AI News

The Palo Alto startup says it has landed multiple seven-figure enterprise contracts within months of launch.

Model-driven synthetic test data for CI/CD and analytics - deterministic, privacy-preserving, and domain-aware. Includes Python APIs, XML pipelines, and MCP/IDE integration to orchestrate realistic datasets for finance, healthcare, and other regulated environments. - rapiddweller/datamimic

E-ink bird frame for Raspberry Pi - real-time bird detection by audio, fully local AI, rendered as real, hand-cut 1800s bird illustrations. - arnegiacomo/fugleramme

Nebius previews NVIDIA Vera Rubin NVL72 in MLPerf® Inference v6.1 and posts first-place DeepSeek R1 results on the full-rack Nebius GB300 NVL72 system.

Explore how AMD connects data center, cloud, and local AI to enable secure, cost-aware agentic workflows across enterprise PCs.

The scaling laws hold that a language model grows more capable with more parameters and more training data, and Mixture-of-Experts (MoE) architectures have ridden these laws to remarkable results, activating only a fraction of an enormous stored parameter bank for each token. That success is built on static pretraining data. A deployed model faces a different world, where much of the data that would make it more useful is not in its training set but in the live interaction it is currently handling, such as the facts a user supplies or the corrections they give. A conventional model cannot learn from this data, because its weights are frozen after training. Instead, the knowledge and behaviour supplied at run time are placed in the prompt, by retrieval or instruction, and re-read on every request only to be discarded once the request ends. We ask how an architecture could learn from live interaction by writing it into its weights. Taking inspiration from MoE, we propose the \textbf{Infinite-Parameter LLM}. A compact hypernetwork turns the data given at run time into a low-rank modulation of a shared base network, so the feed-forward weights are generated from live data rather than stored in a fixed bank. Where prior weight generators read the context once and freeze, we carry a Bayesian belief over the generator's latent code and update it online, so the effective weight is re-derived from that evolving belief as the session proceeds rather than fixed after one read. The stored footprint stays fixed, yet the weights the model can compile are effectively infinite. For the knowledge and behaviour supplied at run time, carrying them in the weights rather than the prompt is amortized in compute, frees the context window, persists across turns, and can generalise better than in-context use. We specify an evaluation protocol that tests exactly this against in-context learning and retrieval.

Scientific code repositories encode decades of human knowledge in executable models, methods, and tools. Yet fragmented toolchains, implicit domain conventions, and specialized correctness criteria make this knowledge difficult to convert into reliable learning experience-a challenge we call the scientific experience bottleneck. We introduce ScienceIDE, infrastructure for turning the world's scientific code into programmable environments for scientific agents. Guided by expert-defined scientific cases and acceptance criteria, agents transform repositories into executable environments that support task generation, execution, and scientific verification. These environments provide a shared foundation for supervised fine-tuning, reinforcement learning, and evaluation. Using verified interaction trajectories, we train PhAI-IDE-72B, PhAI-IDE-9B, and PhAI-IDE-4B. The model family shows gains in held-out scientific-code repair and across selected general-purpose benchmarks in code, reasoning, and knowledge, providing evidence of positive transfer from scientific experience to broader capabilities. ScienceIDE lays the foundation for an integrated workspace for agent learning and scientific practice, making humanity's scientific software a shared substrate for developing scientific intelligence. Code: https://github.com/aitofound/ScienceIDE

Direct preference alignment methods are widely used to align large language models (LLMs) with human preferences because of their computational and memory efficiency. However, likelihood displacement motivates alternative ways to extract information from preference pairs with small likelihood margins. In this paper, we propose and analyze Comparison-based Preference Optimization (ComPO), a zeroth-order alignment method based on comparison oracles. ComPO extracts directional information from these pairs without directly optimizing a differentiable preference loss on them. We establish a convergence guarantee for its basic offline scheme under smoothness, gradient sparsity, and compatibility between the oracle and a latent objective. We further introduce online ComPO, which retains the offline comparison mechanism and uses unlabeled policy generations for reverse-KL control relative to a reference policy. Following the coverage perspective of preference fine-tuning, we establish a performance guarantee for a basic constrained scheme under local coverage and in-distribution pairwise reward accuracy. Experiments on Mistral, Llama, Gemma-2, Qwen3, and Gemma-3 models demonstrate improvements over existing direct alignment methods, including length-controlled win rates, with pair-level diagnostics providing evidence consistent with mitigating likelihood displacement.

Open-weight model developer Arcee AI reaches $1B-plus valuation with new funding - SiliconANGLE

We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI.

The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer inv…

L’intesa triennale della città con lo spin-off del Politecnico Niulinx punta a sperimentare robotaxi, navette e trasporto a chiamata.

Create high-converting AI image & video ads with Creatify. Generate unlimited advertising variations in minutes and let AI handle the rest. Try for free!

