Since Core War is a Turing-complete environment where code and data share the same address space, the “warriors” Redcode programs produced by LLMs, selected through natural selection quickly exhibit rather chaotic self-modifying code dynamics. Full Paper: https://t.co/mynZXzXYOO https://t.co/b7o5forgWU

Digital Red Queen: Adversarial Program Evolution in Core War with LLMs https://t.co/zHNcl2xEtC TL;DR. With MIT, we propose a method for self-evolving assembly code. The process iterates code evolution through battles in Core War. By training new warriors against an ever-changing lineup of opponents rather than a static optimization objective, we can produce warriors that are both robust and general-purpose.

1/🧵 We prompted production LLMs with a short prefix of a book and asked them to complete the rest. How much of the book did they return? For Harry Potter and the Sorcerer’s Stone: (jailbroken) Claude 3.7 Sonnet→95.8%, GPT-4.1→4.0% (not jailbroken) Gemini 2.5 Pro→76.8%, Grok 3→70.3% Read on more details:

AI sovereignty vs. market concentration—can Europe have both? New data shows the EU cloud market shrinking to fewer providers, right as AI demands explode. But there's a path: distributed hybrid architecture lets nations capture economic value without sacrificing control. Is digital sovereignty the next competitive battleground? AI’s future: Plotting a path to competitiveness and digital sovereignty

Trillions are flowing into AI on the belief that progress toward AGI will keep accelerating. That belief may be more fragile than markets assume. Yoshua Bengio warns that AGI development could hit a wall, even as investment continues to surge. If expectations outrun reality, the risk won’t be technical alone. It could turn financial. https://t.co/sMKCVFIL6j

AI is already moving from assistant to operator. Some founders are building companies with AI as part of the core team, not just a tool. Yesim Saydan trained more than a dozen custom GPTs to run her solo consultancy, including one inspired by Steve Jobs to act as a mentor and decision sparring partner. AI may feel scary, but this is the shift to watch. Not AI replacing founders, but founders scaling their thinking. For some, AI is no longer optional. It is the edge. https://t.co/jVEw7atlyB @businessinsider

AI is quietly crossing another threshold. Models are no longer just assisting with math. They are beginning to solve high-level problems end-to-end. Neel Somani tested a new OpenAI model on an advanced math problem. After 15 minutes of reasoning, it produced a complete solution that held up under formal verification. This is not about speed or convenience anymore. When AI can reason deeply and check its own work, the boundary between tool and collaborator starts to blur. https://t.co/DR5GDpTL3f @russellbrandom @techcrunch

OpenAI is taking a step it long resisted. Ads are coming to ChatGPT for free and lower-tier users, while Pro and Enterprise stay ad-free. OpenAI frames this as a necessity, not a pivot. With massive compute costs and only a small share of users paying, ads are meant to fund access while subscriptions fund capability. The company insists ads won’t influence answers or mine conversations. The real test isn’t whether ads appear. It’s whether OpenAI can monetize attention without eroding trust. Can ChatGPT stay useful and credible once ads enter the interface? https://t.co/zizZB0w6Ho @techxplore_com

AI is pushing electricity grids to their limits. As compute demand explodes, energy is becoming the real constraint on AI scale. Many now argue that this pressure could kickstart a new nuclear renaissance. Few energy sources can deliver the constant, carbon-free baseload power that large AI data centers require. The question is no longer whether AI needs nuclear. It’s whether nuclear can move fast enough to meet AI’s timeline. https://t.co/rsWIrzJgG6 @UN_News_Centre @UN

Wall Street banks are aligning on one thing. AI will change head count. They just don’t agree how. Jamie Dimon, Jane Fraser, and David Solomon all highlight AI-driven efficiency, but diverge on outcomes. Some expect fewer roles. Others see jobs shifting, not disappearing. The pattern is clear. AI won’t cut evenly. Routine work shrinks, high-judgment roles expand. The real question isn’t job loss. It’s whether banks can reskill fast enough before the org chart rewrites itself. https://t.co/RU0FLSkg4U @businessinsider @nicollsanddimes @ATecotzky

AGI Is Not Multimodal "In projecting language back as the model for thought, we lose sight of the tacit embodied understanding that undergirds our intelligence." –Terry Winograd, Professor Emeritus in the Computer Science Department at @Stanford https://t.co/x0NPIKzqrm by @superspeeg v/ @gradientpub Cc @DeepLearn007 @SpirosMargaris @aure79lien @YvesMulkers @akwyz @sallyeaves @gvalan

One of the greatest thinkers of all time reminds us that our “constant pursuit of immediate achievement” can massively distracts us from living a meaningful and happy life. https://t.co/obDSIECzTm

https://t.co/VWUzsM9yGc

https://t.co/VWUzsM9yGc

Peter Lynch on how to understand stocks and investing… I often need a reminder of this. Peter Lynch is an investing great. https://t.co/NMccIxtzBc

Rushing into #AIAgents without clear use cases, guardrails or data readiness risks wasted spend, security gaps and backlash instead of business gains.⚠️🏢 #AI #AINews #Business #Tech #ShareForSuccess https://t.co/5QhX8pFVOG

AI in medicine is often framed as either a breakthrough or a threat. In reality, it’s neither. It’s a tool, and its impact depends on how it’s used on the ground. The episode The Pitt gets a few important things right. AI can help with triage, pattern recognition, and decision support when clinicians remain in charge, which is close to how hospitals are actually experimenting with it today. Where it slips is in the numbers. Lines like “98 percent accuracy” sound impressive but miss the point. Medicine is contextual and uncertain by nature. The real question isn’t how accurate AI can be in theory, but how much responsibility we’re willing to hand over when real patients are involved. https://t.co/LLKJPGiT1T @mashable

