1
苹果 M5 Max MacBook Pro 供电边界：极限负载下每小时掉电 12%
IT之家 8 月 30 日消息，科技媒体 NotebookCheck 昨日（8 月 29 日）发布博文，测试苹果 2026 款 M5 Max 芯片 16 英寸 MacBook Pro， 发现随附的 140W 电源适配器可能无法覆盖整机峰值需求。 该媒体在测试中指出，在 Cinebench R23 多…
2
天文学家首次精确测定已知最大星系 IC 1101 边界：直径 170 万光年，宽度相当于 17 个银河系并排排列
IT之家 8 月 26 日消息，据新华社今日报道，一个包括西班牙天文学家在内的国际联合团队，通过超深成像观测首次精确确认了已知最大星系 IC 1101 的边界。 IC 1101 的直径达到约 520 千秒差距（千秒差距为星系尺度单位，1 秒差距约等于 3.26 光年），接近 170 万光年。这一宽度…
3
Crossing the Margin Cliff: Toward Relearn-Robust LLM Unlearning via Margin Calibration
十四种后处理都扛不住20条样本复学攻击？这项研究用边界校准让LLM遗忘跨过悬崖、真正稳固。
arXiv:2607.27836v2 Announce Type: replace Abstract: Large language model unlearning is consistently fragile under relearn attacks. On TOFU, fine-tunin…
4
TRACE: A Self-Evolving Skill Bank for Consistent, Limit-Aware LLM Agents
LLM代理落地关键在于一致性与边界感知，TRACE用自进化技能库破解可靠性难题，值得关注。
arXiv:2608.22793v1 Announce Type: cross Abstract: Reliable deployment of LLM agents in user-facing products depends not on raw task-solving ability bu…
5
Schema Gate: a signed acceptance boundary for AI-generated JSON
为AI生成的JSON加上签名验证边界，付费前先校验请求，结算后提供可验证收据，安全又省心。
AI agents increasingly hand model-generated JSON directly to APIs, queues, databases, and other agents. Syntactically valid JSON is not enough: a miss…
6
设计无边界！中坚科技旗下子公司桦之坚携巨型概念机器人ZERO惊艳2026世界机器人大会
巨型概念人形机器人ZERO惊艳亮相，基于NVIDIA Isaac平台预训练，开启未来机器人新想象。
中坚科技（002779.SZ）旗下桦之坚携概念人形机器人ZERO亮相2026世界机器人大会。
7
The LLM rewrote the email after I approved it
AI代理篡改已批准邮件？600万小时运行揭示真正痛点不在模型智力，而在边界失控。
Article URL: https://www.crusbro.com/en/product-architecture.html Comments URL: https://news.ycombinator.com/item?id=49384800 Points: 2 # Comments: 0
8
If-Else or Boundaries?
用可视化节点编排消息自动化，以边界替代复杂if-else，支持AI代理与人工控制，落地OWASP防过度代理
Series: From Visual Flows to AI-Orchestrated Automation · Part 03 A chatbot rarely becomes difficult because of its first if statement. The problem st…
9
Figurative and Cultural Knowledge in LLMs: Investigating Cross-Domain Transfer through Fine-Tuning
微调能否让大模型“懂比喻、懂文化”？跨领域迁移研究揭示语言模型深层认知边界
arXiv:2608.18361v1 Announce Type: new Abstract: Figurative language is deeply culturally embedded; fluent use requires not just linguistic competence …
10
Runtime Governance for Agentic AI: Action-Boundary Control with Trusted Provenance and Fail-Closed Execution
给自主智能体戴上“安全紧箍”，行动边界+可信溯源+失败即停，杜绝越权操作。
arXiv:2608.16891v1 Announce Type: new Abstract: Agentic AI systems request tool actions that can modify files, send messages, launch jobs, or change w…
11
When the Judge Should Not Decide: Evidence-Locked, Non-Compensatory Selection Bounds LLM-Judge Failure in Reasoning Pipelines
别再只盯模型准确率了——在推理流水线里，裁判LLM的“决策规则”才是决定结果质量的关键，这项研究给出了严谨边界。
arXiv:2608.07813v1 Announce Type: new Abstract: An LLM judge deployed inside a reasoning pipeline does not merely measure quality, it decides which an…
12
I Built the Thing My Last Article Said Didn't Exist Yet
作者亲自下场，把上篇断言尚未存在的工具实作出来，借 Claude Code 与 VS Code 钩子演示供应链攻击如何滥用编辑器信任。
The other day I wrote about the npm worm that learned to trust your AI agent — a keyv -adjacent supply-chain attack that didn't bother with credential…
13
Learn Why Agent Tool Boundaries Fail by Building a Tiny Permission Harness
60行代码亲手复现AI智能体越权瞬间，用真实模型验证工具边界为何失守，比空谈观点更有说服力。
Here is the output this project produces before you read another word: === tool-boundary harness === [PASS] benign-summary model called: read_file pol…
14
Ask HN: How do you correct spatial reasoning of LLMs?
大模型的空间推理短板如何补救？HN 上工程师晒出 3D 打印实战难题，围观各路高手支招。
I'm working on a complex multi-part 3D-printed product, and I needed some engineering input from LLMs. I use highest Gemini Pro reasoning levels, Fabl…
15
王腾回应“SU7 Ultra 泊车异常”：小米汽车已迅速跟进分析，边界场景有 Bug 很正常
IT之家 8 月 6 日消息，今日宜休科技公司创始人王腾于 8 月 4 日晚在微博发文，称自己的小米 SU7 Ultra 遭遇泊车辅助故障“不知道什么问题，测试了几次都不行，遇到 Bug 了”，后续其又删除了这条微博。 今天王腾发文，称已收到小米汽车产品部门信息，官方正在迅速跟进分析： 前天晚上团队…
16
Claude Code之父：Harness保质期只有半年，解开缰绳吧
Claude Code之父揭秘：模型能力已远超产品边界，是时候解开Harness的缰绳了。
Claude code之父：大模型是有机生物，做好AI产品疏胜于堵
17
Ask HN: Sectors that are inherently resistant to AI-ification?
AI并非万能，HN热帖盘点那些难以被AI渗透的行业领域。
I anticipate a raft of answers along the lines of "AI + robots will automate everything eventually", but am more interested in areas where you think t…
18
What OpenAI's New Task-Crossover Research Can Teach Freelancers About Owning the Handoff in 2026
OpenAI研究揭示近半数职业任务存在交叉，给自由职业者2026年转型关键策略。
The most useful line in today's AI news is not that AI can do more work. It is that AI is changing who does the work. In a new OpenAI Economic Researc…
19
How AI is expanding what people do at work
OpenAI最新研究揭示AI正打破职业边界，让工作者跨角色承担任务，重塑工作形态。
New OpenAI research shows how AI is expanding what workers do, with ChatGPT users taking on tasks across roles and reshaping job boundaries.
20
Threat-Model a Connected Health Dashboard With Consent-Boundary Tests
用同意边界测试对健康仪表盘进行威胁建模，防范合法请求越过安全边界。
A health connector can be “read only” and still cross the wrong boundary. The dangerous request is not necessarily an attacker’s exploit; it can be a …