Venture capital is not moving evenly today. It is clustering around businesses that control a scarce input, remove an expensive operational bottleneck, or sit close enough to a measurable economic outcome that investors can underwrite something more tangible than AI adoption curves. The strongest examples range from Impulse Space’s orbital infrastructure and Arcee AI’s open-weight

Capsule packages your entire app — UI, data, and assets — into a single portable file. No cloud. No accounts. Just share it.

Sierra’s multimodal agents bring voice, text, and visuals into the same conversation, so customers get the best of each medium without having to pick just one.

Nvidia's Les Karpas explains how robots are waiting for their ChatGPT moment, only at TechCrunch Disrupt 2026. Register now to save up to $200 on your pass.

At the World Humanoid Robot Games in Beijing, a humanoid robot named Tiangong Ultra ran the 100-meter sprint in a staggering 8.64 seconds, shattering

Mixture-of-Experts (MoE) language models suffer from large parameter counts, which create a significant memory bottleneck. Expert pruning is the most direct approach for reducing this parameter count, yet existing methods make pruning decisions for each expert independently, and assume experts' contributions are purely additive. In reality, expert usage in MoEs is inherently cooperative. We derive HOPE (Higher-Order Pruning of Experts), a second-order pruning objective which provably minimizes an upper bound on the error resulting from pruning. We show that REAP (a state-of-the-art first-order pruning method) is a special case of HOPE where interaction terms are ignored. Across three frontier MoE models (up to 122B parameters), two distinct calibration sets, and multiple benchmarks (including math, instruction following, coding, and an agentic suite), we demonstrate that HOPE produces better pruning decisions than existing methods, and its advantage is most pronounced at high pruning rates and on challenging agentic workloads. At 50% pruning, HOPE outperforms all baselines and achieves an average rank of 1.58 out of 5 methods (versus 2.42 for the next-best method, REAP), with gains of up to +6.1% on agentic coding. Over all conditions, HOPE again achieves the best average rank and surpasses every other method in the majority of head-to-head comparisons. By preserving cooperative expert structure that first-order methods ignore, HOPE enables aggressive compression with minimal degradation, particularly on complex tasks where diverse expert combinations are invoked over long sequences.

Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded. Current vision-language models (VLMs) can generate fluent and detailed image captions, but reliably associating them with image pixels remains challenging. Existing methods that combine dense captioning with pixel-level grounding often produce either incomplete descriptions or inaccurate segmentation masks. We study this problem through panoptic grounded captioning, a task that requires a VLM to describe both foreground objects and background regions while grounding each referring phrase with pixel-level masks. We make three contributions. First, we introduce PanoCaps, a human-annotated benchmark constructed from panoptic segmentation datasets. It provides dense captions with near-complete pixel coverage and image-text alignments at the entity level, supporting both training and evaluation. We further propose a phrase-mask matching protocol and a generalized Panoptic Quality (gPQ) metric that jointly evaluates textual and mask agreement. Second, we formulate phrase grounding as selection from a phrase-conditioned pool of mask proposals and introduce PANORAMA, a VLM that conditions a pretrained segmenter on contextualized phrase representations to obtain candidate masks and learns to select those corresponding to each phrase. Training this interface jointly with caption generation enables PANORAMA to produce high-quality masks while allowing each phrase to refer to a single region or multiple instances. Third, PANORAMA achieves the best overall grounding on PanoCaps and matches or exceeds specialized models across several pixel-level grounding tasks. Experiments show that our method produces precise entity-level segmentations while maintaining detailed, mask-consistent captions. Code, data and models are available at https://www.di.ens.fr/willow/research/panorama/.

Banger report from Salesforce.

Pretty interesting to see more of these custom enterprise models.

Salesforce trained the enterprise agent model from the same files it uses to configure agents.

Koa starts from the open-weight Nemotron-3-Super-120B.

Salesforce takes Agent Script specifications, th…

We just launched the DeepMind Institute, a platform for researchers across @GoogleDeepMind and the wider research community to publish and debate how increasingly capable AI should be built, governed and used.

First 4 essays on reasoning transparency, economic policy and human flourishing at http…

Their startup, Artificial Intelligence Underwriting Company (AIUC) has raised $40 million in a Series A round led by Ribbit Capital, with participation from First Harmonic.

Exein raccoglie 270 milioni e sale a 1,7 miliardi: la cybersecurity italiana punta sulla Physical AI e sulla sicurezza dei dispositivi

Web Search Agents are expert web crawling and research agents for your specific domain (company enrichment, regulations research, etc.). They self-learn your use case to go deeper into the sources that matter most to you, giving your AI deeper and more relevant web context. To get started, give your AI this link: https://docs.nimbleway.com/agent-onboarding.md