Nova Forge lets Amazon's customers train frontier models for different tasks--a potential breakthrough in making AI actually useful for businesses. Amazon has announced a new family of frontier artificial intelligence models--and a new way for customers to build frontier models of their own. The ecommerce giant announced the second generation of its Nova AI models at re:Invent, a company conference held in Las Vegas. The models are nowhere near as popular as those offered by rivals like OpenAI and Google, but Amazon's plan to make them highly customizable could see them gain traction with its cloud users. Amazon detailed two improved large language models, Nova Lite and Nova Pro; a new realtime voice model called Nova Sonic; and a more experimental model called Nova Omni that performs a simulated kind of reasoning using images, audio, and video as well as text.

Welcome back to, TIME's new twice-weekly newsletter about AI. If you're reading this in your browser, why not subscribe to have the next one delivered straight to your inbox? The three most valuable private companies in the U.S. have big reputations: OpenAI, SpaceX, and Anthropic. But the fourth, Databricks, flies a little more under the radar. This company, which is currently raising funds at a valuation of $134 billion according to reports this week, is a quiet workhorse of the AI revolution.

Any far-reaching new technology is always uneven in its adoption, but few have been more uneven than generative AI. That makes it hard to assess its likely impact on individual businesses, let alone on productivity across the economy as a whole. At one extreme, AI coding assistants have revolutionized the work of software developers. At the other extreme, most companies are seeing little if any benefit from their initial investments. That has provided fuel for the skeptics who maintain that--by its very nature as a probabilistic technology prone to hallucinating--generative AI will never have a deep impact on business. To students of tech history, though, the lack of immediate impact is normal.

'The biggest decision yet': Jared Kaplan on allowing AI to train itself Anthropic's chief scientist says AI autonomy could spark a beneficial'intelligence explosion' - or be the moment humans lose control Humanity will have to decide by 2030 whether to take the "ultimate risk" of letting artificial intelligence systems train themselves to become more powerful, one of the world's leading AI scientists has said. Jared Kaplan, the chief scientist and co-owner of the $180bn (£135bn) US startup Anthropic, said a choice was looming about how much autonomy the systems should be given to evolve. The move could trigger a beneficial "intelligence explosion" - or be the moment humans end up losing control. In an interview about the intensely competitive race to reach artificial general intelligence (AGI) - sometimes called superintelligence - Kaplan urged international governments and society to engage in what he called "the biggest decision". Anthropic is part of a pack of frontier AI companies including OpenAI, Google DeepMind, xAI, Meta and Chinese rivals led by DeepSeek, racing for AI dominance. Its widely used AI assistant, Claude, has become particularly popular among business customers.

Masayoshi Son, chairman and chief executive officer of SoftBank Group, speaks during the Future Investment Initiative (FII) Institute Priority Asia conference in Tokyo on Monday. SoftBank Group founder Masayoshi Son said he wouldn't have sold off Nvidia shares if his company had unlimited money to bankroll its next investments in artificial intelligence, which include a big bet on OpenAI. Son, addressing for the first time the surprise November disclosure that SoftBank had unloaded its entire stake in the world's most valuable company, also slammed talk of an AI investment bubble. The Japanese company simply needed to raise capital to fund projects including data center construction, he told a forum in Tokyo Monday. I don't want to sell a single share. I just had more need for money to invest in OpenAI" and other projects, Son said during the FII Priority Asia forum.

Safety-critical environments are inherently dynamic. Distribution shifts, emerging vulnerabilities, and evolving requirements demand continuous updates to machine learning models. Yet even benign parameter updates can have unintended consequences, such as catastrophic forgetting in classical models or alignment drift in foundation models. Existing heuristic approaches (e.g., regularization, parameter isolation) can mitigate these effects but cannot certify that updated models continue to satisfy required performance specifications. We address this problem by introducing a framework for provably safe model updates. Our approach first formalizes the problem as computing the largest locally invariant domain (LID): a connected region in parameter space where all points are certified to satisfy a given specification. While exact maximal LID computation is intractable, we show that relaxing the problem to parameterized abstract domains (orthotopes, zonotopes) yields a tractable primal-dual formulation. This enables efficient certification of updates - independent of the data or algorithm used - by projecting them onto the safe domain. Our formulation further allows computation of multiple approximately optimal LIDs, incorporation of regularization-inspired biases, and use of lookahead data buffers. Across continual learning and foundation model fine-tuning benchmarks, our method matches or exceeds heuristic baselines for avoiding forgetting while providing formal safety guarantees.