In an age dominated by AI, acceleration, and overlapping crises, something analog is quietly gaining value again. Real-life community. Amy Thunig-McGregor argues that while the world feels increasingly unstable, grounding ourselves in home, care, and local community can be a source of strength and clarity. As technology scales intelligence, it does not replace belonging. In the AI era, community may become not just precious, but essential. https://t.co/1mAmdwjFGz @ConversationUS @ConversationEDU

1X World Model | From Video to Action: A New Way Robots Learn Blog: https://t.co/1sPpUJBcrF 1X describes and shows initial results for a new potential way of learning robot policy using video generation based world modeling, compared to VLA which is based on VLM. - How it works: at inference time, the system receives a text prompt and a starting frame. The World Model rolls out the intended future image frames, the Inverse Dynamics Model extracts the trajectory, and the robot executes the sequence in the real world. - The World Model backbone: A text-conditioned diffusion model trained on web-scale video, mid-trained on 900 hours of egocentric human data of first-person manipulation tasks for capturing general manipulation behaviors, and fine-tuned on 70 hours of NEO-specific sensorimotor logs for adapting to NEO’s visual appearance and kinematics. - The Inverse Dynamics Model: similar to architecure used in DreamGen, and trained on 400 hours of robot data on random play and motions. - Results: The model can generate videos aligning well with real-world execution, and the robot can perform object grasping, manipulation with some degree of generalization. - Current limitations: The pipeline latency is high and it’s not lose-loop. Currently the WM takes 11 second to generate 5 second video on a multi-GPU server and IDM takes another 1 second to extract actions.

1X World Model | From Video to Action: A New Way Robots Learn Blog: https://t.co/1sPpUJBcrF 1X describes and shows initial results for a new potential way of learning robot policy using video generation based world modeling, compared to VLA which is based on VLM. - How it works: at inference time, the system receives a text prompt and a starting frame. The World Model rolls out the intended future image frames, the Inverse Dynamics Model extracts the trajectory, and the robot executes the sequence in the real world. - The World Model backbone: A text-conditioned diffusion model trained on web-scale video, mid-trained on 900 hours of egocentric human data of first-person manipulation tasks for capturing general manipulation behaviors, and fine-tuned on 70 hours of NEO-specific sensorimotor logs for adapting to NEO’s visual appearance and kinematics. - The Inverse Dynamics Model: similar to architecure used in DreamGen, and trained on 400 hours of robot data on random play and motions. - Results: The model can generate videos aligning well with real-world execution, and the robot can perform object grasping, manipulation with some degree of generalization. - Current limitations: The pipeline latency is high and it’s not lose-loop. Currently the WM takes 11 second to generate 5 second video on a multi-GPU server and IDM takes another 1 second to extract actions.

What an important post from @om about the blinding conformity of our times. And what we need to do to get more colorful again https://t.co/u7srwfpH5v

Today, we're introducing a working Loveable for video production. It does a video production teams work in minutes by writing, reserching, and editing like a human would. Hiring a video production team of upwork is now optional:

Mike Rowe: “We’ve been telling kids for 15 years to learn to code.” “Well, AI is coming for the coders.” “It’s not coming for the welders, the plumbers, the steamfitters, the pipefitters, the HVAC, or the electricians.” “In Aspen, I sat and listened to Larry Fink say we need 500,000 electricians in the next couple of years—not hyperbole.” “The BlueForge Alliance, who oversees our maritime industrial base—that’s 15,000 individual companies who are collectively charged with building and delivering nuclear-powered subs to the Navy … calls and says, we’re having a hell of a time finding tradespeople. Can you help?” “I said, I don’t know, man … how many do you need? He says, 140,000.” “These are our submarines. Things go hypersonic, a little sideways with China, Taiwan, our aircraft carriers are no longer the point of the spear. They’re vulnerable.” “Our submarines matter, and these guys have a pinch point because they can’t find welders and electricians to get them built.” “The automotive industry needs 80,000 collision repair and technicians.” “Energy, I don’t even know what the number is, I hear 300,000, I hear 500,000.” “There is a clear and present freakout going on right now. I’ve heard from six governors in the last six months. I’ve heard from the heads of major companies.”

PSA for people using amp. Huge if you use amp on mobile a lot like me! https://t.co/gMlI5bIBdR

Announcing hameval, made by @randal_olson It's a meme site 🤣 have fun! https://t.co/oQ1KWAjspT https://t.co/TxAscaqhJS

Love this transcript project by @noamseg AdviceArena: When product legends give contradictory advice, which wisdom wins? https://t.co/Ivim5bged1 https://t.co/eJG5jqHfjy

Woah! https://t.co/LMgMpT5PRj

@pamelafox It’s here , point Claude code at it and tell it to install the tools and guide you through using it etc https://t.co/NghIsGO0dG

@pamelafox Also @isaac_flath has a SaaS app for this https://t.co/qwkXEqHpzj and it’s really good if you want to jump right into that

It's easy to get lost when reading traces. An effective heuristic is to stop at the first (most upstream) error you find and move on when doing error analysis. You can then modify and curate traces for evals. BTW this is flashcard 9/12 (links in reply). https://t.co/4pPnCig4ez

This MCP is so dam useful I have conversations with friends over iMessage all the time. Its amazing to be able to yeet this into a coding agent or claude. Thanks to @mattt for making it https://t.co/ouHHg7QGQ6 https://t.co/esnAgEgy8V

There are three categories of evals - and we've found people get confused between these! The trick is to try to use code based assertions whenever you can get away with it (even by being a bit clever). https://t.co/BWcKNRfVVw