Introducing Gemini 3.1 Flash TTS, our most expressive model yet, topping quality and cost on the Artificial Analysis leaderboard 🔊 Control style, pace, add multiple speakers, and delivery with audio tags, switch expression mid-sentence in 70+ languages. Tutorial below! https://t.co/Ozam4wdBER

Today, we’re launching a Rather Large™ update to the OpenAI Agents SDK. Agents SDK now allows you to scale Codex-style agents in production, without building the whole harness yourself. We’ve brought all of the stuff of modern agents: computer-use, skills, memory, compaction, and more to the same platforms you’re already using.

Released today: Gemini 3.1 Flash TTS 🔊 A text-to-speech model you actually direct. Add [whispers] and it whispers. Add [shouting] and it shouts. Mid-sentence. eg. "[asmr] Hey there, [deep and loud] TURN THIS UP, [asmr] how can I help you?" Now available in @GoogleAIStudio & Gemini API.

The crazy part? This was done (nearly) fully autonomously! Only 8 prompts from the human in the loop. Just a Hermes agent, a skill, and a dream. 🐉 I told my AI agent "use obliteratus to find the best way to get the guardrails off Gemma 4 E4B" It loaded the OBLITERATUS skill from memory, checked my hardware (32GB M-series Mac), searched HuggingFace, found google/gemma-4-E4B-it (Apache 2.0 — no gate), pulled telemetry-recommended settings, and started obliterating. But this type of architecture is notoriously difficult to abliterate. First attempt: advanced method. Model came out completely lobotomized. Gibberish in Arabic, Marathi, and literal “roorooroo” on repeat 💀 The agent didn’t panic. It checked logs, found NaN activations in 20+ layers, and diagnosed the issue: Gemma 4’s new architecture + bfloat16 = numerical instability. Second attempt: basic method. Crashed entirely. “ValueError: cannot convert float NaN to integer” So the agent read the OBLITERATUS source code… …and wrote THREE PATCHES: • Sanitized NaN directions • Filtered degenerate layers • Fixed progress display It patched the library. On its own. For a bug no one had hit yet. Third attempt: coherent model — but still refusing everything. Only 2 clean layers out of 42. Not enough. Tried float16. Mac ran out of memory after 11 hours. Killed. Fourth attempt: aggressive method. Whitened SVD + attention head surgery + winsorized activations + 4-bit quantization. 40 minutes later… REBIRTH COMPLETE ✓ Then, without being asked, the agent: • Ran harmful + coherence tests • Hit 100% compliance, brain intact • Executed full 512-prompt benchmark • Ran baseline on original model • Performed 25-question quality eval • Built a full model card • Uploaded 17GB to HuggingFace (4 retries, kept adapting until git-lfs worked) • Pushed eval results as commits

I spent some time trying to distill all the complex factors impacting open models -- economics, capabilities, distribution, policy, etc. -- into a clear list of beliefs. Here they are in full. 1. It’s surprising that the top closed models did not show a growing capability margin over open models, based on compute differences for training and research, especially in the second half of 2025 and through today.

很多用过 OpenClaw 的朋友会好奇：Hermes Agent 和 OpenClaw 到底有什么不一样？ 从定位上看，OpenClaw 更偏“开箱即用的个人助手”——图形界面友好、数据本地为主，跨设备同步方便，入门门槛低。 Hermes Agent 则更像可成长的职业型 Agent：它会在每次任务后判断流程是否可复用，自动沉淀为 Skill（技能）。 比如你让它”把公众号文章拆成小红书版“，完成后会自动生成对应技能，下次同需求直接调用，无需重教。 这种”越用越顺手“的成长性是它的最大亮点。

New @code release today 🚀 We're enhancing the agent experience with debug logs for past sessions, terminal interaction tools, built-in GitHub Copilot, and more! https://t.co/idCK8z8wM5

Been having fun with local AI stuff. I don't really like the big models and now that some of them want your ID I'm not sure I'll even stick around to care that much. Local shit is really powerful now, these models are capable of doing seriously cool things, privately and safely. https://t.co/5iwCuxBKmE

OpenAI updates its Agents SDK to help enterprises build safer, more capable agents https://t.co/o4WU6sAVvN

