2026-07-14

The startup, PrismML, said it has shrunk down Qwen 3.6 , an open-source large language model developed by Chinese internet giant Alibaba , to run on an iPhone 17 Pro. The model has 27 billion parameters , which are roughly similar to the synapses in a brain and can help determine the complexity of the data a model can process. In contrast, most models that run on mobile phones have only a few billion parameters active at a time. The largest AI models, which can measure in the trillions of parame

Read a long piece on the future of LLM reasoning that makes a provocative claim: Chain of Thought is a useful hack but we've started to confuse a readable trace with the actual computation. All in all, "generating text is not the same as thinking." There are two practical problems here: Faithfulness: CoT style traces can decouple from what the model actually "did." u can get plausible steps with a wrong answer, or messy steps with a right answer (so the trace isnt a reliable audit trail) Systems

Hey everyone, I recently switched from DS4 Flash to Qwen3.5-122B on my M3 Ultra Mac Studio for long-context agentic coding. While the model fit better, I hit a wall where follow-up messages took 3-5 minutes to start generating (cold fills) despite having a "warm" context. Turns out the issue wasn't the model, but three specific bugs in my serving stack (qMLX fork of rapid-mlx): Prompt Instability: A unique message ID in the system prompt broke byte-exact KV cache matching, forcing a full re-comp

Apple’s revamped Siri is more than a voice assistant; it’s now the backbone of the iPhone user experience. You can try it now through the iOS 27 public beta.

Apple’s trade secrets lawsuit against OpenAI contains allegations that range from employees joking about unauthorized access to Apple’s systems to claims that job candidates were asked to bring Apple hardware to interviews. Here are the complaint’s most eye-catching claims.

Angelo Martino helped scam ransomware victims out of over $75 million, officials said.

arXiv:2607.09175v1 Announce Type: new Abstract: Deployed LLM agents rely on agentic context, the model-external textual control content assembled by an operational harness. In this work, the mutable component of that context is a persistent system-level instruction that is updated from operational experience while the model, tools, and harness remain fixed. Over long evolution horizons, flat-text maintenance makes verification increasingly difficult as accumulated instructions grow and interact.

arXiv:2607.09076v1 Announce Type: new Abstract: Cyberattacks on operational technology are increasingly causing costly downtime and physical damage, exposing the limitations of traditional rule-based monitoring in industrial IoT environments. While Large Language Models (LLMs) have strong semantic reasoning abilities to assist in decision support, their hallucinatory nature presents unacceptable safety liabilities for closed-loop control. This paper introduces a neuro-agentic control framework,

I only spent 2 hours making the bios accept the dual gpus, only 5 hours configuring VLLM to run deepseek v4 flash dspark, but totally worth it. I truly believe in the near future we will have to rely on ourselves.

Apple has just released public betas for iOS 27 and other major OS updates that are set to publicly launch this fall. The big new feature this year is Siri AI, the delayed AI-powered revamp to Siri. It actually works - which is big praise! - though it keeps things brief. Other betas available now […]

iOS 27 escaped the developer world today with the launch of the first public beta. I've been testing the new operating system since early June, looking for quirks and seeing if it can live up to the hype Apple promised in the keynote. This year's iOS upgrades are what one might call a Snow Leopard […]

Siri has been on the Apple Watch since day one, though I'm usually hard-pressed to find people who actually make good use of it. It's kind of just… been there - mostly as a way to set timers when my hands are full. But after playing around with the watchOS 27 developer beta, I get […]

What does a world of total user-aligned AI actually look like?

