NO/FOMO

每天一次，过滤 AI 噪音

值得打开的
AI 日报。

发布日期2026-01-16中文版本
本期阅读
—
累计阅读
—

Hacker News

6 stories
01

How to wrangle non-deterministic AI outputs into conventional software? (2025)

The article addresses the critical engineering challenge of integrating inherently non-deterministic AI outputs into conventional, deterministic software systems, a growing concern as AI models become more sophisticated and ubiquitous. With the rise of large language models and other generative AI, their probabilistic and often unpredictable nature clashes with the reliability requirements of traditional software development. The piece delves into architectural strategies and design patterns aimed at "wrangling" these AI outputs, transforming them into structured, predictable data or actions that can be consumed by deterministic components. Key considerations include implementing robust validation layers, employing bounded non-determinism, designing effective retry mechanisms, and strategically incorporating human-in-the-loop processes. This discussion provides essential guidance for developing resilient software architectures capable of harnessing AI's capabilities without compromising system stability, which is vital for the widespread integration of AI components anticipated by 2025. It underscores the necessity for evolving software engineering practices to effectively manage AI's unique operational characteristics.

AI IntegrationNon-deterministic SystemsSoftware ArchitectureAI ComponentsSystem DesignArtificial IntelligenceMachine LearningGenerative AI
02

Our approach to advertising and expanding access to ChatGPT

OpenAI has publicly shared its strategic framework for integrating advertising into the ChatGPT platform while simultaneously expanding its global accessibility. This comprehensive approach aims to strike a critical balance between fostering sustainable commercial growth for its advanced large language model technology and ensuring wider public and enterprise access to its AI capabilities. The discussion is anticipated to delve into the specific mechanisms for advertising implementation, emphasizing user experience, data privacy, and ethical considerations to maintain user trust. Furthermore, OpenAI is expected to detail its plans for scaling ChatGPT's availability, potentially through new subscription tiers, partnerships, or localized versions, thereby reaching diverse user bases and application domains. This dual strategy underscores OpenAI's commitment to both the responsible evolution of AI and the exploration of innovative monetization models necessary to fund continuous research and development in the rapidly advancing field of artificial intelligence.

ChatGPTAI MonetizationLarge Language ModelAdvertising StrategyAccess ExpansionLarge Language ModelArtificial IntelligenceOthers
03

The Toxic Modernity Narrative

The article 'The Toxic Modernity Narrative' critically examines the prevalent societal discourse that often portrays contemporary technological and cultural advancements as inherently detrimental. It delves into the underlying arguments and philosophical underpinnings of this narrative, which frequently highlights the negative consequences of rapid innovation, including the perceived dehumanizing effects of digital transformation and the ethical challenges posed by emerging technologies such as artificial intelligence. The piece likely analyzes how this 'toxic modernity' perspective influences public perception of progress, fostering skepticism towards advancements in automation, data science, and AI-driven solutions. It may explore concerns related to privacy erosion, algorithmic bias, and the potential for technological unemployment, advocating for a more balanced and nuanced evaluation of modernity's complexities rather than succumbing to a purely pessimistic outlook. The analysis likely encourages a re-evaluation of how society frames technological progress, emphasizing the need for critical engagement over wholesale rejection.

Technological EthicsSocietal Impact of AIDigital Transformation CritiqueAlgorithmic BiasTechnological UnemploymentArtificial IntelligenceMachine LearningLarge Language Model
04

DuckDuckGo is asking for a Yes or No vote on AI

DuckDuckGo, a search engine renowned for its unwavering commitment to user privacy, has launched an innovative public 'Yes or No' vote regarding artificial intelligence. This proactive step underscores the company's intention to directly involve its user community in shaping its stance and potential future integration of AI technologies. The initiative suggests a strategic move to gauge user sentiment on AI's role within DuckDuckGo's ecosystem, particularly amidst growing industry discussions surrounding AI ethics, data privacy, and algorithmic transparency. While the precise scope of the AI applications under consideration for the vote remains unspecified, it reflects a broader industry trend where technology companies are increasingly grappling with the implications of AI adoption. The results of this public consultation are expected to provide valuable insights that could guide DuckDuckGo's product development roadmap, influencing decisions on how or if AI-powered features might be implemented, always in alignment with its foundational privacy-first principles. This approach exemplifies a commitment to user-driven decision-making in the evolving landscape of AI.

Artificial IntelligenceAI EthicsData PrivacyUser EngagementSearch EnginesArtificial IntelligenceMachine LearningNatural Language Processing
05

Show HN: Aventos – An experiment in cheap AI SEO

Aventos, an experimental project, aims to offer a cost-effective solution for tracking company mentions within Large Language Models (LLMs). Developed after exploring the AI search, GEO, and AEO sectors, Aventos addresses the current landscape of tools that typically simulate or scrape LLM query results. The creators note a recent shift towards scraping live ChatGPT outputs for their cost efficiency and realism, despite the inherent challenges of building and maintaining robust scrapers. Many existing SaaS products in this category function as wrappers around third-party scraping APIs for platforms like ChatGPT, Perplexity, and Google AIO. Despite their reliance on these underlying data providers, these tools often command high monthly fees ranging from $70 to $200+. Aventos seeks to challenge this pricing model by providing a significantly cheaper alternative, leveraging similar data sources without the inflated costs, thus making LLM-based mention tracking more accessible for businesses.

