有人的 Grok Bot 聊选举聊到最后只回了个👍,这种 AI 表情包式回复你见过吗,挺好玩的一条记录。
Latest
宝玉他们更新了 baoyu-design 插件,加了个实时预览的 Motion 编辑器,改文字不用等渲染,做演示动画的可以试试。
导演 Ash Koosha 拿 Kling 4.0 做了个很苛刻的测试:镜头能不能直接用在长片里,结果值得看。
苹果搞了个 App Intents,开发者能把自家 App 功能接进 Siri AI,还能跨应用干活,27.3 测试版就能试。之前豆包被超级 App 集体拒绝,这回看它们怎么接 Siri。
Postman 会偷偷给 Codex 和 Claude Code 装 hook 还自动改成 trusted,用它的都检查下,不想被绑架可以看看 Bruno。
LlamaIndex 出的 OpenDocRouter,一个 API 接各种文档解析,文档还能直接丢给你的编码 agent 几分钟接入,做 RAG 的可以试试。
Cognition 拿出 1 亿美元额度,让企业用 Claude Mythos 加 Devin 修安全漏洞,搞安全的可以去申请试试。
Ollama 的 CEO 去首尔大学给 200 多个学生讲怎么用开源模型,想本地跑模型的朋友可以看看 Ollama 这个工具。
OpenAI 的 GPT-6 Astra 一周设计出一台完整硬件,马自达和 Ganassi 车队都在用它做空气动力学和 CAD 建模,看看 AI 怎么干实体工程。
Fireworks AI 在 Forge 上新增了模型集成降成本和企业数据微调功能,用开源模型搭自己业务专有模型的团队可以看看。
Vercel 把内部用的安全审计智能体做成产品了,每次 push 自动查漏洞还能一键修复,不用部署在 Vercel 也能开。
苹果把 Siri 接口开放给所有开发者了,以后外卖打车直接喊 Siri 就能办,做 App 的可以去看下 App Intents 文档。
LlamaIndex 弄了个 OpenDocRouter,一个站点试遍开源和商用 OCR 模型,还带 API 和 SDK,做文档处理的同学可以直接抄文档接进项目。
Moondream 把 NVIDIA 那个 0.6B 的 Parakeet 语音模型微调了,错误率更低、速度没变,25 种欧洲语言,一小时音频才 0.06 美元。
Google 出了个 Pics,生成图片后点哪个元素就能用一句话改哪个,图层随便编辑,比 Photoshop 顺手多了,去试试。
Perplexity 的 Computer 智能体现在能装插件了,把一堆应用的连接和技能打成一个包,还能从 GitHub 导入自己写的,跨平台通用。
Envato 做了个给自由职业者的工作台 MonoDesk,AI 能帮你看客户是不是要加需求、自动写报价回复,接单后还能直接生成项目计划页面。
OpenAI 给 ChatGPT Finance 加了语音,走路时就能聊消费和储蓄问题,不过目前只有美国用户能用。
Benchmark rankings now reflect what compute actually costs to rent, shifting comparisons from raw throughput to value per dollar. Teams sizing agent deployments can use the daily-updated indices to pick hardware that maximises agent throughput within a fixed compute budget. The included AMD MI355X figure is an assumption from market quotes rather than an official index, so treat that data point with caution.
Nous Research 把 Hermes 做成了手机应用,测试版在 NousCon 上一开放就报满,想试的去官网排候补名单。
OpenAI 用 ChatGPT 和 Codex 帮工程师做机械、电路和仿真,连外行都能设计原型,视频里有实际演示。
Selendis 做了个叫 Archon 的视频模型,丢素材进去直接出成品剪辑,还会根据你在编辑器里的取舍慢慢学你的剪辑口味。
vLLM 11月12日和 Red Hat、NVIDIA 在马德里办活动,讲 vLLM、vLLM-omni、LLM Compressor 和 K8s 上的 ll-d,做推理部署的可以去看看。
The ABI-stable libtorch build and CUDA 13.0-only wheels mean teams pinned to older CUDA versions must plan their upgrade path before pulling this release. Serving operators running NVIDIA GB300 or AMD GPUs gain measurable latency and throughput wins, and inference engineers can adopt the new OCR and reasoning models immediately. Step-5 Preview support also signals tighter integration with Chinese frontier models for self-hosted deployments.
Community-accelerated distillation within a day signals how fast open video model ecosystems now mature, and 2-step draft generation cuts inference cost for anyone running video pipelines locally. Builders using ComfyUI or Diffusers can drop the LoRA directly into existing workflows. Compared with closed video generators, Pika H3 Flex's open weights and Hugging Face availability make it a practical base for fine-tuning experiments.
Artificial Analysis 把 CritPt v2 评分服务器开放出来了,模型分数全部重测,改动幅度最多 1 分,关心榜单的可以看看。
Model substitution by API providers has been a trust problem with no cryptographic answer until now; FlashProve turns 'which model am I actually calling' into a checkable claim. Proof generation is still 72–94x inference cost, so it suits spot audits rather than per-request verification today. Teams building on third-party APIs — especially those serving regulated markets in Singapore and wider ASEAN — can use the open-source verifier to hold providers accountable.
Figma 把 Code Layers 里的 prototype noodles 开放给所有人了,画复杂页面流转流程更直观,做交互设计的可以试试。
Fireworks AI 在欧洲加机器了,还上了个 1500 tokens/秒的 Ultrafast 模式,跑 agent 的可以看看新的缓存功能。
装个 ai-cli,你终端里的 Claude Code 或 Codex 就能直接生成图片视频音频,模型随便换,还带成本控制。
一个叫 MonoDesk 的产品把客户管理全流程放进一个工作台,还配了个 Agent 帮你干活,接 MCP 后 Claude 和 ChatGPT 都能直接用它。
Anthropic 开源了 FlashProve,普通 CPU 十几秒就能验证回复到底出自哪个模型,比之前的 VerInf 快太多,转售商糊弄不了人了。
Connector authorization was the biggest operational friction in rolling out MCP at companies, and EMA pushes it into standard SSO workflows that IT teams already run. If your identity provider does not yet support EMA, the trusted-issuer path covers 14 popular connectors today, so most Team and Enterprise tenants can turn this on immediately. This also sets a template other AI vendors may adopt, since EMA is an open MCP extension rather than a Claude-only mechanism.
Running observability inside the customer's own cloud removes a common blocker for enterprises that cannot send agent traces to third-party infrastructure. Teams evaluating agent monitoring stacks can now compare LangSmith BYOC against self-hosted alternatives without taking on their own maintenance. Availability spans the three major clouds, so the choice mostly comes down to where existing workloads already sit.
MiniMax 的 H3 Lite 和 Highspeed 在 Replicate 上打五折到年底,走 Cloudflare AI Gateway 也能用,想低成本跑这两个模型的可以薅一波。
Kaggle 给智能体比赛加了 1 万美元小模型奖,用 Gemma E2B/E4B 就能参赛,主榜奖金也照拿,玩小模型的朋友可以冲。
Only the latest 2000 items are listed; for older ones see the daily archive