Post-training quantization (PTQ) aims to preserve model-level behavior; however, most methods focus on individual linear layers. Even recent extensions, such as QEP and LoaQ, which mitigate error propagation or target specific submodules, still rely on layer-wise formulations and fail to capture the behavior of larger submodules. We introduce Layer-Projected Coordinate Descent (LPCD), a unified framework that extends PTQ beyond layers by optimizing relaxed objectives across arbitrary submodules and projecting the solutions with layer-wise quantizers. LPCD generalizes existing methods and provides a principled approach to quantizing complex submodules while maintaining the efficiency and compatibility of layer-wise PTQ pipelines. Across diverse LLM architectures and bit-widths, LPCD-based submodule quantization consistently enhances both layer-wise PTQ methods and existing submodule approaches.

Foundation models, and in particular large language models, can generate highly informative responses, prompting growing interest in using these ''synthetic'' outputs as data in empirical research and decision-making. This paper introduces the idea of a foundation prior, which shows that model-generated outputs are not as real observations, but draws from the foundation prior induced prior predictive distribution. As such synthetic data reflects both the model's learned patterns and the user's subjective priors, expectations, and biases. We model the subjectivity of the generative process by making explicit the dependence of synthetic outputs on the user's anticipated data distribution, the prompt-engineering process, and the trust placed in the foundation model. We derive the foundation prior as an exponential-tilted, generalized Bayesian update of the user's primitive prior, where a trust parameter governs the weight assigned to synthetic data. We then show how synthetic data and the associated foundation prior can be incorporated into standard statistical and econometric workflows, and discuss their use in applications such as refining complex models, informing latent constructs, guiding experimental design, and augmenting random-coefficient and partially linear specifications. By treating generative outputs as structured, explicitly subjective priors rather than as empirical observations, the framework offers a principled way to harness foundation models in empirical work while avoiding the conflation of synthetic ''facts'' with real data.

Large language model (LLM) reinforcement learning has increasingly relied on group-based policy optimization frameworks, such as GRPO and GSPO, to achieve stable fine-tuning at scale. However, a fundamental trade-off persists between optimization granularity and training stability. While GSPO improves robustness via sequence-level optimization, its monolithic treatment of sequences introduces severe inefficiencies: its conservative clipping mechanism indiscriminately discards valid training samples-a phenomenon we term gradient underutilization-and its uniform credit assignment fails to capture the heterogeneous contributions of critical reasoning steps. In this work, we propose Entropy Importance Sampling Policy Optimization (ESPO), a novel framework that reconciles fine-grained control with training stability. ESPO decomposes sequences into groups based on predictive entropy, enabling (1) Entropy-driven Importance Sampling to capture intra-sequence heterogeneity, and (2) Entropy-adaptive Clipping to dynamically allocate trust regions based on model uncertainty. Extensive experiments on mathematical reasoning benchmarks demonstrate that ESPO not only accelerates convergence but also achieves state-of-the-art performance, notably improving accuracy on the challenging HMMT benchmark from 4.4% to 13.13%.

Tabular data drive most real-world machine learning applications, yet building general-purpose models for them remains difficult. Mixed numeric and categorical fields, weak feature structure, and limited labeled data make scaling and generalization challenging. To this end, we introduce Orion-Bix, a tabular foundation model that combines biaxial attention with meta-learned in-context reasoning for few-shot tabular learning. Its encoder alternates standard, grouped, hierarchical, and relational attention, fusing their outputs through multi-CLS summarization to capture both local and global dependencies efficiently. A label-aware ICL head adapts on the fly and scales to large label spaces via hierarchical decision routing. Meta-trained on synthetically generated, structurally diverse tables with causal priors, Orion-Bix learns transferable inductive biases across heterogeneous data. Delivered as a scikit-learn compatible foundation model, it outperforms gradient-boosting baselines and remains competitive with state-of-the-art tabular foundation models on public benchmarks, showing that biaxial attention with episodic meta-training enables robust, few-shot-ready tabular learning. The model is publicly available at https://github.com/Lexsi-Labs/Orion-BiX .