AI SEOLarge Language ModelsWeb ScrapingSaaSAI Search ToolsArtificial IntelligenceLarge Language ModelNatural Language Processing
06

Cursor's latest "browser experiment" implied success without evidence

A recent Hacker News post critically examines Cursor's latest "browser experiment," asserting that the company implied significant success without substantiating these claims with adequate evidence. This critique underscores a broader concern within the tech community regarding transparency and the necessity for verifiable results, especially from companies operating in innovative and rapidly evolving sectors. The context of a related discussion, "Scaling long-running autonomous coding," strongly suggests that Cursor is involved in developing AI-powered coding assistants or automated development tools that likely integrate with browser environments. The central issue highlighted is the importance of integrity in product announcements and the presentation of performance metrics for advanced capabilities, such as those related to autonomous agents. Without clear, data-driven proof, assertions of success can erode user confidence and raise questions about the true efficacy and reliability of new features, urging for a more rigorous approach to demonstrating technological advancements in the field of AI and software development.

CursorBrowser ExperimentAutonomous CodingAI AgentSoftware Development ToolsArtificial IntelligenceAI AgentGenerative AI

huggingface

6 stories
01

STEP3-VL-10B Technical Report

We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-VL-10B is realized through two strategic shifts: first, a unified, fully unfrozen pre-training strategy on 1.2T multimodal tokens that integrates a language-aligned Perception Encoder with a Qwen3-8B decoder to establish intrinsic vision-language synergy; and second, a scaled post-training pipeline featuring over 1k iterations of reinforcement learning. Crucially, we implement Parallel Coordinated Reasoning (PaCoRe) to scale test-time compute, allocating resources to scalable perceptual reasoning that explores and synthesizes diverse visual hypotheses. Consequently, despite its compact 10B footprint, STEP3-VL-10B rivals or surpasses models 10times-20times larger (e.g., GLM-4.6V-106B, Qwen3-VL-235B) and top-tier proprietary flagships like Gemini 2.5 Pro and Seed-1.5-VL. Delivering best-in-class performance, it records 92.2% on MMBench and 80.11% on MMMU, while excelling in complex reasoning with 94.43% on AIME2025 and 75.95% on MathVision. We release the full model suite to provide the community with a powerful, efficient, and reproducible baseline.

Multimodal IntelligenceFoundation ModelVision-LanguageReinforcement LearningParallel Coordinated ReasoningMultimodalLarge Language ModelComputer Vision
02

A Safety Report on GPT-5.2, Gemini 3 Pro, Qwen3-VL, Doubao 1.8, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5

The rapid evolution of Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) has produced substantial gains in reasoning, perception, and generative capability across language and vision. However, whether these advances yield commensurate improvements in safety remains unclear, in part due to fragmented evaluation practices limited to single modalities or threat models. In this report, we present an integrated safety evaluation of 7 frontier models: GPT-5.2, Gemini 3 Pro, Qwen3-VL, Doubao 1.8, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5. We evaluate each model across language, vision-language, and image generation settings using a unified protocol that integrates benchmark evaluation, adversarial evaluation, multilingual evaluation, and compliance evaluation. Aggregating our evaluations into safety leaderboards and model safety profiles across multiple evaluation modes reveals a sharply heterogeneous safety landscape. While GPT-5.2 demonstrates consistently strong and balanced safety performance across evaluations, other models exhibit pronounced trade-offs among benchmark safety, adversarial alignment, multilingual generalization, and regulatory compliance. Both language and vision-language modalities show significant vulnerability under adversarial evaluation, with all models degrading substantially despite strong results on standard benchmarks. Text-to-image models achieve relatively stronger alignment in regulated visual risk categories, yet remain brittle under adversarial or semantically ambiguous prompts. Overall, these results show that safety in frontier models is inherently multidimensional--shaped by modality, language, and evaluation scheme, underscoring the need for standardized safety evaluations to accurately assess real-world risk and guide responsible model development and deployment.

AI SafetyLarge Language ModelsMultimodal Large Language ModelsModel EvaluationAdversarial EvaluationLarge Language ModelMultimodalGenerative AI
03

Toward Ultra-Long-Horizon Agentic Science: Cognitive Accumulation for Machine Learning Engineering

