周报 2026-08-03 ~ 2026-08-09
生成时间：2026/8/9 11:02:08（UTC: 2026-08-09T03:02:08.650Z）
本周自动总结未启用或调用失败，以下为原始内容合并。
2026-08-03
Section titled “2026-08-03”内部采集快照，不作为站点文章发布。
生成时间：2026/8/3 09:30:03（UTC: 2026-08-03T01:30:03.456Z）
AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis
Section titled “AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis”👍 292 · arXiv
Chemistry literature synthesis often requires assembling specific findings scattered across many publications, yet existing literature-search systems primarily return ranked document lists. As a resul…
Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
Section titled “Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents”👍 285 · arXiv
GUI agents have the potential to become a general purpose executor over existing digital devices. To advance them toward real-world use, we envision agents that operate reliably on real devices, execu…
Metis: Memory Foundation Model
Section titled “Metis: Memory Foundation Model”👍 257 · arXiv
Recent advances in AI agents have increasingly internalized native capabilities into their underlying foundation models, giving rise to multimodal foundation models and large reasoning models. However…
Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering
Section titled “Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering”👍 168 · arXiv
Recursive self-improvement (RSI) requires AI systems that improve the process of building AI (i.e., AI4AI); machine learning engineering (MLE) offers a concrete, executable testbed for studying this c…
PhiZero: A World Model Built Around Physical Language
Section titled “PhiZero: A World Model Built Around Physical Language”👍 159 · arXiv
We introduce PhiZero, a physical world model built around physical language, a compact discrete representation of world-state transitions. Existing physical world models typically predict future video…
OpenClaw v2026.7.2-beta.7
Section titled “OpenClaw v2026.7.2-beta.7”2026.7.2
Section titled “2026.7.2”Highlights
Section titled “Highlights”- State safety and recovery: protect persisted data with a quarantine store that survives primary-database damage, crash-recoverable SQLite snapshots, crash-durable fi…
链接：https://github.com/openclaw/openclaw/releases/tag/v2026.7.2-beta.7
Sam Altman and AI’s decel debate
Section titled “Sam Altman and AI’s decel debate”On the latest episode of Equity, we discuss why Sam Altman has calling on the industry to “pace the rate of AI development.”
Judge denies xAI’s request to block Minnesota ban on ‘nudify’ apps
Section titled “Judge denies xAI’s request to block Minnesota ban on ‘nudify’ apps”Despite a lawsuit from xAI, a Minnesota ban on apps that allow users to “nudify” images can move forward.
YouTuber Hank Green says his AI usage is ‘not healthy’
Section titled “YouTuber Hank Green says his AI usage is ‘not healthy’”Green offered a remarkable apology, saying that “the level of dopamine that I’ve been getting from interacting with LLMs … is not healthy for me or good for the world.”
Sam Altman is still making the case for parenting via ChatGPT
Section titled “Sam Altman is still making the case for parenting via ChatGPT”OpenAI’s CEO seemed excited to share a “cool use case” for parents.
This $9 key physically locks your most addictive apps
Section titled “This $9 key physically locks your most addictive apps”This $9 NFC key requires you to physically scan it to unlock distracting apps on your phone.
OpenAI reportedly finds evidence that more of its agents ran amok
Section titled “OpenAI reportedly finds evidence that more of its agents ran amok”OpenAI has reportedly found evidence of additional agent misbehavior as it looks into the incident that occurred with Hugging Face.
India is starting to pay for apps, not just download them
Section titled “India is starting to pay for apps, not just download them”India’s app market generated a record $345 million in Q2.
Google nixes its Earth AI feature one day after launch, amid criticism it would spread misinformation
Section titled “Google nixes its Earth AI feature one day after launch, amid criticism it would spread misinformation”A tool that allowed anyone to generate fake AI-generated imagery and superimpose it over real Google Earth maps quickly spurred backlash.