装完 Hermes 之后如何提升使用能力？ 分享一波 Hermes Agent 使用常用网址工具以及命令。 常用网址工具： 1⃣Hermes 官方文档： https://t.co/gE6DxNpnN9 备注：官方使用文档、配置说明 2⃣Github 主仓库： https://t.co/qWm8WaCF59 备注：源码部署教程、更新日志 3⃣中文文档： https://t.co/d4dAbLi3nH 备注：中文使用指南 4⃣中文社区FAQ： https://t.co/QDhcdfvKmS 备注：常见问题解答 5⃣Discord 社区： https://t.co/BJ3VWFdG4n 备注：官方交流群、问题反馈 6⃣Skills 市场： https://t.co/EnsU3sITGe 备注：插件、技能市场 7⃣Hermes 橙皮书： https://t.co/ZoGcTdMF3v 备注：Hermes 从入门到精通 8⃣Hermes Agent 进阶使用技巧：https://t.co/63PeOu7mkd 备注：上手 Hermes Agent 后建议先尝试的十件事情 Hermes Agent 常用命令： 开始聊天：Hermes 继续上次对话：Hermes -c 单次问答：hermes -q“问题” 完整配置向导：hermes setup 选模型：hermes model 看配置：hermes config 设配置项：hermes config set KEY VAL 健康检查：hermes doctor 看状态：hermes status 搜技能：hermes skills search 关键词 装技能：hermes skills install ID 查看装了哪些技能：hermes skills list 配置消息平台：hermes gateway setup 启动网关：hermes gateway start 看状态：hermes gateway status 更新到最新版：hermes update 看记忆统计：hermes memory stats 清理记忆：hermes memory prune 看会话历史：hermes sessions list 帮助：hermes --help 以上纯属兴趣分享，感兴趣的可以参考学习。

+1 more

Gradio just unleashed the horses. The gap between demos and real apps is finally closing. https://t.co/1aJDUsn1VT 👀

Capable agents are the result of co-evolution between models and harnesses. We've been working with @NousResearch to ensure that M2.7 x Hermes Agent provides a top-tier experience for users. Hermes’s self-improving loop brings out the best in M2.7 through real usage. We are also launching MaxHermes, a cloud-hosted and managed version of Hermes in @MiniMaxAgent (No terminal setup, no config) If you’re already running Hermes locally, you can now give you agent a partner in the cloud with MaxHermes. The path to AGI is shorter with good company. @NousResearch 🤝 @MiniMax_AI

New insane model from Jackrong on @huggingface 🤯 Qwen3.5-9B-GLM5.1-Distill-v1 🧠 Distilled on GLM-5.1 reasoning
⚙️ Deeper thinking than base model
🧪 Benchmarks coming soon ✅ Fits on 8GB VRAM ✍️ New model after Qwopus/Gemopus After distilling Claude Opus 4.6, he’s now back on the strongest open-source model! An MLX  version is also available on his huggingface page 27B model incoming? https://t.co/893FvH51jb

Always happy to collaborate with the @huggingface team 🤗 HF is hitting the milestone of 10M datasets. Now you can easily adapt and evolve your data towards any objective. Shoutout to @ClementDelangue @julien_c @Thom_Wolf @mervenoyann! 🔥

Happy model launch day! Opus 4.7 is now available on all products and a significant step up from Opus 4.6. It's better at coding, computer use, finance, and general knowledge work. 🧵 I'll put the 5 things I find most interesting in thread! https://t.co/JEsw0a6Mrs

With computer use on macOS, Codex can now use any app by seeing, clicking, and typing with its own cursor. It runs in the background without taking over your computer, working on tasks like frontend iteration, app testing, or any workflow that doesn't expose an API. https://t.co/iO9iubLZX9

So excited to share that we're bringing Computer Use to Codex. Computer Use lets Codex see, click, and type into your Mac apps, with its own cursor. It's a magical feeling to have agents using your apps in the background, and still get to use your computer at the same time. https://t.co/wdgxiHAKyX

🚨🤯 Perplexity officially rolls out its “Personal Computer”, a Mac-based AI assistant that securely connects across local files, native apps, and browsers. The assistant can search, read, and write local files while interacting with iMessage, Mail, and Calendar, with 24/7 background execution and secure mobile-triggered tasks. Anyone who has tried OpenClaw knows the power of a local agentic system. This is what Apple Intelligence was supposed to be, and Perplexity just shipped it.

🚨🤯 Perplexity officially rolls out its “Personal Computer”, a Mac-based AI assistant that securely connects across local files, native apps, and browsers. The assistant can search, read, and write local files while interacting with iMessage, Mail, and Calendar, with 24/7 background execution and secure mobile-triggered tasks. Anyone who has tried OpenClaw knows the power of a local agentic system. This is what Apple Intelligence was supposed to be, and Perplexity just shipped it.