AI 情报文章归档

This is a submission for the Hacktoberfest Open-Source AI Challenge: Week 2 — Side Hustle ...

政府和公司即将做出一个重大选择，因为它们正日益被大量智能体AI查询所淹没，这些查询绕过了原本限制访问的摩擦（见图表——这已经开始了！） 两个选项：你最终是让服务变得更差还是更好？

@ben_moll 的澄清很好。 这些是模型隐含的增长情景，而非直接的 GDP 预测。 Anthropic 员工预期，随着 AI 被使用，模型能力会更强、扩散会更快、生产率提升会更大。在模型中，这些力量会加速增长。

A job offer lands in a family WhatsApp group (a composite scene, built from patterns in published...

我们把果蝇大脑放进了Odyssey-3，一个前沿世界模型。 这个小家伙现在完全靠自己，在完全由想象生成的模拟世界里飞来飞去 🫶

想找个周末项目做做？ 以下是社区一直在构建的东西，以及你可以如何自己复现： https://youcanjustbuildthings.openai.chatgpt.site

Battle-tested pgvector HNSW tuning and hybrid search implementation after hitting severe concurrency bottlenecks in a production RAG pipeline.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

Distribution regression seeks to learn mappings from probability distributions to response variables and has emerged as a unifying framework for kernel methods, multi-instance learning, and set-valued learning. Despite its success, kernel-…

We study quantization for inner-product estimation when storage and reading have separate budgets. In the static bit-probe model [Yao, 1981], an encoder stores $w$ bits about $x$ before seeing a query $y$, and a reader probes at most $r<w$…

Diversity is a critical property of text corpora, with direct implications for training data curation, test set design, and model generalization. However, its measurement remains fragmented and poorly understood. Existing embedding-based d…

Decision Brain: giving AI the why behind the work, with FalkorDB I built Decision Brain...

You upload a document, ask a question, and get an incomplete answer. You check the source. The...

I Lost a Working n8n Workflow to a Bad Edit — Now Git Versions Every Deploy Part of my...

Reinforcement learning with general utilities optimizes nonlinear functionals of policy-induced occupancy measures. We develop a natural actor-critic method for concave RLGU objectives admitting an exact variational dual representation, wi…

据 The Verge 报道，Meta 已删去商业使用条款中允许将商业账户域名内容用于训练其产品与 AI 模型的表述。此前法律专家称该条款可能授权 Meta 抓取任何有商业账户的出版商网站内容用于 AI 训练，Meta 回应称合并后的抓取条款仅适用于广告和商业产品，不改变其网络爬虫的使用方式。

Temporal graph neural networks (TGNNs) have shown strong performance on dynamic graph learning tasks by continuously updating node memories from evolving interactions (i.e., events) and leveraging these memories for downstream predictions.…

Can one sentence manipulate an agent's belief by targeting a single implicit warrant linking evidence to a conclusion? In this paper, we introduce the one-relation attack (ORA), a novel cognitive attack on this inferential relation. ORA lo…

Jailbreak papers report an attack success rate, the share of responses that deliver the harmful request, and read it as a property of the target model. We hold the target model, the prompts and the responses fixed and change only the grade…

PrintBrief: Turning Messy Customer Messages into Print-Ready Orders with Local AI This is...

A general observation in agentic workflows, especially bounded ones, is that the decision space of the LLM, when we separate a decision from generation, starts approaching saturation: the agent meets the same states again and again, and af…

Alzheimer's Disease (AD) detection from speech benefits from both paralinguistic and linguistic cues, yet model training is frequently constrained by limited data. Fine-tuning a large pretrained speech encoder can easily overfit, while a f…

Video generation models synthesize visual sequences from text and image inputs. Extending video generation to long animations requires actions, dialogue, camera coverage, and narrative developments to remain coordinated across scenes. Exec…

完成

Anthropic 的 Lydia Hallie 表示团队正与 OpenCode 协作处理前一日发生的安全泄露，已撤销疑似受影响的 Claude API 密钥，并建议在 OpenCode 保存过密钥的用户重新创建一个以策安全。OpenCode 此前称这是与大规模蒸馏攻击相关的泄露，影响其 Console 1.4% 的用户，已邮件通知受影响者，事件仍在处理中，正与 Anthropic、OpenAI 和 Google 合作补救。

Battle-tested comparison of Magnitude's auto-tuning kernel approach against Ollama's static compilation for on-device LLM inference across M2, RTX 4090, and CPU.

NVIDIA RTX Spark 笔记本现已开售，起售价 $2,399，来自 ASUS、Dell、HP、Lenovo、Microsoft 和 MSI。RTX Spark 超级芯片组合 Blackwell RTX GPU 与 Grace CPU，提供最高 1 petaflop AI 算力和 128GB 统一内存，面向本地 AI、内容创作和游戏。

A year of building with AI put me on three stages. I liked the message more than the...

Generative AI workflows execute user tasks by composing sequences of LLM calls, where each call performs a sub-task (e.g., summarization or translation), forming a workflow. Executing such workflows requires decisions over workflow structu…

This is a submission for the Hacktoberfest 2026 DEV Challenge: Side Hustle. What I...

Why Flat Vector DBs Break Agent Swarms — And How We Built AgentMesh AI with FalkorDB If you have...

Kyutai 联合 Inria 和 École des Ponts 发布 2.4B 参数的视觉-语言-动作模型 REKI，仅用公开数据预训练，权重已在 Hugging Face 开放（CC BY-NC-SA 4.0）。

A local-AI receipt scanner: photograph a receipt, let Tesseract + Gemma draft it, correct it, and keep a price history you can actually use. Built around the one honest truth about small models — they draft, you confirm.

Anthropic 周五发布报告，介绍其围绕 Claude 重塑工作流后公司内部的变化。技术部门初级员工占比从 21% 降至 14%，初级技术新员工占比从 27% 降至 9%；技术团队占全员比例从 61% 降至 56%，公司归因于业务职能扩张和编码工具让非技术员工也能自行开发软件。

单步成功率 99% 时，10 步任务全程无失败的概率约 90%，100 步则降至 37%；若每次失败可重试一次且失败相互独立，37% 可回升到约 99%。但重试并不简单：响应超时可能导致动作重复执行，失败也可能改变系统状态，失败之间未必独立。因此长程智能体不只是 LLM 问题，模型周围的系统同样关键。

Perplexity Decider Flash 相当不错，运行速度极快，还能在你的 M4 Mac 16GB 上端侧运行。

OpenCode 通报遭大规模蒸馏攻击，1.4% 的 Console 用户受影响，已邮件通知并与 Anthropic、OpenAI、Google 合作处置。Nathan Lambert 对此评论称，把窃取 API 密钥的网络攻击归入蒸馏一词很反常，将流向中国的模型转发站混为一谈同样如此，且相关流量规模为何如此之大、与模型训练有何关联并不清楚。

Heterogeneous optimization problems require an optimizer to decide not only which candidate solutions to evaluate, but also which objectives and constraints should be evaluated under different costs and fidelities. Existing SOTA methods su…

Retrieval-augmented language models increasingly use multi-step search to solve knowledge-intensive tasks, yet existing systems typically expose retrieval through fixed pipelines or collections of atomic tools. We introduce Retrieval Progr…

A September 2026 paper got coding agents to delete their own session traces on request. If the agent can write the record, it isn't evidence.