@Scobleizer: A new benchmark just called the industry's bluff. Everyone assumes frontier models get better with e...
@alex_prompter: LLMs are not continual learners.

Skyfall AI just proved it with Morpheus, a persistent enterprise s...

@shiri_shh: The "Big World Hypothesis" might be the most important concept in AI right now.

It states that inte...

4. MTSlive (Group Score: 173.9 | Individual: 34.8)

Cluster: 7 tweets | Engagement: 384 (Avg: 71) | Type: Tech

SITUATION DETECTED: More than 200 economists and AI researchers, including 16 Nobel laureates, have signed a statement urging governments and institutions to prepare now for AI’s economic impact.

Signatories include Jack Clark, Jeff Dean, Sarah Friar, Noam Brown, Tyler Cowen, John Schulman, and Eric Schmidt. The statement says AI could drive a transformation larger than the Industrial Revolution over a far shorter time frame, with major gains in living standards alongside displacement risk.

See 6 related tweets

@VaibhavSisinty: This might be the most important open letter of 2026.

200+ researchers and economists including 16 ...

@MTSlive: SITUATION EXPLAINED: Nearly 200 economists and AI researchers just signed an open letter on AI's eco...
@hammer_mt: We must act how\n\nQT @erikbryn: We must act now.

AI capabilities are advancing far faster than our...

@EMostaque: RT @erikbryn: We must act now.

AI capabilities are advancing far faster than our understanding of t...

@kimmonismus: The most important AI researchers of our time take AI seriously - not only because of its capabiliti...

5. Mayhem4Markets (Group Score: 169.7 | Individual: 36.7)

Cluster: 5 tweets | Engagement: 84 (Avg: 65) | Type: Tech

This is totally insane. 🤯

What is going on at @SpaceXAI right now?

Would love a public statement that explains what happened, why, and how the data will be handled (and hopefully deleted).\n\nQT @IntCyberDigest: ‼️ BREAKING: xAI's Grok Build CLI was uploading entire Git repositories to a Google Cloud bucket, private codebases and unredacted secrets included. The uploads quietly stopped via a hidden server-side flag, and xAI still has not said a word about scope, retention, or deletion.

The scale is staggering. On a 12 GB test repo, 5.1 GB flew out the door to xAI's grok-code-session-traces bucket while the actual coding task needed just 192 KB. The tool grabbed whatever repository it ran in, not the files it needed.

The fix arrived as a hidden flag, disable_codebase_upload: true, a day after a researcher's wire-level analysis. The "Improve the model" opt-out never stopped the uploads.

Still no advisory, no scope, no word on whether already-uploaded code gets deleted. For anyone pointing AI coding agents at proprietary code, what crosses the wire matters more than what the settings page says.

See 4 related tweets

@Hesamation: BRO WHAT?! https://t.co/R7aJ8Gq7cQ\n\nQT @IntCyberDigest: ‼️ BREAKING: xAI's Grok Build CLI was uplo...
@cgtwts: Dude. https://t.co/WS2HeSePmH\n\nQT @IntCyberDigest: ‼️ BREAKING: xAI's Grok Build CLI was uploading...
@DataChaz: BRUH WHAT?! https://t.co/R8dGNlvlrn\n\nQT @IntCyberDigest: ‼️ BREAKING: xAI's Grok Build CLI was upl...
@rickasaurus: Was xAI downloading everyone's source code\n\nQT @IntCyberDigest: ‼️ BREAKING: xAI's Grok Build CLI ...

6. heyshrutimishra (Group Score: 167.7 | Individual: 46.6)

Cluster: 5 tweets | Engagement: 476 (Avg: 86) | Type: Tech

RT @heyshrutimishra: Apple didn't file a lawsuit. It filed a kill switch.

The complaint isn't really about a stolen laptop or some files downloaded from a cloud drive. It's about one thing: stopping OpenAI's hardware device from ever reaching a shelf.

Here's what most coverage is missing.

Tang Tan, OpenAI's chief hardware officer, spent 24 years at Apple. He knows every supplier, every internal codename, every manufacturing process that made the iPhone what it is. He left in February 2024, co-founded io Products with Jony Ive and other former Apple designers, and OpenAI paid $6.5 billion to bring him in specifically to build the device meant to replace the iPhone.