2026-08-04
Section titled “2026-08-04”内部采集快照，不作为站点文章发布。
生成时间：2026/8/4 09:19:37（UTC: 2026-08-04T01:19:37.142Z）
Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
Section titled “Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents”👍 294 · arXiv
GUI agents have the potential to become a general purpose executor over existing digital devices. To advance them toward real-world use, we envision agents that operate reliably on real devices, execu…
From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement
Section titled “From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement”👍 74 · arXiv
Reinforcement Learning with Verifiable Rewards (RLVR) has driven recent progress in reasoning-oriented large language models (LLMs) by enabling large-scale optimization. However, its applicability rem…
Mental World Modeling
Section titled “Mental World Modeling”👍 53 · arXiv
World models enable a predictive substrate for planning and action, yet existing formulations merely answer a physical question: what/where it is, and how will it evolve. Human behavior, however, is d…
Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory
Section titled “Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory”👍 51 · arXiv
Decoder-only language models entangle long-term memory and reasoning in a single parameter set, making it difficult to scale memory capacity independently. Memory Decoder introduces a parametric long-…
N_0-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens
Section titled “N_0-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens”👍 50 · arXiv
We present N_0-VTLA, a vision-tactile-language-action (VTLA) foundation model capable of (1) fine-grained contact-rich manipulation with tactile perception and tactile-feedback control, and (2) offlin…
OpenClaw v2026.7.1-2
Section titled “OpenClaw v2026.7.1-2”- npm plugin updates: accept singleton-array metadata from newer npm clients so tracked official plugins can install and update to correction releases. (#108336)
…
链接：https://github.com/openclaw/openclaw/releases/tag/v2026.7.1-2
OpenAI Codex CLI rust-v0.147.0-alpha.6
Section titled “OpenAI Codex CLI rust-v0.147.0-alpha.6”Release 0.147.0-alpha.6
…
链接：https://github.com/openai/codex/releases/tag/rust-v0.147.0-alpha.6
AI’s debt binge can’t last, hidden borrowing reaches $1.65T
Section titled “AI’s debt binge can’t last, hidden borrowing reaches $1.65T”Article URL: https://fortune.com/2026/07/31/ai-debt-hypescalers-capex-capital-spending-hidden-borrowing-bond-issuance/ Comments URL: https://news.ycombinator.com/item?id=49160699 Points: 112
Comments: 132
Section titled “Comments: 132”What’s the largest software project AI can complete on its own?
Section titled “What’s the largest software project AI can complete on its own?”Article URL: https://epoch.ai/MirrorCode Comments URL: https://news.ycombinator.com/item?id=49157786 Points: 66
Comments: 74
Section titled “Comments: 74”The AI bubble is popping; we just don’t know it yet
Section titled “The AI bubble is popping; we just don’t know it yet”Article URL: https://www.theregister.com/ai-and-ml/2026/08/03/the-ai-bubble-is-already-popping-we-just-dont-know-it-yet/5282004 Comments URL: https://news.ycombinator.com/item?id=49154601 Points: 75
Comments: 135
Section titled “Comments: 135”SQLite Critical CVEs or LLM Slop?
Section titled “SQLite Critical CVEs or LLM Slop?”Article URL: https://research.jfrog.com/post/sqlite-critical-cves-or-llm-slops/ Comments URL: https://news.ycombinator.com/item?id=49154332 Points: 698
Comments: 347
Section titled “Comments: 347”Show HN: Nightcrawler – A local AI pentesting agent running on a smartphone
Section titled “Show HN: Nightcrawler – A local AI pentesting agent running on a smartphone”Article URL: https://github.com/garagehq/nightcrawler/ Comments URL: https://news.ycombinator.com/item?id=49154127 Points: 102
Comments: 30
Section titled “Comments: 30”Prevent cognitive debt by manually retyping LLM-generated code
Section titled “Prevent cognitive debt by manually retyping LLM-generated code”Article URL: https://ankursethi.com/blog/prevent-cognitive-debt-by-manually-retyping-llm-generated-code/ Comments URL: https://news.ycombinator.com/item?id=49153374 Points: 380
Comments: 319
Section titled “Comments: 319”The AI Productivity Gap
Section titled “The AI Productivity Gap”Article URL: https://bjorg.bjornroche.com/management/ai-productivity-gap/ Comments URL: https://news.ycombinator.com/item?id=49152222 Points: 105
Comments: 99
Section titled “Comments: 99”AI migrated legacy COBOL programs to Java, bugs included
Section titled “AI migrated legacy COBOL programs to Java, bugs included”Article URL: https://arxiv.org/abs/2607.28271 Comments URL: https://news.ycombinator.com/item?id=49150773 Points: 87
