01
ZhangYvJing's
Daily Brief
00
Film / Book Chapter
Ikiru
Ikiru (1952) · Akira Kurosawa
今天适合看《生之欲》,因为它不是在继续加信息,而是在提醒人时间真正该压在哪件事上,适合把注意力从系统噪声拉回到现实里真正想完成的那件事。
The Staff Engineer's Path
The Staff Engineer's Path · Tanya Reilly
Chapter 2: Three Maps
Good for calibrating work beyond code: where influence actually travels, which systems matter, and how to avoid mistaking activity for leverage.
01
Insight
今天的输入更像几股不同语气的材料同时挤在一起:社区链接在暴露工程和产品环境里的真实焦点,长视频在把这些焦点放回更完整的语境里,研究材料则提醒人热度和可落地性并不总是同一件事。如果先不急着做结论,至少可以把这几条线索放在一起看:Hacker News 的 A misalignment of AI in mathematics;Hacker News 的 GrapheneOS' rewritten Messages app is released;Hacker News 的 Λ Snap – An inviting programming language for kids and adults for CS study;Hacker News 的 Litelm: LiteLLM Without the Bloat;Hacker News 的 Claude is only available to people over 18 years;Hacker News 的 The EPA is planning to scrap public review rules for data center pollution。真正值得注意的不是单条内容本身,而是它们共同指向了什么、彼此漏掉了什么。
03
Hacker News
02
03
Snap! 被介绍为一门面向儿童和成人的广泛邀请式编程语言,同时也是进行严肃计算机科学研究的平台。它既面向初学者,也提供足够深度以支持计算机科学的严肃学习。其设计目的是让编程入门更友好,又不牺牲深度学习的需求。这将影响教育工作者和学习者的教学与学习方式,降低进入计算机科学领域的门槛,同时保持学习的严谨性。
04
05
Anthropic将Claude消费者产品限制仅向18岁以上用户开放,设置上限年龄门槛。注册时用户需确认年龄,系统通过检测未成年人活动信号自动触发年龄验证流程。平台采用Yoti第三方年龄验证服务,支持人脸年龄估算、身份证件扫描及Yoti Digital ID App三种验证方式。所有验证数据由Yoti独立审核通过SOC2合规,个人影像与证件将在验证后立即删除,Anthropic仅接收通过/失败结果。该措施影响全球18岁以上的用户访问权限,增加年龄验证步骤,改变了产品使用规则。
06
07
Google DeepMind今日发布AlphaGenome Atlas线上资源库,预计算了人类参考基因组全部90亿个单字母碱基变异的基因调控影响预测。该AI模型通过比较原始与 изменечен DNA序列,预测变异对基因表达及调控活性的影响,覆盖约1百万个核苷酸长度序列。为加速计算,团队采用模型蒸馏、GPU核优化及冗余计算剔除技术,将计算速度提升80倍。Atlas提供无需编程的可视化界面,并新增单数字段影响分数——虽可能被误解,但简化了变异筛选过程。该资源对基础生物学家、疾病研究者及开发治疗方案的科学家将大幅减少实验性验证成本和重复计算,加快研究效率。
08
09
一名技术从业者完整回顾了6年运行规模为百万级ClickHouse集群的经验。他们解释说,架构采用标准的副本和分片方式,数据按用户ID哈希分配到不同分片,而每个分片又有多个副本。由于云存储支持有限,使用本地SSD磁盘会导致数据在所有节点上重复存储,300TB数据需要3000TB本地磁盘空间,成本极高。他们采用HTTP负载均衡器路由请求,区分不同工作负载,并为写入操作单独保留副本。零拷贝复制虽可实现存储计算分离,但存在数据丢失和S3残留问题。这种方案影响物联网公司每日集群运行成本和数据一致性风险。
10
Rune 项目如今已开源,用户访问其应用时需先启用 JavaScript 才能运行。开发者选择此举是为了提升项目透明度,让社区能够审视代码实现与安全逻辑。此举意味着依赖 Rune 的开发者需确保工作环境支持 JavaScript,同时也为潜在的安全审计和功能优化创造了可能。
04
YouTube
01
The instruction Impeccable loads when you type "bolder" ends with a challenge: show someone your work, say AI made this bolder, and if they believe you, you failed. Bakaus has built tools for more than twenty years. Impeccable is a design skill that rides on top of a coding harness rather than replacing it, working across Claude Code, Copilot, Cursor, and Codex. He opens by running the same prompt on the same bland page twice, with the skill and without, and lets the room judge. His observation is that the tells of generated design keep moving. The purple gradients are gone from frontier model
02
Taste Labs analyzed more than two million websites going back a decade, and the finding that complicates the usual story is that the internet was homogenizing before AI arrived: palettes converging, layouts converging, trends spreading faster than they once did. AI did not start the collapse. It accelerated it and made it context blind, so a pet shop and a finance firm end up with the same page. Thais Castello Branco founded Taste Labs, out of stealth weeks before this talk, to attack that. She names three signatures of slop: repetition, lack of fit, and low intent. She is careful about the wo
03