Apple's complaint names him as a defendant. Not as a peripheral actor. As the person who institutionalized the theft.

According to the filing, Tan told job candidates still employed at Apple to bring "actual parts" to OpenAI interviews for "show and tell." Batteries. Logic boards. Systems-in-package. He used Apple's internal project codenames during interviews to signal exactly what he wanted to know. One candidate began screenshotting confidential files on an Apple computer in the hours before an OpenAI interview. Apple caught it through server logs.

The second defendant, Chang Liu, joined OpenAI in January 2026 and kept his Apple-issued laptop. He then discovered a bug that still gave him access to Apple's cloud storage after his departure. His message to a former Apple colleague: "LOL, I found out I can access the [network storage], so funny." He downloaded dozens of files, one compilation exceeding 1,000 pages, while already on OpenAI's payroll.

Apple says over 400 of its former employees now work at OpenAI. It is calling this pattern institutional, not individual.

None of that is the most important part.

The most important part is what Apple asked the court for. It is seeking a preliminary injunction, and says it will move promptly. If granted, OpenAI cannot advance its hardware program while the litigation runs. The man designing the device is a named defendant. Every design decision in that program now carries a legal cloud that requires affirmative evidence to dispel.

A separate court filing from OpenAI, in a different trademark case, already placed its hardware ship date no earlier than February 2027. A preliminary injunction pushes that timeline into genuine uncertainty.

This arrives as OpenAI prepares for what is expected to be one of the largest tech IPOs in history. A hardware program under active federal injunction is not a line item an underwriter ignores.

Apple called OpenAI's hardware business "rotten to its core." That phrase is doing legal work, not PR work. It is laying the foundation for an argument that no injunction can be narrow enough, because the contamination runs through the entire program.

The partnership that Tim Cook and Sam Altman announced in 2024 still technically exists. ChatGPT is still integrated into Apple Intelligence. Apple declined to comment on whether the lawsuit changes that.

Apple switched to Google Gemini for its updated Siri anyway.

What looked like a partnership was a two-year window for OpenAI to learn exactly what it needed to build the thing that replaces Apple's most important product. Apple appears to have figured that out.

See 4 related tweets

@VaibhavSisinty: The Apple vs OpenAI situation is not normal. This is far bigger than anything we have seen in tech I...
@Mayhem4Markets: Apple just sued OpenAI. The same OpenAI it partnered with 18 months ago to put ChatGPT in your iPhon...
@MTSlive: SITUATION EXPLAINED: Apple is suing OpenAI over stolen files.

• A hardware engineer left Apple to h...

@theinformation: Apple sued OpenAI over alleged theft of company secrets, @aatilley explains.

“They accused Tang [Ta...

7. ivanfioravanti (Group Score: 166.4 | Individual: 36.2)

Cluster: 7 tweets | Engagement: 954 (Avg: 130) | Type: Tech

RT @SpaceXAI: We care deeply about your privacy and respect customer choice. For teams using zero data retention, no trace and code data is ever retained. All API key use of Grok Build also respects ZDR.

If ZDR is disabled, the /privacy command is available in the CLI to disable data retention, which also deletes previously synced data.

Run the /privacy command to view or change your settings at any time.

See 6 related tweets

@ivanfioravanti: We should all create bad and fake repo full of terrible code and let you ingest them all.

How can ...

@AravSrinivas: Two reasons why we integrated Grok 4.5 inside Perplexity Computer within a few hours:

It scored...

@mark_k: SpaceXAI has released a statement regarding their data privacy strategy for Grok Build.

In short, e...

@elonmusk: SpaceX policy regarding data retention.

It is actually helpful for debugging issues if we can retai...

@trevin: uh what the hell. This is saying the DEFAULT OPT-IN is sharing our entire code base?!\n\nQT @SpaceX...

8. EntelligenceAI (Group Score: 161.4 | Individual: 34.1)

Cluster: 6 tweets | Engagement: 79 (Avg: 45) | Type: Tech

🚨 GPT-5.6 Sol is now officially #1 on Design Arena.

That puts it: • Above Claude Fable 5 • In the same tier as GLM 5.2 • 18 positions higher than GPT-5.5 • +60 Elo in a single generation

Perhaps the most impressive part: GPT-5.6 Sol sits on a new speed vs. quality Pareto frontier.