Comments: 85
Section titled “Comments: 85”2026-08-05
Section titled “2026-08-05”内部采集快照，不作为站点文章发布。
生成时间：2026/8/5 09:23:42（UTC: 2026-08-05T01:23:42.230Z）
SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks
Section titled “SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks”👍 140 · arXiv
Speech and audio generation is often needed in animation dubbing, audio drama, movies, advertising, games, podcasts, and short-video production. In these scenarios, creators may need to design voices …
LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks
Section titled “LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks”👍 127 · arXiv
Large language model (LLM) agents increasingly undertake long-horizon tasks that require sustained reasoning, tool use, and revision across many interdependent steps. However, existing agent harnesses…
DAPD: Dual-Anchored Policy Distillation
Section titled “DAPD: Dual-Anchored Policy Distillation”👍 56 · arXiv
On-policy (self) distillation (OPSD) is increasingly adopted for language-model post-training. It strengthens the teacher with privileged information but can induce a privilege illusion: the student l…
Weak-to-Strong On-Policy Distillation
Section titled “Weak-to-Strong On-Policy Distillation”👍 50 · arXiv
On-policy distillation (OPD), which aligns a student with the teacher’s token-level distribution on the student’s own rollouts, is an effective paradigm for transferring capabilities across LLMs. Prev…
Progressive Agent Skill Generation via Reinforcement Learning
Section titled “Progressive Agent Skill Generation via Reinforcement Learning”👍 49 · arXiv
Existing skill generation methods largely rely on heuristics or pipeline-style consolidation, which must be specially designed for different evidence sources. In contrast, learning-based approaches of…
OpenClaw v2026.7.1-2
Section titled “OpenClaw v2026.7.1-2”- npm plugin updates: accept singleton-array metadata from newer npm clients so tracked official plugins can install and update to correction releases. (#108336)
…
链接：https://github.com/openclaw/openclaw/releases/tag/v2026.7.1-2
Ollama v0.32.6-rc0
Section titled “Ollama v0.32.6-rc0”What’s Changed
Section titled “What’s Changed”- Qwen3.5 is faster on Apple GPUs: the MLX engine now uses the model’s MTP head for speculative decoding automatically
/v1/chat/completionsstreaming now matches OpenAI’s w…
链接：https://github.com/ollama/ollama/releases/tag/v0.32.6-rc0
OpenAI Codex CLI rust-v0.147.0-alpha.7
Section titled “OpenAI Codex CLI rust-v0.147.0-alpha.7”Release 0.147.0-alpha.7
…
链接：https://github.com/openai/codex/releases/tag/rust-v0.147.0-alpha.7
AI fuels more than half of cybercrime in Africa as scams surge – Interpol
Section titled “AI fuels more than half of cybercrime in Africa as scams surge – Interpol”https://www.interpol.int/Media/Documents/Publications/Cyberc…
Comments URL: https://news.ycombinator.com/item?id=49175826 Points: 129
Comments: 87
Section titled “Comments: 87”Security Incident INC-2026-07-28-01 – UK AI Security Institute [pdf]
Section titled “Security Incident INC-2026-07-28-01 – UK AI Security Institute [pdf]”Article URL: https://cdn.prod.website-files.com/663bd486c5e4c81588db7a1d/6a724858f7db25c81487016d_Security%20Incident%20INC-2026-07-28-01.pdf Comments URL: https://news.ycombinator.com/item?id=49175717 Points: 56
Comments: 47
Section titled “Comments: 47”AI Data Centers Are Driving Up Power Bills – This Map Shows Where
Section titled “AI Data Centers Are Driving Up Power Bills – This Map Shows Where”Article URL: https://www.gadgetreview.com/ai-data-centers-are-driving-up-power-bills-this-map-shows-where Comments URL: https://news.ycombinator.com/item?id=49172433 Points: 61
Comments: 20
Section titled “Comments: 20”The Warp Agent CLI
Section titled “The Warp Agent CLI”Article URL: https://www.warp.dev/blog/introducing-the-warp-agent-cli-coding-agent Comments URL: https://news.ycombinator.com/item?id=49171766 Points: 93
Comments: 59
Section titled “Comments: 59”When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation
Section titled “When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation”Article URL: https://arxiv.org/abs/2602.16763 Comments URL: https://news.ycombinator.com/item?id=49170915 Points: 78
Comments: 83
Section titled “Comments: 83”The AI Demand Bubble
Section titled “The AI Demand Bubble”Article URL: https://www.wheresyoured.at/the-ai-demand-bubble/ Comments URL: https://news.ycombinator.com/item?id=49170648 Points: 104
Comments: 121
Section titled “Comments: 121”Agent skills that bring team coding standards to Claude Code and Codex
Section titled “Agent skills that bring team coding standards to Claude Code and Codex”Article URL: https://github.com/tikalk/adlc-team-skills Comments URL: https://news.ycombinator.com/item?id=49169640 Points: 74