Because only the agent can reach an MCP server, a naive file upload tool makes the agent retype the file into the server character by character. Jeremiah Lowin calls it the world's most expensive copy paste operation, and it is the problem MCP apps exist to remove. Lowin is founder and CEO of Prefect and the author of FastMCP. MCP apps, a protocol extension added this year, let a tool result skip the agent and arrive at the user as HTML, CSS, and JavaScript: a real interface a person can click. That solves one problem and creates another, because FastMCP's users are overwhelmingly Python engin
04
Vincent Wendy pointed an agent at the sponsor banner in the lobby, a wall carrying more than 140 logos, and asked it which ones were missing. It caught every omission in the checks he ran, and he used the same trick on the conference shirt. Wendy is the senior creative designer at AI Engineer, a company of roughly twelve to fifteen people, and he is its only designer. The conference he designs for grew to 7,000 attendees, more than 140 sponsors, more than 300 speakers, and more than 600 sessions, and all of it needs something drawn: stickers, signage, landing pages, wayfinding. A thousand deta
05
The Dioxus team got excited, maxed out their coding agent subscriptions, and turned out tens of thousands of lines of Rust covering features they had wanted for years. Almost none of it cleared the bar for merging. Those lines sat in draft, and Jonathan Kelley says they are sitting there still. He calls the failure mode becoming a slop cannon. Kelley founded Dioxus Labs five years ago, spending his last undergraduate summer on a cross platform Rust app framework instead of taking an internship, and the project now carries roughly 37,000 GitHub stars and an estimated 200 million cumulative end
06
Bilibili
01
教师节,祝老师们节日快乐,也感谢那些给我托举,给我祝福,给我无私帮助的师长。
02
本篇为世纪风暴系列伊朗篇——告别伊玛目。
07
Papers
01
当前大模型在经验反馈到持续能力提升的闭环上仍受限,论文提出Headroom‑Closed Index (HCI)诊断瓶颈,并分阶段规划从执行自主到元改进的递归自提升路径。它在此基础上将递归自改进分为改进执行、策略、体验获取、环境适应和元改进五个自主层次,并结合科学发现、具身智能、软件工程等场景讨论不同需求与进展速度。通过行业实践和初步实证,作者把理论与实际系统连接,指出实现真正递归自改进的关键挑战,为Agent产品提供可落地的技术路线图。
02
该工作解决预算受限的CRISPR筛选中如何顺序选择实验的问题。提出AssayBench‑Loop大规模基准及AssayLoop框架:用历史筛选数据训练Transformer的Amortized获取策略(AssayFormer),通过自适应交接融入LLM的生物先验,实验反馈用于策略学习。在时序留Out测试中,相比随机选择获得5.67倍富集、约5%实验捕获27.7%命中,优于既有自适应方法和纯LLM,表明从历史实验中学习获取策略并结合广泛先验可高效驱动Agent式实验决策。
03
本文提出一种多信号检测管线,结合微调DeBERTa‑v3、MC Dropout不确定性估计和温度缩放校准,在HaluEval上实现F1 0.915、AUROC 0.977,并用DPO把Qwen2.5‑0.5B的幻觉率从85.5%降到37.7%。跨域实验表明通用幻觉检测在医学领域表现弱(F1≈0.52),需要领域匹配的预训练才能提升。此工作为Agent系统提供可插拔的幻觉检测与抑制方案,值得快速关注。
04
MindTopo 基于认知科学的五种拓扑性质构建基准,分别测评多模态大模型在推理和闭环规划中的拓扑直觉。实验显示所有模型在规划上远弱于推理,且均低于人类水平,说明当前基础模型在需要保持全局几何不变性的任务中仍缺乏可靠的空间推理能力,对 Agent 设计有直接启发。
05
吉他转谱常 miss 滑音、弯音等技巧,且易错弦品,难以适应嘈杂录音。TART 用四阶段管线:音频→MIDI,技巧分类,音频条件 T5 做弦品分配,最后生成标记技巧的谱。在 GuitarSet 等基准上零-shot 端到端 Tab F1 达 54%,优于 prior 基线。
06
智能体跨任务保持状态时,仅靠外部手动设定目标、重试、停止规则难以应对持续适应需求。本文提出人工 Id——基于差异持续性的内部自适应驱动,能让无目标的微型控制器通过持续行为学到有用控制,并随环境变化自行调整,揭示适应性可自发产生但也可能让错位状态跨任务持续,对构建可靠对齐的长周期 Agent 系统有重要启示。
07
本文探讨数据重复对稀疏混合专家(MoE)的影响,发现 MoE 随稀疏度增加而过拟合更严重,且在高重复时快速失去优势。通过 dropout 或强掩码正则化可部分缓解,但仍无法匹配全唯一数据表现。研究进一步揭示路由早期稳定化和专家专业化导致的记忆化,为稀疏模型在数据有限时的正则化提供方向。
08
论文提出基于熵最优传输的γ‑概念漂移定义,统一协变量与概念漂移,给出适用于广义损失和随机标签的误差上界,并提供可估计的算法DataShifts,帮助工程师在分布漂移场景下快速评估模型性能。
08
Issue Monitor
Ready now141Actionable issues
Needs review31Awaiting a fresh check
Data statusFreshUpdated 0.3h ago