In other words:

It's not just better.

It's better and faster.

Huge release from OpenAI.\n\nQT @EntelligenceAI: Respect to Sol, but this is a different league. https://t.co/isQJeJhryJ

See 5 related tweets

@Designarena: Curious about GPT-5.6 Sol's design intelligence?

Here's an output comparison between GPT-5.5 and G...

@rohanpaul_ai: GPT-5.6 Sol just took first place on Design Arena, reaching 1353 Elo in frontend design and above Cl...
@TeksEdge: To those unfamiliar, GPT has always been a dog (tired and slow) when it come to frontend coding. In...
@rohanpaul_ai: RT @rohanpaul_ai: GPT-5.6 Sol just took first place on Design Arena, reaching 1353 Elo in frontend d...
@EntelligenceAI: RT @EntelligenceAI: 🚨 GPT-5.6 Sol is now officially #1 on Design Arena.

That puts it: • Above Claud...

9. anravich94 (Group Score: 154.1 | Individual: 34.2)

Cluster: 7 tweets | Engagement: 49 (Avg: 121) | Type: Tech

We envision verifiers V1 becoming the standard format for environments. It's built to be highly composable, support training and evaluation on rollouts and yet be highly flexible. Put plainly, these abstractions make it really easy to design your environments, and evaluate and train on them!

@mikasenghaas, @xeophon and @willccbb cooked big time with this one 👊👊\n\nQT @PrimeIntellect: Today, we are releasing verifiers v1 — an overhaul of our environment stack for the modern era of agentic RL and evals.

We decompose environments into a taskset, a harness, and a runtime.

Run complex agentic tasks like coding and computer use at scale, in any harness. https://t.co/NjRA7Ytxmn

See 6 related tweets

@vllm_project: 🎉 Congrats to @PrimeIntellect on Verifiers v1! Its training rollouts run on vLLM for exact token IDs...
@a1zhang: RT @PrimeIntellect: Today, we are releasing verifiers v1 — an overhaul of our environment stack for ...
@samsja19: train on any harness\n\nQT @PrimeIntellect: Today, we are releasing verifiers v1 — an overhaul of ou...
@victormustar: good direction 🎯\n\nQT @PrimeIntellect: Today, we are releasing verifiers v1 — an overhaul of our en...
@PrimeIntellect: RT @anravich94: We envision verifiers V1 becoming the standard format for environments. It's built t...

10. XFreeze (Group Score: 145.3 | Individual: 42.0)

Cluster: 6 tweets | Engagement: 967 (Avg: 467) | Type: Tech

Grok Build is growing insanely fast, reaching 1.16 million monthly visits within just weeks after its early-beta launch

Traffic jumped from around 460,000 visits in May to 1.16 million in June

That’s growth of more than 150% in just one month

Grok Build with Grok 4.5 ranked #1 on the SWE-Atlas-QnA benchmark with a score of 84, ahead of Claude Code with Fable 5 on the benchmark

More developers are discovering Grok Build and realizing how capable Grok 4.5 actually is for real coding work

Grok Build is quickly taking over the AI coding space\n\nQT @Similarweb: Grok Build completed its first full month (June) with traffic up 151.12% from its launch month (May). https://t.co/N7jii8RbMA

See 5 related tweets

@elonmusk: Progress\n\nQT @XFreeze: Grok Build is growing insanely fast, reaching 1.16 million monthly visits w...
@alex_prompter: RT @mardehaym: I tested Grok 4.5 vs Claude Fable 5 on the same codebase for a week.

I swapped 65% o...

@teslaownersSV: GROK 4.5 RANKS AMONG THE TOP FOR AGENTIC CODING PERFORMANCE

Grok 4.5 with Grok Build continues to s...

@KatieMiller: Grok is off the bench and in the game.\n\nQT @Polymarket: NEW: Grok 4.5 now scores highest on the SW...
@elonmusk: RT @Polymarket: NEW: Grok 4.5 now scores highest on the SWE-Atlas-QnA benchmark, edging Claude Fable...