Comments: 39
Section titled “Comments: 39”It’s not a fear of “AI communism”; it’s a fear of competitive market capitalism
Section titled “It’s not a fear of “AI communism”; it’s a fear of competitive market capitalism”Article URL: http://observationalepidemiology.blogspot.com/2026/07/its-not-fear-of-ai-communism-its-fear.html Comments URL: https://news.ycombinator.com/item?id=49169227 Points: 78
Comments: 71
Section titled “Comments: 71”2026-08-06
Section titled “2026-08-06”内部采集快照，不作为站点文章发布。
生成时间：2026/8/6 09:21:04（UTC: 2026-08-06T01:21:04.906Z）
MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations
Section titled “MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations”👍 87 · arXiv
Large language model agents are increasingly evaluated as autonomous tool users, yet most benchmarks focus on bounded tasks with immediate success criteria. Real-world deployments often require Long-T…
JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion
Section titled “JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion”👍 79 · arXiv
Real-time video editing requires low-latency causal generation with bounded computational resources while preserving source fidelity and long-term temporal consistency. We present JoyAI-Video-Edit, a …
AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling
Section titled “AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling”👍 74 · arXiv
Language remains an outlier in generative modeling: while images, video, and audio are increasingly modeled in continuous latent spaces, text generation still relies predominantly on discrete tokens. …
Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing
Section titled “Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing”👍 71 · arXiv
Recent advances in image generation have demonstrated the potential of unified multimodal models that integrate understanding, generation, and editing. However, unified 3D modeling remains constrained…
InfiniSplat: Implicit Gaussian Decoding for Large-Baseline Monocular View Synthesis
Section titled “InfiniSplat: Implicit Gaussian Decoding for Large-Baseline Monocular View Synthesis”👍 56 · arXiv
Single-image feed-forward 3D Gaussian Splatting (3DGS) aims to directly generate a renderable 3D scene representation from one input image, avoiding the cost of multi-view capture and per-scene optimi…
LangChain langchain-anthropic==1.5.4
Section titled “LangChain langchain-anthropic==1.5.4”Changes since langchain-anthropic==1.5.3
release(anthropic): 1.5.4 (#39277) fix(anthropic): handle tool schemas with unsupported top-level composition (#39273) chore: bump the minor-and-patch group a…
链接：https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.5.4
Ollama v0.32.6
Section titled “Ollama v0.32.6”What’s Changed
Section titled “What’s Changed”- Qwen3.5 is faster on Apple GPUs: the MLX engine now uses the model’s MTP head for speculative decoding automatically
/v1/chat/completionsstreaming now matches OpenAI’s w…
链接：https://github.com/ollama/ollama/releases/tag/v0.32.6
CrewAI 1.15.12
Section titled “CrewAI 1.15.12”What’s Changed
Section titled “What’s Changed”Features
Section titled “Features”- Bump Flow canary on release
- Add URLReadTool for reading arbitrary URLs
- Add app metadata to platform action tools
- Unify scaffolding under `crewai create <resourc…
链接：https://github.com/crewAIInc/crewAI/releases/tag/1.15.12
OpenAI Codex CLI rust-v0.147.0-alpha.13
Section titled “OpenAI Codex CLI rust-v0.147.0-alpha.13”Release 0.147.0-alpha.13
…
链接：https://github.com/openai/codex/releases/tag/rust-v0.147.0-alpha.13
Prime Agent: A self-improving RLM agent
Section titled “Prime Agent: A self-improving RLM agent”Article URL: https://www.primeintellect.ai/blog/prime-agent Comments URL: https://news.ycombinator.com/item?id=49189075 Points: 95
Comments: 17
Section titled “Comments: 17”Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery
Section titled “Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery”Article URL: https://www.wired.com/story/meta-ran-ads-that-contained-ai-generated-child-sexual-abuse-imagery/ Comments URL: https://news.ycombinator.com/item?id=49187977 Points: 244
Comments: 198
Section titled “Comments: 198”Born Against, or why hobby programming communities are against LLM usage
Section titled “Born Against, or why hobby programming communities are against LLM usage”Article URL: https://blog.fogus.me/llm/born-against.html Comments URL: https://news.ycombinator.com/item?id=49187061 Points: 123
Comments: 137
Section titled “Comments: 137”Microsoft’s AI Sales Mostly Come from OpenAI, Disclosures Show
Section titled “Microsoft’s AI Sales Mostly Come from OpenAI, Disclosures Show”https://www.bloomberg.com/news/articles/2026-08-05/microsoft…
