1
占 16 字节：曝微软 Win11 画图 / 照片向 AI 生图嵌入 128-bit 隐藏标识符
IT之家 8 月 26 日消息，科技媒体 NeoWin 昨日（8 月 25 日）发布博文，报道称安全研究员 Xusheng Li 通过逆向分析， 发现微软画图（Paint）和微软照片（Photos）会把 GDID 标识符嵌入到 AI 生成的图片中。 IT之家注：GDID 最早于今年 7 月报道，在针…
2
Show HN: A Ruby IDE running on a microcontroller
在仅512KB内存的微控制器上跑起了Ruby IDE，硬核嵌入式极客不容错过。
I built AREA512, software that lets you write, compile, and run Ruby code entirely on the M5Stack Cardputer. https://github.com/engneer-hamachan/area5…
3
Show HN: I let an AI agent brick my $2.51 ESP32 on purpose
让AI代理主动搞崩ESP32，却能精准定位传感器ISR栈溢出，支持千款开发板，嵌入式调试神器。
Article URL: https://github.com/GLechevalier/nff-core Comments URL: https://news.ycombinator.com/item?id=49366373 Points: 2 # Comments: 0
4
The Arduino UNO Q is an almost perfect Hermes Agent host
用Arduino UNO Q当Hermes Agent宿主，双芯片Linux+Arduino，4GB内存低成本搞定个人AI代理，省去VPS开销。
I’ve wanted to try out Hermes agent for a while now but I never wanted to run it on my own personal machine. I thought about maybe using a hosted VPS …
5
Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings
借密集语义嵌入给LLM提示装上低延迟“安全护栏”，看它如何兼顾速度与防护。
arXiv:2608.17556v1 Announce Type: cross Abstract: Large Language Models (LLMs) in real-world applications often face the risks of specially crafted pr…
6
Geometric Filtering of LLM-Generated Samples for Few-Shot Text Classification
用几何过滤筛掉低质量LLM生成样本，少样本分类精度立竿见影。
arXiv:2608.13866v1 Announce Type: new Abstract: Large language models (LLMs) can generate synthetic training data for text classification, but the qua…
7
TraVEL: Trajectory-Guided Video Embedding Learning for Driving-Video Retrieval
用车辆轨迹引导视频嵌入学习，一举提升驾驶视频检索精度。
arXiv:2608.13495v1 Announce Type: cross Abstract: Efficiently retrieving relevant clips from large-scale driving logs is essential for data curation, …
8
LLM-Based Embeddings for Program Analysis and Optimization
用LLM嵌入打通程序分析与优化，让代码理解更智能，优化更精准。
arXiv:2608.07894v1 Announce Type: new Abstract: Recent advances have highlighted the potential of machine learning, particularly Large Language Models…
9
SCALE: Scientific Concept Aggregation via LLMs and Embeddings for Fine-Grained Taxonomy Extension
用LLM与嵌入聚合科学概念，高效扩展细粒度分类法，科研知识管理新思路。
arXiv:2608.07254v1 Announce Type: cross Abstract: The increasing specialization of scientific research challenges existing classification systems, whi…
10
Show HN: Llmem – Local persistent memory for AI coding, no embeddings
本地持久记忆层，用BM25替代向量嵌入，给AI编程加上轻量记忆，15个工具即插即用。
Hi folks. This part of AI coding always felt really dumb: after working on many projects it repeats the same mistakes, uses awkward patterns or looks …
11
Beyond Initialization Loss: A Systematic Study of Token Embedding Initialization Strategies for LLM Vocabulary Extension
系统评估20余种初始化策略，为LLM词汇扩展提供实用指南。
arXiv:2608.03494v1 Announce Type: cross Abstract: Vocabulary extension is an efficient way to adapt pretrained large language models (LLMs) to new lan…
12
Douyin Multimodal Embedding Model Technical Report
抖音多模态嵌入模型技术报告，揭秘工业级搜索推荐背后的向量表示学习。
arXiv:2608.02148v1 Announce Type: cross Abstract: Multimodal representation learning is a cornerstone of modern AI. By encoding multimodal queries and…
13
Visualising Information Flow in Word Embeddings with Diffusion Tensor Imaging
用医学影像技术透视大模型词嵌入，跨界可视化信息流的新视角。
arXiv:2601.05713v2 Announce Type: replace-cross Abstract: Understanding how large language models (LLMs) represent natural language is a central chall…
14
LLM2Vec-Gen: Generative Embeddings from Large Language Models
不再丢弃大模型输出语义，用自监督生成式嵌入让LLM直接产出向量，重新定义嵌入学习。
arXiv:2603.10913v3 Announce Type: replace Abstract: Fine-tuning LLM-based text embedders via contrastive learning maps inputs and outputs into a new r…
15
Show HN: WiPhone Lab – Doom on an Open ESP32 Wi-Fi Phone
ESP32开源手机固件，刷机就能玩《Doom》，四兆内存跑出怀旧体验，极客DIY必看。
Article URL: https://github.com/mulfyx/wiphone-lab Comments URL: https://news.ycombinator.com/item?id=49114317 Points: 2 # Comments: 0
16
HN: I ran a 56M parameter LLM across three ESP32-S3 boards via ESP-NOW
在3块ESP32-S3开发板上用ESP-NOW分布式运行5600万参数LLM，低成本边缘AI新玩法
Article URL: https://github.com/wladimiravila/esp32s3-distributed-ai Comments URL: https://news.ycombinator.com/item?id=49111137 Points: 2 # Comments:…
17
铠侠出货 UFS 5.0 嵌入式闪存商业样品，预计 2026 年底量产
IT之家 7 月 29 日消息，KIOXIA（铠侠）当地时间今日宣布已开始出货新一代嵌入式闪存 UFS 5.0 的商业样品，预计 2026 年底量产。铠侠此前在今年 2 月宣布了 UFS 5.0 功能评估与互操作性测试样品的交付 。 JEDEC UFS 5.0 标准在物理层采用 MIPI M-PHY…
18
Show HN: Full complex text shaping and rendering on ESP32 with 320KB usable RAM
在不到10美元的ESP32上实现完整的OpenType文本整形与渲染，支持阿拉伯语、天城文等多种复杂文字，仅用320KB内存，无需预制资源，极具工程突破性。
Article URL: https://github.com/waruyama/flattype-cyd-demo Comments URL: https://news.ycombinator.com/item?id=49075016 Points: 3 # Comments: 1
19
Show HN: Codex Plays NetHack
通过终端录像观看AI Codex自动游玩经典游戏NetHack，展示人工智能在复杂环境中的决策能力
I'll add a link to the completed recording after the stream completes. Comments URL: https://news.ycombinator.com/item?id=49065825 Points: 1 # Comment…
20
群晖推出 RS1226+ 机架式八盘位存储服务器，基于 AMD 锐龙嵌入式 V1500B
IT之家 7 月 27 日消息，群晖科技 (Synology) 本月 22 日推出了机架式八盘位存储服务器 RackStation RS1226+。这一型号 面向拥有边缘和远程业务、需求精简且可靠数据管理的中小企业 ，另提供双电源版本 RS1226RP+。 RS1226+ 采用 19" 2U 外形规…