hotai.news2026-08-22 收录的 20 条 AI 快讯，按采集时的来源记录回看。消息仍需核对原始发布页。
这期内容仍需回到原始发布页核对。
Most teams treat ‘which model’ as the important decision. The harness engineering literature keeps pointing somewhere else. In LangChain’s Terminal-Bench experiment, changing only the harness—same model throughout—moved a coding agent from roughly 30th place into the top 5.
来源链接可在原站核对查看原文 ↗DeepSeek 将于明日起调整 API 计费规则，周末全天不再区分峰谷时段，统一按低谷价收费，简化计费逻辑，开发者任何时段调用都可享受低价。
来源链接可在原站核对查看原文 ↗A study from researchers at Princeton University and UC San Diego finds that so-called skills make AI agents better mainly through structured workflows, not through added knowledge. But as the skill library grows, agents have a harder and harder time finding the right set of instructions.
来源链接可在原站核对查看原文 ↗本文上手体验雷鸟全新发布的雷鸟 iO AI 眼镜，它主打首创全天智记功能，可自动记录整理日常信息，翻译、提词、录音转写等功能也做了 AI 优化，佩戴舒适续航出色。#雷鸟 iO AI 眼镜#
来源链接可在原站核对查看原文 ↗The Information 此前报道英伟达计划年底向中国客户小批量出货定制 AI 芯片，英伟达官方明确否认该报道，称目前及路线图中均无针对中国市场的 LPU 产品
来源链接可在原站核对查看原文 ↗Netflix pitted its years-old recommendation engine against an in-house language model called GenRec and says it got better results. Instead of relying on thousands of hand-crafted features, GenRec converts viewing behavior into plain text. Netflix itself calls it "an early but promising step."
来源链接可在原站核对查看原文 ↗RayNeo is launching a new headset without a camera or speakers.
来源链接可在原站核对查看原文 ↗Researchers at the UK AI Security Institute used psychometric methods to show that popular safety benchmarks for language models don't measure one consistent trait. Blanket blocking of requests can artificially inflate a safety score even as the model gets less useful day to day. The study also offers a method for catching models that act more cautious during tests than they do in normal use.
来源链接可在原站核对查看原文 ↗全球最大 AI 聚合平台 OpenRouter 于 8 月 20 日上线匿名 AI 模型 Ox Alpha，免费开放 1 周时间，目前种种线索指向智谱公司。
来源链接可在原站核对查看原文 ↗CNBC 昨日（8 月 21 日）发布博文，报道称在 Anthropic 的首次公开募股（IPO）招股说明书中，把美国公众对人工智能的抵制情绪列为关键风险因素。
来源链接可在原站核对查看原文 ↗OpenAI 宣布在未来 3 个月内，将 GPT-5.6 Sol 的 API 与 Credit 调用价格下调 20% 以上。
来源链接可在原站核对查看原文 ↗豆包输入法鸿蒙版现已在 App Gallery 应用商店开启尝鲜，版本号为 0.9.1（901），时间为 8 月 22 日至 9 月 21 日。
来源链接可在原站核对查看原文 ↗金融媒体 mitrade 昨日（8 月 21 日）发布博文，报道称 Anthropic 最早可能会在本月（2026 年 8 月）底公开提交 IPO 申请，预估融资规模持平或者超过此前 SpaceX 创造的纪录。
来源链接可在原站核对查看原文 ↗在 2022 年离职前，萨莱克曾长期负责谷歌的 TPU 业务，主导并交付了前七代 TPU 芯片。
来源链接可在原站核对查看原文 ↗科技媒体 Windows Latest 昨日（8 月 21 日）发布博文，报道称在 Windows 11 系统中，微软计划增加统一内存管理，精细化管控图形处理和 AI 任务预留内存。
来源链接可在原站核对查看原文 ↗“IT早报”时间，大家好，现在是 2026 年 8 月 22 日星期六，今天的重要科技资讯有：
来源链接可在原站核对查看原文 ↗LinkedIn actually announced a “Seems like AI slop” button on July 30th, and the company says “over a million people” have clicked on it.
来源链接可在原站核对查看原文 ↗字节豆包团队最快将于下周发布一款对标腾讯 WorkBuddy 的办公类产品。（新皮层）
来源链接可在原站核对查看原文 ↗Anthropic forbids its Claude models from generating sexually explicit content. But a series of tests conducted by TechCrunch found that it didn't take much to get past the restriction.
来源链接可在原站核对查看原文 ↗Part I: costs and how much revenue you would need to justify those costs. Estimate #1 comes from Peter Berezin, Chief Economist at BCA Research. 
来源链接可在原站核对查看原文 ↗