Comments URL: https://news.ycombinator.com/item?id=49186766 Points: 61
Comments: 16
Section titled “Comments: 16”Beating GPT-5.6 Sol on retrieval with 100x cheaper open models
Section titled “Beating GPT-5.6 Sol on retrieval with 100x cheaper open models”Article URL: https://neon.com/blog/how-castform-neon-beats-frontier-models-on-price-and-efficiency Comments URL: https://news.ycombinator.com/item?id=49186762 Points: 219
Comments: 39
Section titled “Comments: 39”Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)
Section titled “Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)”Article URL: https://arxiv.org/abs/2510.01395 Comments URL: https://news.ycombinator.com/item?id=49186720 Points: 70
Comments: 50
Section titled “Comments: 50”TIME Is Serving AI Bots a Different Website, with Ads Built In
Section titled “TIME Is Serving AI Bots a Different Website, with Ads Built In”Article URL: https://www.vincentschmalbach.com/time-serves-ai-bots-a-different-website/ Comments URL: https://news.ycombinator.com/item?id=49182041 Points: 230
Comments: 97
Section titled “Comments: 97”Anthropic AI created fake profiles and impersonated people in attempted hack
Section titled “Anthropic AI created fake profiles and impersonated people in attempted hack”Article URL: https://www.bbc.co.uk/news/articles/c1w1lvn7d9go Comments URL: https://news.ycombinator.com/item?id=49181773 Points: 50
Comments: 20
Section titled “Comments: 20”2026-08-07
Section titled “2026-08-07”内部采集快照，不作为站点文章发布。
生成时间：2026/8/7 10:01:00（UTC: 2026-08-07T02:01:00.211Z）
ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment
Section titled “ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment”👍 55 · arXiv
Long-horizon search agents must make multiple sequential actions (steps) to search, retrieve, verify, and integrate evidence to reach a final answer. However, existing methods for training these agent…
ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation
Section titled “ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation”👍 46 · arXiv
Text-to-image (T2I) models can produce visually compelling images, yet they remain limited on open-world tasks that require complex semantic understanding, multi-step reasoning, and the integration of…
Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes
Section titled “Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes”👍 42 · arXiv
Vision offers a critical axis for advancing foundation models, driving a shift towards natively unified multimodal pretraining. Despite this momentum, the design space and the fundamental mechanisms o…
The Personalization Mirage: How LLMs Fabricate User Profiles, and Why Self-Monitoring Misleads
Section titled “The Personalization Mirage: How LLMs Fabricate User Profiles, and Why Self-Monitoring Misleads”👍 37 · arXiv
Personalized LLMs with persistent memory are increasingly deployed, yet the faithfulness of their user models remains unexamined. We study over-inference (OI): the phenomenon where LLMs fabricate user…
OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents
Section titled “OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents”👍 28 · arXiv
LLM agents are increasingly applied to open-ended everyday requests that span work, study, and life. These tasks are long-horizon, cross-environment, and multimodal, forcing the agent to preserve goal…
LangChain langchain-anthropic==1.5.4
Section titled “LangChain langchain-anthropic==1.5.4”Changes since langchain-anthropic==1.5.3
release(anthropic): 1.5.4 (#39277) fix(anthropic): handle tool schemas with unsupported top-level composition (#39273) chore: bump the minor-and-patch group a…
链接：https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.5.4
CrewAI 1.15.12
Section titled “CrewAI 1.15.12”What’s Changed
Section titled “What’s Changed”Features
Section titled “Features”- Bump Flow canary on release
- Add URLReadTool for reading arbitrary URLs
- Add app metadata to platform action tools
- Unify scaffolding under `crewai create <resourc…
链接：https://github.com/crewAIInc/crewAI/releases/tag/1.15.12
OpenAI Codex CLI rust-v0.147.0
Section titled “OpenAI Codex CLI rust-v0.147.0”New Features
Section titled “New Features”- Install portable Agent Plugins and search across local, personal, workspace, and remote plugin catalogs. (#36544, #36409, #36919, #36796)
- Organize conversations into persistent, ma…
链接：https://github.com/openai/codex/releases/tag/rust-v0.147.0
OpenAI’s new AI smart speaker will reportedly sell for between $300 and $400
Section titled “OpenAI’s new AI smart speaker will reportedly sell for between $300 and $400”Additional details about OpenAI’s mysterious new AI device make it sound like a pricey smart speaker.