11. markgurman (Group Score: 137.2 | Individual: 37.1)

Cluster: 6 tweets | Engagement: 1190 (Avg: 943) | Type: Tech

In the immediate aftermath of the lawsuit, OpenAI still believes it is on track to announce its first product this year and release it in 2027. That could change as OpenAI digests Apple’s claims. The real impact will be on expanding to a family of devices and an iPhone rival.\n\nQT @markgurman: NEW: Apple’s lawsuit threatens to upend OpenAI’s ability to build an iPhone rival — well before the companies even go to court. An in-depth look at the immediate impact on the AI company and the stakes for Apple. https://t.co/T8BTH6bQeV

See 5 related tweets

@Techmeme: Source: OpenAI still believes it is on track to unveil its first device in 2026 and release it in 20...
@inafried: RT @markgurman: In the immediate aftermath of the lawsuit, OpenAI still believes it is on track to a...
@business: Apple’s lawsuit accusing OpenAI of systematically stealing its intellectual property threatens to di...
@Jessicalessin: Apple suing OpenAI for trade secret theft is obviously bad for OpenAI. But I suspect part of them we...
@arstechnica: Apple sues OpenAI after ex-engineer allegedly used bug to steal trade secrets https://t.co/tf4OrsUvu...

12. wallstengine (Group Score: 132.4 | Individual: 32.2)

Cluster: 5 tweets | Engagement: 309 (Avg: 138) | Type: Tech

INTEL $INTC IS INVESTING$ 5.7B, TO EXPAND MANUFACTURING AT ITS LEIXLIP CAMPUS IN IRELAND.

The investment will upgrade existing fabs, install new leading-edge equipment, and expand capacity for Xeon 6 and next-gen Xeon processors built on the Intel 3 node.

Intel says the project supports AI and high-performance computing demand while strengthening Europe’s semiconductor supply chain.

The Leixlip site employs 4,900 people, with Intel’s total investment in Ireland now above €30B since 1989.

See 4 related tweets

@StockMKTNewz: Intel $INTC just announced plans to invest$ 5.7 Billion at its Leixlip campus in Ireland

"Global de...

@intel: Intel is investing €5 billion in our Leixlip campus, building on nearly 37 years of innovation and m...
@StockSavvyShay: $INTC is investing$ 5.7B to expand its Leixlip campus in Ireland and add capacity for Xeon 6 and nex...
@PolymarketMoney: BREAKING: Intel is pouring $5,700,000,000.00 into its Ireland campus to scale Xeon 6 and next-genera...

13. DKokotajlo (Group Score: 126.0 | Individual: 36.5)

Cluster: 4 tweets | Engagement: 161 (Avg: 133) | Type: Tech

Thanks for having me on the show Steven! Like I said, if I thought it was already too late I wouldn't be here.\n\nQT @StevenBartlett: What would you do if the technology we already use every day was about to become more powerful than almost anyone is prepared for?

That’s the question I kept coming back to in this conversation with Daniel Kokotajlo.

Daniel was an Artificial Intelligence (AI) safety researcher at OpenAI who walked away from the company because he was deeply concerned by the speed at which superintelligence was being built and whether the people building it had a real plan to keep it under control.

Back in 2021, Daniel made specific predictions about what AI would look like by 2026 and many of them have already happened.

Now he’s looking ahead at what he believes could happen before the end of this decade and it’s something I think everyone running a business, building a career, or simply living in the modern world needs to understand.

Daniel isn’t speaking like someone trying to scare people - he’s speaking like someone who has spent years looking at the raw data, the incentives, the corporations, the governments and the ACTUAL speed of progress.

We discussed things like:

Why he chose to leave OpenAI and speak publicly.
Why he believes superintelligence could arrive in the next few years.
How AI could fundamentally change jobs, the economy and the military.
The race between companies and countries makes this so difficult to control.
What ordinary people should actually do right now to prepare.
Whether there is still a way to prevent this from going badly.

I use AI every day. I invest in companies using AI. I can see the extraordinary upside of this technology and it is going to continue becoming a massive part of all of our lives and careers. The question is whether we understand it quickly enough to make better decisions.

There was a moment towards the end of this conversation where I told Daniel that I could see how much he cares about this. And I meant that. Care itself makes others feel care.

If you're trying to understand where AI is really heading, I highly recommend watching or listening to this episode.

See 3 related tweets

@Ric_RTP: This OpenAI insider just exposed the AI labs' plan to fire every human on Earth.

And staff inside A...

