Skip to content

周报 2026-07-27 ~ 2026-08-02

生成时间:2026/8/2 12:24:45(UTC: 2026-08-02T04:24:45.372Z)

本周自动总结未启用或调用失败,以下为原始内容合并。

内部采集快照,不作为站点文章发布。

生成时间:2026/7/31 15:15:05(UTC: 2026-07-31T07:15:05.311Z)

TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM

Section titled “TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM”

👍 122 · arXiv

Vision-language-action (VLA) models commonly adopt an LLM-centric V to L to A pathway, where visual observations are projected into the representation space of a large language model before being deco…

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”

👍 100 · 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…

DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillation

Section titled “DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillation”

👍 87 · arXiv

Existing autoregressive video distillation methods commonly adopt a Distribution Matching Distillation (DMD)-based multi-stage pipeline. However, they typically decouple the initialization and DMD sta…

👍 86 · 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…

AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis

Section titled “AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis”

👍 79 · 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…

Changes since langchain-core==1.5.2

release(core): 1.5.3 (#39145) fix(core): fall back to LANGSMITH_API_KEY for gateway (#39115)…

链接https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.5.3

  • Surface tool failures instead of reporting them as success
  • Emit FlowFailedEvent when a flow execution fails
  • Implement progressive disclosure for skills

链接https://github.com/crewAIInc/crewAI/releases/tag/1.15.9

链接https://github.com/aaif-goose/goose/releases/tag/v1.45.0

Release 0.147.0-alpha.2

链接https://github.com/openai/codex/releases/tag/rust-v0.147.0-alpha.2

The AI trade now runs on borrowed money, and the lenders are repricing it

Section titled “The AI trade now runs on borrowed money, and the lenders are repricing it”

Article URL: https://greyswansignals.com/?theme=dark Comments URL: https://news.ycombinator.com/item?id=49118933 Points: 129

来源Hacker News AI

Article URL: https://blog.jim-nielsen.com/2026/ai-aesthetic/ Comments URL: https://news.ycombinator.com/item?id=49117099 Points: 279

来源Hacker News AI

Agent Skill to Force Docs in ASD-STE100 Simplified Technical English

Section titled “Agent Skill to Force Docs in ASD-STE100 Simplified Technical English”

Article URL: https://github.com/AminBlg/SimpleEnglish Comments URL: https://news.ycombinator.com/item?id=49114639 Points: 286

来源Hacker News AI

Show HN: Distilling DeepSeek into GPT-OSS doesn’t transfer censorship. Try it

Section titled “Show HN: Distilling DeepSeek into GPT-OSS doesn’t transfer censorship. Try it”

We recently used DeepSeek V4 Flash as a teacher for finance tasks with GPT-OSS-120B. Distillation works well on this problem. At a constrained 8k token budget, our self-distilled 120B scores 83.61% on FinanceReasoning, above Kimi K3 (81.93%) and Inkling (65.13%). We released the 20B open weights. Wi

来源Hacker News AI

We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447

Section titled “We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447”

Article URL: https://www.bottlenecklabs.com/blog/autonomously-run-businesses Comments URL: https://news.ycombinator.com/item?id=49113059 Points: 347

来源Hacker News AI

Advancing the price-performance frontier with GPT‑5.6

Section titled “Advancing the price-performance frontier with GPT‑5.6”

Article URL: https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6/ Comments URL: https://news.ycombinator.com/item?id=49112867 Points: 555

来源Hacker News AI

Citadel Buys Situational Awareness’s Stock Portfolio After Big Losses in AI

Section titled “Citadel Buys Situational Awareness’s Stock Portfolio After Big Losses in AI”

Article URL: https://www.wsj.com/finance/citadel-buys-situational-awarenesss-stock-portfolio-after-big-losses-in-ai-5117159b Comments URL: https://news.ycombinator.com/item?id=49111879 Points: 51

来源Hacker News AI

Article URL: https://openjdk.org/legal/ai Comments URL: https://news.ycombinator.com/item?id=49109165 Points: 70

来源Hacker News AI


内部采集快照,不作为站点文章发布。

生成时间:2026/8/1 09:44:04(UTC: 2026-08-01T01:44:04.906Z)

AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis

Section titled “AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis”

👍 285 · 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”

👍 267 · 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…

👍 241 · 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…

PhiZero: A World Model Built Around Physical Language

Section titled “PhiZero: A World Model Built Around Physical Language”

👍 150 · 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…

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”

👍 145 · 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…

Changes since langchain-core==1.5.2

release(core): 1.5.3 (#39145) fix(core): fall back to LANGSMITH_API_KEY for gateway (#39115)…

链接https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.5.3

  • Collect skill usage events
  • Remove migrated AMP documentation
  • Update security reporting guidelines
  • Snapshot and changelog for v1.15.9

链接https://github.com/crewAIInc/crewAI/releases/tag/1.15.10

Release 0.147.0-alpha.4

链接https://github.com/openai/codex/releases/tag/rust-v0.147.0-alpha.4

Everyone is building LLM routers, we deprecated ours

Section titled “Everyone is building LLM routers, we deprecated ours”

Article URL: https://manifest.build/blog/why-we-deprecated-our-llm-router/ Comments URL: https://news.ycombinator.com/item?id=49126630 Points: 90

来源Hacker News AI

Article URL: https://github.com/yc-software/qm Comments URL: https://news.ycombinator.com/item?id=49126604 Points: 457

来源Hacker News AI

Is AI reasoning right for the wrong reasons?

Section titled “Is AI reasoning right for the wrong reasons?”

Article URL: https://www.quantamagazine.org/is-ai-reasoning-right-for-the-wrong-reasons-20260731/ Comments URL: https://news.ycombinator.com/item?id=49124358 Points: 119

来源Hacker News AI

Situational Awareness down 67% in July in AI stock rout

Section titled “Situational Awareness down 67% in July in AI stock rout”

Article URL: https://www.wsj.com/finance/investing/situational-awareness-down-67-in-july-in-ai-stock-rout-cd19901f Comments URL: https://news.ycombinator.com/item?id=49122994 Points: 140

来源Hacker News AI

The Maxwell Conjecture Is False (GPT 5.6 Sol)

Section titled “The Maxwell Conjecture Is False (GPT 5.6 Sol)”

Article URL: https://arxiv.org/abs/2607.27197 Comments URL: https://news.ycombinator.com/item?id=49121868 Points: 139

来源Hacker News AI

Google fixed more Chrome bugs in June than over the past two years, thanks to AI

Section titled “Google fixed more Chrome bugs in June than over the past two years, thanks to AI”

Article URL: https://blog.google/security/chrome-stronger-with-every-update/ Comments URL: https://news.ycombinator.com/item?id=49120097 Points: 481

来源Hacker News AI

Show HN: What should the GUI for AI agents look like?

Section titled “Show HN: What should the GUI for AI agents look like?”

Hi HN! We’re Akilan and Miguel, the creators of MarbleOS.The inspiration for Marble comes from the GUI work at Xerox PARC, the 1984 Macintosh, and later NeXTSTEP, which became the foundation for Mac OS X. Before GUIs, interacting with a computer was limited to strange terminal commands:C:> DIRC:>

来源Hacker News AI

The AI trade now runs on borrowed money, and the lenders are repricing it

Section titled “The AI trade now runs on borrowed money, and the lenders are repricing it”

Article URL: https://greyswansignals.com/?theme=dark Comments URL: https://news.ycombinator.com/item?id=49118933 Points: 140

来源Hacker News AI