ChatGPT brings unlimited text chats to free users
Section titled “ChatGPT brings unlimited text chats to free users”OpenAI said that ChatGPT free and Go users are also getting a new think button for complex queries.
Naïve raises $28.5M to automate the grunt work of setting up and running a company
Section titled “Naïve raises $28.5M to automate the grunt work of setting up and running a company”Taking vibe-coding a step further, Naïve claims its infra can automate most of the work in setting up and running a business.
Gen Z dating apps like Ditto ditch swiping in favor of AI matchmaking
Section titled “Gen Z dating apps like Ditto ditch swiping in favor of AI matchmaking”This generation of twentysomethings is so disillusioned with swipe-based dating apps that they’ll try literally anything else — even an AI matchmaker.
OpenAI says Apple’s own security practices undermine its trade secrets case
Section titled “OpenAI says Apple’s own security practices undermine its trade secrets case”Newly filed court exhibits show OpenAI’s legal strategy in Apple’s trade secrets lawsuit: argue that Apple’s own security and offboarding practices — including allowing an Apple manager to access a former engineer’s iCloud account after he left the company —undermine its claims that the allegedly st
Amid legal battles, Suno says it will start watermarking songs
Section titled “Amid legal battles, Suno says it will start watermarking songs”Suno’s watermarking feature comes as the company is fighting legal battles on several fronts.
Ex-Spotify employees raise $10M to bring the AI behind its recommendations to e-commerce
Section titled “Ex-Spotify employees raise $10M to bring the AI behind its recommendations to e-commerce”The startup’s platform predicts which product a shopper wants next, learns their general taste, and fine-tunes continuously based on what they do in real time.
Exclusive: Mirendil inks $100M+ Google Cloud deal to scale self-improving AI
Section titled “Exclusive: Mirendil inks $100M+ Google Cloud deal to scale self-improving AI”Mirendil has signed a $100 million-plus Google Cloud partnership to expand its compute infrastructure, powering research into self-improving AI systems designed to accelerate scientific discovery and AI development.
2026-08-08
Section titled “2026-08-08”内部采集快照，不作为站点文章发布。
生成时间：2026/8/8 08:44:41（UTC: 2026-08-08T00:44:41.845Z）
Recursive Synthesis for Long-Horizon Terminal Tasks
Section titled “Recursive Synthesis for Long-Horizon Terminal Tasks”👍 212 · arXiv
High-quality long-horizon training data for terminal agents is expensive to produce, often costing hundreds to thousands of dollars per task, because each task must keep the instruction, environment, …
AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning
Section titled “AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning”👍 67 · arXiv
Reinforcement learning (RL) with verifiable rewards constructs trajectory-level advantage estimates, yet it often fails to credit the few pivotal decisions that determine outcomes in long-horizon, mul…
ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment
Section titled “ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment”👍 61 · arXiv
Long-horizon search agents must make multiple sequential actions (steps) to search, retrieve, verify, and integrate evidence to reach a final answer. However, existing methods for training these agent…
OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
Section titled “OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models”👍 52 · arXiv
Computer-using agents (CUAs) are advancing rapidly across the digital world. A CUA trajectory records the agent’s actions, states, and reasoning. Verifying whether it fulfilled the task instruction is…
Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval
Section titled “Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval”👍 46 · arXiv
Short segments of perceived speech can be retrieved from non-invasive magnetoencephalographic (MEG) recordings by deep networks trained with a CLIP-style objective against wav2vec 2.0 audio embeddings…
LangChain langchain-openai==1.4.2
Section titled “LangChain langchain-openai==1.4.2”Changes since langchain-openai==1.4.1
release(openai): 1.4.2 (#39322)
fix(openai): handle ContextWindowExceededError (#39300)
chore: bump the minor-and-patch group across 3 directories with 7 updat…
链接：https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.4.2
CrewAI 1.15.13
Section titled “CrewAI 1.15.13”What’s Changed
Section titled “What’s Changed”Bug Fixes
Section titled “Bug Fixes”- Fix preservation of provider on LiteLLM-routed models.