Experts explain how they work, what they can do, and what's still unsettled.

arXiv:2607.09195v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly expected to play a central role in AI-driven scientific discovery. Equipped with broad knowledge, flexible reasoning, and tool use, they have the potential to autonomously explore and solve scientific problems by repeatedly proposing hypotheses, testing them, and revising their beliefs in the light of the evidence. In current agents, however, these hypotheses, tests, and belief updates are buried i

arXiv:2607.08773v1 Announce Type: new Abstract: In this work we present a rigorous theoretical framework to a foundational problem of AI safety, namely adversarial robustness. In particular, we show that the adversarial robustness problem can be reduced to a lattice traversal problem. Each element of this lattice corresponds to an interval, i.e., an axis-aligned hyper-rectangle, containing an input point $\mathbf{x}$. Consider a multilayered perceptron classifier (MLP). An interval $I$ constitut

" Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity " This paper was accepted to ICML this year. Its main idea is a very simple prompt-engineering trick: "changing the prompt this way led to more diverse sampling". Naturally, it is difficult to provide a rigorous theoretical analysis for something like this. Even if it works, I’m not sure this kind of prompt engineering belongs at a top-tier machine learning conference. Some people seems to call this kind of work “moder

Am I missing something? It seems like some people think distillation is magic and will raise the quality of output above what the base model is actually capable of. It's especially weird to me to see all these Fable fine-tunes, because as far as I understand it, they miss the fact that the reasoning traces you get from Anthropic's models are completely different from the actual chain of thought the model outputs, which makes it pretty much a guarantee that the result will be worse than before.

This weekend I got to what I consider a shareable state with this experiment to create little characters that can run commands to do funny things. I want something to play around with local AI on my laptop which is a HP Omen from like 6 or 7 years ago and has a 2060. The idea is that since it's a very small model, the NPCs are dumb. So I imagine I can lean into that to with the concept of dumb NPCs doing silly things and make interesting little demos and hopefully find a way to get other people

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Sperm donors need limits, says a European fertility group Ties van der Meer doesn’t know how many siblings he has. The 47-year-old was conceived at a private fertility clinic using sperm…

Recovering high-quality video from sparse event streams is a challenging task. Regression methods often blur textures, while existing generative models struggle with long-term stability. We propose LongE2V, a novel approach that leverages pre-trained video diffusion priors to jointly handle event-based video reconstruction, prediction, and frame interpolation. By fine-tuning a foundational video model, our approach achieves high data efficiency and superior perceptual quality. We introduce Autor

arXiv:2607.09142v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in online medical consultation, yet existing benchmarks remain poorly aligned with real clinical practice. Many rely on synthetic conversations or patient simulators, omit patient-uploaded medical images, or evaluate open-ended clinical responses using multiple-choice or lexical-overlap metrics that poorly reflect clinical quality. We introduce \textbf{MedRealMM}, a large-scale benchmark for mu

arXiv:2607.09330v1 Announce Type: new Abstract: Embodied agent teams powered by heterogeneous large language models (LLMs) are being widely deployed in physical artificial intelligence such as smart factories, warehouses, and service robotics. To enable collaboration among such an agent team, efficient coordination mechanisms that operate reliably under limited network resources are required. However, existing heterogeneous LLM-agent coordination frameworks that rely on multi-round natural-langu

Big goals are hard to achieve all at once; breaking them into small steps is wiser. We present Trust Region Policy Distillation (TOP-D), which transforms the notoriously unstable, high-variance On-Policy Distillation (OPD) into a stable training paradigm by dynamically constructing a proximal teacher. Theoretically, we establish a rigorous framework demonstrating that TOP-D inherently controls gradient variance. By providing a formal global convergence analysis alongside a monotonic improvement

Release: shot-scraper 1.11 Some minor improvements, mainly around command option consistency and making the server: mechanism used by both shot-scraper video and shot-scraper multi work if the server takes longer than a second to start serving traffic. server: processes used by shot-scraper multi and shot-scraper video now wait up to 30 seconds for the target URL to accept connections, polling for port availability and replacing the previous fixed one-second delay. #197 The shot-scraper , pdf ,

Generalist robot manipulation policies have advanced rapidly, yet existing benchmarks remain limited in systematically evaluating their capabilities. Many rely on simple, short-horizon, or skill-narrow tasks with limited capability coverage, and are often conducted only in simulation or only in the real world. Simulation enables scalable feedback but misses physical deployment challenges, while real-world evaluation is costly, time-consuming, and difficult to reproduce. We introduce RoboDojo, a