The advancement of artificial intelligence toward agentic science is currently bottlenecked by the challenge of ultra-long-horizon autonomy, the ability to sustain strategic coherence and iterative correction over experimental cycles spanning days or weeks. While Large Language Models (LLMs) have demonstrated prowess in short-horizon reasoning, they are easily overwhelmed by execution details in the high-dimensional, delayed-feedback environments of real-world research, failing to consolidate sparse feedback into coherent long-term guidance. Here, we present ML-Master 2.0, an autonomous agent that masters ultra-long-horizon machine learning engineering (MLE) which is a representative microcosm of scientific discovery. By reframing context management as a process of cognitive accumulation, our approach introduces Hierarchical Cognitive Caching (HCC), a multi-tiered architecture inspired by computer systems that enables the structural differentiation of experience over time. By dynamically distilling transient execution traces into stable knowledge and cross-task wisdom, HCC allows agents to decouple immediate execution from long-term experimental strategy, effectively overcoming the scaling limits of static context windows. In evaluations on OpenAI's MLE-Bench under 24-hour budgets, ML-Master 2.0 achieves a state-of-the-art medal rate of 56.44%. Our findings demonstrate that ultra-long-horizon autonomy provides a scalable blueprint for AI capable of autonomous exploration beyond human-precedent complexities.

ultra-long-horizon autonomyagentic sciencemachine learning engineeringHierarchical Cognitive CachingAI agentsAI AgentMachine LearningLarge Language Model
04

Think-Then-Generate: Reasoning-Aware Text-to-Image Diffusion with LLM Encoders

Recent progress in text-to-image (T2I) diffusion models (DMs) has enabled high-quality visual synthesis from diverse textual prompts. Yet, most existing T2I DMs, even those equipped with large language model (LLM)-based text encoders, remain text-pixel mappers -- they employ LLMs merely as text encoders, without leveraging their inherent reasoning capabilities to infer what should be visually depicted given the textual prompt. To move beyond such literal generation, we propose the think-then-generate (T2G) paradigm, where the LLM-based text encoder is encouraged to reason about and rewrite raw user prompts; the states of the rewritten prompts then serve as diffusion conditioning. To achieve this, we first activate the think-then-rewrite pattern of the LLM encoder with a lightweight supervised fine-tuning process. Subsequently, the LLM encoder and diffusion backbone are co-optimized to ensure faithful reasoning about the context and accurate rendering of the semantics via Dual-GRPO. In particular, the text encoder is reinforced using image-grounded rewards to infer and recall world knowledge, while the diffusion backbone is pushed to produce semantically consistent and visually coherent images. Experiments show substantial improvements in factual consistency, semantic alignment, and visual realism across reasoning-based image generation and editing benchmarks, achieving 0.79 on WISE score, nearly on par with GPT-4. Our results constitute a promising step toward next-generation unified models with reasoning, expression, and demonstration capacities.

Text-to-Image Diffusion ModelsLLM EncodersReasoning-Aware GenerationThink-Then-Generate (T2G)Semantic AlignmentGenerative AILarge Language ModelMultimodal
05

FlowAct-R1: Towards Interactive Humanoid Video Generation

Interactive humanoid video generation aims to synthesize lifelike visual agents that can engage with humans through continuous and responsive video. Despite recent advances in video synthesis, existing methods often grapple with the trade-off between high-fidelity synthesis and real-time interaction requirements. In this paper, we propose FlowAct-R1, a framework specifically designed for real-time interactive humanoid video generation. Built upon a MMDiT architecture, FlowAct-R1 enables the streaming synthesis of video with arbitrary durations while maintaining low-latency responsiveness. We introduce a chunkwise diffusion forcing strategy, complemented by a novel self-forcing variant, to alleviate error accumulation and ensure long-term temporal consistency during continuous interaction. By leveraging efficient distillation and system-level optimizations, our framework achieves a stable 25fps at 480p resolution with a time-to-first-frame (TTFF) of only around 1.5 seconds. The proposed method provides holistic and fine-grained full-body control, enabling the agent to transition naturally between diverse behavioral states in interactive scenarios. Experimental results demonstrate that FlowAct-R1 achieves exceptional behavioral vividness and perceptual realism, while maintaining robust generalization across diverse character styles.

Interactive Video GenerationReal-time SynthesisDiffusion ModelsTemporal ConsistencyHumanoid AgentsGenerative AIComputer VisionAI Agent
06

Rewarding the Rare: Uniqueness-Aware RL for Creative Problem Solving in LLMs

Reinforcement learning (RL) has become a central paradigm for post-training large language models (LLMs), particularly for complex reasoning tasks, yet it often suffers from exploration collapse: policies prematurely concentrate on a small set of dominant reasoning patterns, improving pass@1 while limiting rollout-level diversity and gains in pass@k. We argue that this failure stems from regularizing local token behavior rather than diversity over sets of solutions. To address this, we propose Uniqueness-Aware Reinforcement Learning, a rollout-level objective that explicitly rewards correct solutions that exhibit rare high-level strategies. Our method uses an LLM-based judge to cluster rollouts for the same problem according to their high-level solution strategies, ignoring superficial variations, and reweights policy advantages inversely with cluster size. As a result, correct but novel strategies receive higher rewards than redundant ones. Across mathematics, physics, and medical reasoning benchmarks, our approach consistently improves pass@k across large sampling budgets and increases the area under the pass@k curve (AUC@K) without sacrificing pass@1, while sustaining exploration and uncovering more diverse solution strategies at scale.

Reinforcement LearningLarge Language ModelsExploration CollapseCreative Problem SolvingSolution DiversityLarge Language ModelMachine LearningNatural Language Processing