- Harden brittle LLM event-bus mocks.
- Fix underreporting of Anthropic cache token usage.
- Bump h2 to versio…
链接：https://github.com/crewAIInc/crewAI/releases/tag/1.15.13
OpenAI Codex CLI rust-v0.147.0
Section titled “OpenAI Codex CLI rust-v0.147.0”New Features
Section titled “New Features”- Install portable Agent Plugins and search across local, personal, workspace, and remote plugin catalogs. (#36544, #36409, #36919, #36796)
- Organize conversations into persistent, ma…
链接：https://github.com/openai/codex/releases/tag/rust-v0.147.0
Managing AI Coding Costs at Scale
Section titled “Managing AI Coding Costs at Scale”Article URL: https://www.databricks.com/blog/managing-ai-coding-costs-scale Comments URL: https://news.ycombinator.com/item?id=49214468 Points: 155
Comments: 158
Section titled “Comments: 158”Oracle bans AI-generated code from OpenJDK
Section titled “Oracle bans AI-generated code from OpenJDK”Article URL: https://app.dealroom.co/news/feed/oracle-bans-ai-generated-code-from-openjdk-despite-ellison-s-claim-oracle-isn-t-writing-its-own-code Comments URL: https://news.ycombinator.com/item?id=49213754 Points: 371
Comments: 247
Section titled “Comments: 247”AI psychosis is the new leadership blind spot
Section titled “AI psychosis is the new leadership blind spot”Article URL: https://www.fastcompany.com/91576086/ai-psychosis-is-the-new-leadership-blind-spot-ai-leadership-blind-spots Comments URL: https://news.ycombinator.com/item?id=49210077 Points: 159
Comments: 102
Section titled “Comments: 102”Kitesurf: Agent-first browser that runs in V8 isolates
Section titled “Kitesurf: Agent-first browser that runs in V8 isolates”Article URL: https://blog.cloudflare.com/kitesurf/ Comments URL: https://news.ycombinator.com/item?id=49208393 Points: 160
Comments: 42
Section titled “Comments: 42”I won’t read LLM authored fiction
Section titled “I won’t read LLM authored fiction”Article URL: https://mccormick.cx/news/entries/why-i-won-t-read-llm-authored-fiction Comments URL: https://news.ycombinator.com/item?id=49207146 Points: 70
Comments: 112
Section titled “Comments: 112”New Orleans is testing Carbyne’s AI-powered Emergency Call Triage software
Section titled “New Orleans is testing Carbyne’s AI-powered Emergency Call Triage software”Article URL: https://www.shreveporttimes.com/story/news/local/louisiana/2026/07/28/is-new-orleans-using-ai-to-answer-911-calls-instead-of-human-dispatchers-impacts-emergencies-crime/91065014007/ Comments URL: https://news.ycombinator.com/item?id=49204546 Points: 72
Comments: 117
Section titled “Comments: 117”Inside vLLM: Anatomy of a High-Throughput LLM Inference System (2025)
Section titled “Inside vLLM: Anatomy of a High-Throughput LLM Inference System (2025)”Article URL: https://www.aleksagordic.com/blog/vllm Comments URL: https://news.ycombinator.com/item?id=49202852 Points: 142
Comments: 9
Section titled “Comments: 9”xAI, SpaceX, and the Race for AI Buildout
Section titled “xAI, SpaceX, and the Race for AI Buildout”Article URL: https://illegal.solutions/posts/xai_pollution Comments URL: https://news.ycombinator.com/item?id=49201342 Points: 144