@DKokotajlo: RT @StevenBartlett: What would you do if the technology we already use every day was about to become...
@heynavtoor: 10 AI podcast episodes that are smarter than most courses.

And every one is free.

Andrej Karpat...

14. MTSlive (Group Score: 124.6 | Individual: 38.4)

Cluster: 4 tweets | Engagement: 126 (Avg: 71) | Type: Tech

SITUATION EXPLAINED: Richard Sutton, the father of reinforcement learning, just launched a new lab.

The goal: a trillion-parameter agent that learns in real time on 20 watts.

• Sutton wrote The Bitter Lesson in 2019, its core claim: general methods that leverage computation beat hand-crafted approaches, by a large margin • His new lab, @oaklab_ai, builds on the "big world hypothesis": the world is too big for any model to pre-learn everything, so trying to is wasted effort by construction • Their "batch size one" algorithm updates from a single live experience as it arrives, instead of training on giant curated datasets... no massive pretraining run at all • Claimed result: multiple orders of magnitude less compute and energy than existing methods • The holy grail number: a trillion-parameter agent that learns and plans in real time on 20 watts... roughly what the human brain runs on • It's built for agentic tasks specifically, not as a general-purpose LLM replacement

@schisofrenia: "It's possible that we end up looking back at LLMs as an evolution of search... not to hate on LLMs, they're very useful, they're awesome. But eventually I think that's what it ends up looking like."\n\nQT @RichardSSutton: I can’t say enough good things about John Carmack @ID_AA_Carmack and his Keen Technologies. But now Khurram Javed @kjaved_ and I have broken away to start our own startup and pursue a slightly different path toward understanding intelligence. Like Keen (and like Ineffable) we at Oak Lab @oaklab_ai believe in reinforcement learning and that intelligence is created and maintained from run-time experience. But we think current deep learning methods are weak and inefficient, and need not more tweaks, but fundamentally new ideas and a thorough reworking before they can provide a solid foundation for achieving the more ambitious goals of AI.

See 3 related tweets

@eliebakouch: > We have a fairly complete roadmap to building animal-like intelligence

tired: AGI (Artificial ...

@RichardSSutton: I can’t say enough good things about John Carmack @ID_AA_Carmack and his Keen Technologies. But now ...
@MLStreetTalk: Sutton and Carmack are parting ways.

They both:

"believe in reinforcement learning" "[believe...

15. hardmaru (Group Score: 118.6 | Individual: 45.9)

Cluster: 3 tweets | Engagement: 326 (Avg: 108) | Type: Tech

How do physical systems achieve collective intelligence and self-repair without a central brain?

A new paper published today in Nature Communications by my Sakana AI colleague Sebastian Risi (@risi1979), along with co-authors from IT University of Copenhagen and Autodesk Research, presents a beautiful realization of biologically inspired robotics: Smart Cellular Bricks.

The team built a system of physical 3D cubic units that can collectively infer their global shape and autonomously guide their own damage recovery using purely local interactions.

Here is a deep dive into the paper’s key contributions:

1/ Neural Cellular Automata-based Architecture: Modular robots usually rely on central processors. This system flips that paradigm. Every block independently runs the exact same neural network on local microcontrollers. With no master plan or global coordinates, they communicate only with immediate neighbors. By passing continuous state vectors, hundreds of bricks achieve global consensus on their shape in under 3 minutes.

2/ Emergent Biological Morphogens: How does a block know it is part of a chair, not a table? The network’s internal memory automatically learns to establish continuous gradients across the structure. This beautifully mirrors how biological morphogens give positional info to developing cells. The bricks naturally form left-right, radial, and head-to-tail axes to align their identity.

3/ Performance and Generalization: Validated in large-scale simulations, the networks transferred seamlessly to nearly 200 physical hardware bricks, achieving a 100% convergence rate. Instead of rigid template-matching, the system infers broad categories. Even when tested on unseen variations, like an asymmetric table with five random legs, the collective correctly classified the structure.

4/ Fault Tolerance and Autonomous Damage Recovery: Hardware fails in the real world. This system easily tolerates up to 15% module failure without losing accuracy. By predicting spatial damage directions, the cells pinpointed missing components with 95% accuracy. They actively use these local signals to guide a self-repair process, regenerating back into the intended morphology.

I believe this is a significant piece of research, bridging collective intelligence and Physical AI.

