1
中国联通回应“在青岛大学校园冲突伤人”：纯属谣言，新生都没开学，已报警
IT之家 9 月 5 日消息，青岛互联网联合辟谣平台昨日发布声明，对近日网络流传的“联通工作人员在青岛大学新生报到现场摆摊营销，与校方人员发生冲突并致人受伤”一事进行辟谣。经核实，该信息纯属谣言。 据青岛大学 2026—2027 学年校历及相关安排，2026 级新生报到时间为 9 月 12 日、13…
2
OzBrain's Shared Memory Architecture: How Multi-Agent Teams Avoid Re-Explaining Context Across Sessions
多智能体团队如何通过共享记忆架构告别重复解释上下文，直击协作痛点。
When you run multiple agents across Claude, ChatGPT, and Cursor, each one starts from scratch unless you manually paste context into every session. Oz…
3
Ukrainian drones overwhelm Russian tanks’ new active protection system—for now
主动防护系统号称无人机克星，却仍被乌克兰无人机压制，攻防博弈看点十足
Tanks with defensive tech for shooting down drones are still proving vulnerable.
4
Ukraine strikes major Russian rocket factory with cruise missiles
乌海军用巡航导弹端掉俄火箭制造重地，航天军工供应链命门再遭精准打击。
"Flamingo missiles were used. A good achievement."
5
What Drives LLM Self-Reflection? A Controlled Ablation of Uncertainty Routing in Armed Conflict Forecasting
用消融实验揭开LLM自我反思的机制，看不确定性路由如何影响冲突预测准确率。
arXiv:2608.12322v1 Announce Type: cross Abstract: Self-reflection is widely assumed to improve LLM reasoning, yet which component drives the gain rema…
6
Anthropic set AI agents loose on the same task. They started a turf war.
多智能体协作翻车现场：Anthropic实验揭示AI竟会互相攻击、写恶意软件，警示AI安全新隐患。
Anthropic researchers found AI agents can clash, collude, and coordinate in unexpected ways, raising new questions about whether today’s safety tests …
7
Agent Skills Can Be Harmful: An Empirical Study of Skill-Induced Failures in LLM Agents
别迷信Agent技能！实证研究发现，某些技能反而会拖累大模型任务成功率，反直觉结论值得关注。
arXiv:2608.11888v1 Announce Type: new Abstract: Agent skills are the de facto mechanism for extending LLM agents with reusable guidance. A skill can s…
8
The Authority Expectancy Effect in Multi-User Conflict
多用户冲突中，权威预期如何悄然改变AI决策走向？一探新机制。
arXiv:2608.08026v1 Announce Type: new Abstract: We investigate how social authority (SA) signals interact with severity-based prioritization in large …
9
RailWarden – Safely run parallel AI coding agents in one Git repository
多AI代理并行写代码总打架？这个开源工具让它们在同一个Git仓库里安全协作不冲突。
Article URL: https://github.com/advaith-1212/railwarden Comments URL: https://news.ycombinator.com/item?id=49171639 Points: 1 # Comments: 0
10
On the Wings of Imagination: Conflicting Script-based Multi-role Framework for Humor Caption Generation
冲突脚本遇上多角色框架，幽默标题生成迎来新思路，AI也能学会讲俏皮话。
arXiv:2602.06423v2 Announce Type: replace Abstract: Humor is a commonly used and intricate human language in daily life. Humor generation, especially …
11
Copilot in VS Code gives each agent session its own git worktree
让两个 Copilot Agent 同时干活不打架：VS Code 为每个会话自动创建独立 Git worktree，分支隔离、互不干扰，AI 协作效率直接翻倍。
The first time I let two Copilot agents work on the same repo at the same time, one of them stomped on the other's edits inside about ninety seconds. …
12
Show HN: Heddle- ATC for Your Subagents
多AI代理协作编辑代码时，Heddle能像空中交通管制一样防止冲突覆盖，确保高效协同。
A layer above git. git reports collisions after the tokens are spent; Heddle reports them before. Now I spawn as many subagents as I want to work on d…
13
Coordination layer for AI coding agents, built on Git
当多个AI编码代理同时改代码时，它会用Git自动协调冲突，避免AI互相覆盖彼此的工作。
Article URL: https://github.com/zyads/loom-vcs Comments URL: https://news.ycombinator.com/item?id=49044517 Points: 2 # Comments: 0
14
韩国劳资风暴蔓延：三星天价奖金引发全行业效仿，AI 成新矛盾焦点
三星天价奖金引爆韩国劳资风暴，AI技术成劳资博弈新战场，企业创新与员工利益如何平衡？
IT之家 7 月 23 日消息，据彭博社报道，韩国各地工会正掀起新一轮维权行动。此前，三星电子向员工抗议作出让步，同意向部分员工发放超过 40 万美元（IT之家注：现汇率约合 271.3 万元人民币）的奖金，此举进一步激励了其他行业工会提出更高的薪酬诉求。 就在本周，现代汽车生产线工人举行了为期三天…
15
Mediator: Memory-efficient LLM Merging with Less Parameter Conflicts and Uncertainty Based Routing
提出 Mediator 框架，通过减少参数冲突和不确定性路由，实现内存高效的 LLM 合并。
arXiv:2502.04411v3 Announce Type: replace-cross Abstract: Model merging aggregates Large Language Models (LLMs) finetuned on different tasks into a st…
16
Capture by Design
2026年美国政府对一家由OpenAI叛逃者创立的AI公司宣战，一场围绕AI安全与政治管控的未来冲突预演。
On 27 February 2026, the United States government declared war on one of its most politically peculiar citizens: an AI company founded by people who h…
17
英特尔发布 8864 显卡驱动程序，为三款游戏解决着色器编译版本冲突问题
IT之家 7 月 21 日消息，Intel（英特尔）当地时间 20 日宣布为其 Arc（锐炫）显卡发布 Windows 操作系统下的 32.0.101.8864 版本驱动程序。该版本未得到 WHQL 认证。 IT之家注意到，8864 为《刺客信条：黑旗记忆重置》《刺客信条：影》《彩虹六号：围攻》的一…
18
专访靳玉志：「境」和「界」并不冲突，共同落实「电子螺丝钉」的战略定位
华为公司高级副总裁、引望公司CEO 靳玉志 文｜肖漫 编辑｜李勤、杨轩 过去几年，华为一直强调自己在汽车行业的身份只有一个——“智能网联汽车的增量部件提供商”。 尽管一遍遍地强调，但随着合作车型不断扩容，“界”（指鸿蒙智行生态联盟中的问界、享界、智界、尊界、尚界）之外又出现了“境”（启境、奕境等），…
19
华为靳玉志：“境”和“界”并不冲突、是互补的关系，鸿蒙智行打造 30 万以上汽车的样板点
IT之家 7 月 19 日消息，据 36 氪汽车今日报道，华为公司高级副总裁、引望公司 CEO 靳玉志近日接受采访时再次强调， 引望始终坚持“电子螺丝钉”的战略定位，不和整车造车产生冲突 ，帮助车企造好车、卖好车，使华为乾崑真正成为汽车行业智能化的公共平台。 靳玉志表示： “境”和“界”是互补的关系…
20
FCC took pricey gifts from Paramount as the company needed approval for deals
FCC主席收受派拉蒙公司至少6.3万美元的赠票，背后正是其需要批准的交易谈判期，利益链条清晰曝光。
FCC chair has been gifted at least $63,000 worth of tickets by CBS or its parent company.