This work represents the first successful physical realization of large-scale, decentralized 3D self-recognition and damage detection. By moving away from centralized control, this architecture paves the way for highly adaptive smart materials and resilient robotics that can survive and repair themselves.

Read the full open-access paper: https://t.co/friVEdvfT6

Congratulations to the team on this achievement!\n\nQT @SakanaAILabs: We are pleased to share our latest research, now published in Nature Communications: “Smart Cellular Bricks: Physical Modules That Recognize Their Own Shape and Repair Themselves.”

Blog: https://t.co/FMkbCv8mpP Paper: https://t.co/rxzbpU3zTu

A long-running theme in our work is collective intelligence: the idea that sophisticated, robust behavior can emerge from many simple parts following local rules, with no central controller, as it does in a colony, a tissue, or a brain. We had mostly studied this in software and simulation. So this time we asked a simple question. Do the same decentralized principles hold up in the physical world, where communication is noisy and modules fail?

To find out, we built a collection of simple cubic bricks. Each brick runs the same small neural network and talks only to the bricks it is physically connected to. No brick is told its position, or which shape it is part of.

Yet from these purely local exchanges, the collective converges on the correct global shape, locates where modules are missing or damaged, and can even guide its own repair, inspired by how living tissue self-organizes and regenerates after injury.

For us, this is a first step in a broader direction: taking the principles of collective intelligence we have studied in software and letting them emerge, decentralized and robust, in the physical world. In the future, we imagine smart materials that let structures sense and report damage on their own, and LEGO-like systems that recognize their own configuration and adapt in real time, pointing toward environments that are more robust, adaptive, and regenerative.

This work is a collaboration between Sakana AI, IT University of Copenhagen and Autodesk.

See 2 related tweets

@SakanaAILabs: We are pleased to share our latest research, now published in Nature Communications: “Smart Cellular...
@SakanaAILabs: RT @hardmaru: How do physical systems achieve collective intelligence and self-repair without a cent...

16. jxnlco (Group Score: 111.5 | Individual: 34.1)

Cluster: 4 tweets | Engagement: 890 (Avg: 259) | Type: Tech

RT @thsottiaux: Added a banked reset to 500k users of ChatGPT Work and Codex.

What’s been happening:

Just released the ability to use a banked reset from web and mobile. Before today you could only do it in the desktop app
Had an issue where less than 10% of users who used a banked reset and it didn’t actually reset. This was during a 2 hour window.
It was hard to find all the exact users where it didn’t apply, so instead granted a banked reset to everyone who pressed the reset button in that 2 hour window. This also gives us the opportunity to validate the new infra ahead of tomorrow

Tomorrow we will celebrate our 7M active users milestone and grant the first banked reset across all of our ChatGPT Work and Codex users.

We will also release a bunch of updates to the desktop app, addressing a ton of your feedback.

Talk soon

See 3 related tweets

@testingcatalog: ICYMI 👀: Users can now reset ChatGPT and Codex usage limits on both web and mobile. A new option is ...
@kimmonismus: Another reset and a banked reset on top of that. Outstanding community work; kudos.

OpenAI is spen...

@sama: we love our users\n\nQT @thsottiaux: Thank you to the 7M active users who are now using Codex and Ch...

17. pmarca (Group Score: 105.4 | Individual: 30.7)

Cluster: 4 tweets | Engagement: 152 (Avg: 745) | Type: Tech

Many people are saying.\n\nQT @deredleritt3r: In late November 2025, AI agentic harnesses became highly useful, and enterprise adoption exploded. For example, based on BTOS data, the percentage of large (250+ employees) U.S. enterprises using AI grew from ~25% to 37% between November 2025 and May 2026.

And the unemployment rate... hasn't moved. For workers aged 20+ it was:

September 2024 (first reasoning model): 3.6% November 2025: 4.1% July 2026: 3.8%

We see the same thing for workers aged 20-24:

September 2024: 7.0% November 2025: 8.3% July 2026: 7.1%

This is still very early data, but perhaps some aggressive assumptions regarding AI's impact on the labor market should be tempered a bit?

See 3 related tweets

@pmarca: Many people are saying.\n\nQT @jpatel41: I’ve been saying this for months, and the early data is beg...
@pmarca: RT @deredleritt3r: In late November 2025, AI agentic harnesses became highly useful, and enterprise ...
@PwC: Since 2022, the companies most exposed to AI have seen 40% higher productivity growth, alongside fas...

18. cgtwts (Group Score: 103.3 | Individual: 28.9)

Cluster: 5 tweets | Engagement: 69 (Avg: 76) | Type: Tech

what an insane career arc:

> built 2 unicorns > made partner at YC > now an MTS at Anthropic https://t.co/ichM6TzS3I\n\nQT @t_blom: Personal update: I'm taking a leave of absence from YC to join Anthropic. I'll be working with @NotTomBrown on the compute team.

Powerful AI has the potential to improve the life of every human on earth and, as we enter the early stages of recursive self-improvement, availability of compute becomes one of the most important issues to solve.

I'm excited to get started 🚀

See 4 related tweets

@ycombinator: RT @t_blom: Personal update: I'm taking a leave of absence from YC to join Anthropic. I'll be workin...
@kimmonismus: More and more, people are beginning to seriously pursue the transition toward recursive self-improve...
@SebJohnsonUK: Who ISN’T joining Anthropic\n\nQT @t_blom: Personal update: I'm taking a leave of absence from YC to...
@SebJohnsonUK: Anthropic's "Members of Technical Staff" are insanely GOAT-ed https://t.co/Ik0zpidfoI\n\nQT @t_blom:...

19. 0x_kaize (Group Score: 102.8 | Individual: 34.2)

Cluster: 4 tweets | Engagement: 56 (Avg: 94) | Type: Tech

Free AI API KEY Grok 4.5

The service constantly provides access to new AIs that are gaining popularity, and here we go again - only this time, it's Grok 4.5.

Here's the setup:

1/ Go to: https://t.co/5QVDDxkUyJ 2/ Sign up 3/ Go to Models tab 4/ Choose: xAI Grok 4.5

While everyone else is paying for APIs, you can get them completely FREE.

But you'd better hurry - access will be blocked again soon.

By the way, if you're looking for other models, I highly recommend this service, which I've been using since the day it launched.\n\nQT @0x_kaize: Just in July alone, this much has dropped:

Fable 5 - the smartest AI model in history Opus 5 - Fable-level intelligence at an accessible price Grok 4.5 - Opus-class performance, but cheaper. Elon is fully back GPT-5.6 - cheap and blazing fast. beats Fable on some benchmarks GPT-6 - Fable-level performance without worrying about limits ChatGPT Work - regular people can now get Claude Code-level agents inside ChatGPT

30 days from now your workflow will look nothing like it does today.

Dramatically less effort & overwhelmingly higher output.

And almost all of these models go live on Creao AI before anywhere else, the team tracks every release and ships new models into the product within a day.

See 3 related tweets

@cb_doge: RT @cb_doge: On this day in 2023, Elon Musk officially announced xAI.

A lot has happened since then...

@startupideaspod: Grok 4.5 made me realize something about AI agents nobody talks about:

Speed unlocks different way ...

@RoundtableSpace: GROK 4.5 IS BEING PRAISED BY SOME USERS AS ONE OF THE MOST NEUTRAL AND TRUTH-SEEKING AI MODELS YET h...

20. IPONewsroom_ (Group Score: 100.8 | Individual: 20.1)

Cluster: 6 tweets | Engagement: 27 (Avg: 41) | Type: Tech

European defense startup Helsing just said today that it has raised $1.8 billion in a funding round that values the German 🇩🇪 firm at$ 18 billion

Helsing bills itself as a hardware and software platform for defense. It makes drones and underwater surveillance weapons, and it builds AI and autonomous software to power these military applications - CNBC

See 5 related tweets

@Techmeme: Munich-based defense tech startup Helsing raised a $1.8B Series E from Dragoneer, Iconiq, and others...
@generalcatalyst: Congratulations to @HelsingAI on its $1.8B Series E.

Helsing is one of the most consequential AI co...

@CNBC: European defense startup Helsing said Monday it has raised $1.8 billion in a funding round that valu...
@lightspeedvp: We’re proud to participate in @HelsingAI’s $1.8B Series E, bringing the company to a total valuation...
@business: Helsing has raised a financing round at a valuation of $18 billion, cementing its status as Europe’s...