正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Muse Glimmer can complete multi-step agentic tasks end-to-end from a single natural language prompt. In this demo, it autonomously discovers a local Home Assistant instance via network tool calls, queries device APIs, writes a responsive HTML/CSS/JS dashboard from scratch, and deploys a local server for verification.

AI at Meta

861,326 subscribers

52,924 次观看 • 2 个月前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

11 条评论

AI at Meta 的头像

AI at Meta2 个月前

For a local agent to be practical, generation latency must be low enough to maintain workflow continuity. To run Muse Glimmer on consumer hardware without degrading quality, we used quantization to shrink the language model to under 20GB and a lightweight DFlash drafter model to accelerate token generation. As a result, Muse Glimmer is fast enough for fluid conversation and real-time agent interaction, all running entirely on your device.

AI at Meta 的头像

AI at Meta2 个月前

Introducing Muse Glimmer, an open-weight 30B-parameter model optimized for local, always-on agent workflows. Muse Glimmer delivers strong performance on key agentic use cases and benchmarks compared with leading models in its size category, and is designed to run entirely on consumer hardware like a Mac or PCs with performant GPUs. In keeping with our long tradition of sharing fundamental AI research, we’re releasing model weights under a permissive Apache 2.0 license. 🧵👇

AI at Meta 的头像

AI at Meta2 个月前

🔗 Download Muse Glimmer on @huggingface: 🔗 Read the technical blog: 🔗 Find resources:

Krish Dasgupta 的头像

Krish Dasgupta2 个月前

You guys are so backkkk !!!!!

AI at Meta 的头像

AI at Meta2 个月前

Now its time to build 🚀

Krish Dasgupta 的头像

Krish Dasgupta2 个月前

Yessssss !!!! Building Something on SRE. Time to put Glimmer in it too !!! Let’s goooooooo. Will be posting the demo here soon !!!

Sanjay Bhadra 🇮🇳 的头像

Sanjay Bhadra 🇮🇳2 个月前

This Muse Glimmer demo is compelling because it completed the entire local-agent loop: discover Home Assistant, understand its APIs, build a dashboard, launch it, and verify it. If it holds up beyond the demo, I can see many private home networking, camera, and automation projects being attempted by non-tech or semi-tech people!

d5niel 的头像

d5niel2 个月前

Make a deal with @opencode to give every user $120 usage with meta models atleast you will have users 😭

Mohith kumar 的头像

Mohith kumar15 天前

Glimmer finding a local Home Assistant instance through network tools is a cool demo. I'm really glad I came across this, so I put you up on ShipWithMuse, my showcase of the best Muse builds. Take a look 👉

Normarayr 的头像

Normarayr2 个月前

Can you guys tell your demo person to use iTerm2 or Ghostty? The rendering artifacts really paints a bad picture for opencode like apple using 2011 android mockups

LADI ⟡⭐ 的头像

LADI ⟡⭐2 个月前

Big move for open source Making strong models like Muse Glimmer available to run locally is a huge step

相关视频

Advice for AI engineers 💡 A local AI assistant is just a while loop, an LLM and a set of tools. Here's a step-by-step lesson on how to build one from scratch with Liquid AI LFM2-24B-A2B running on llama.cpp. No cloud APIs. Enjoy ↓

Advice for AI engineers 💡 A local AI assistant is just a while loop, an LLM and a set of tools. Here's a step-by-step lesson on how to build one from scratch with Liquid AI LFM2-24B-A2B running on llama.cpp. No cloud APIs. Enjoy ↓

Pau Labarta Bajo

15,463 次观看 • 5 个月前

QVAC first demo (android mobile) of its local on-device fine-tuning generalized framework, which aims to bring LORA fine-tuning to any device, scaling from the smallest embedded device to any smartphone, laptop, server, operating system. In combination with WIP multi-node features, we expect QVAC to become the standard for local on-device single/multi node inference and fine-tuning. Imagine a horde of AI agents being able to learn from your emails, notes and messages, maintaining 100℅ privacy, leveraging only your own devices. A P2P private intelligence network at your fingertips. That's how QVAC personal assistant will look like. Future of Intelligence won't be centralized. Infinite Ingelligence vibes.

QVAC first demo (android mobile) of its local on-device fine-tuning generalized framework, which aims to bring LORA fine-tuning to any device, scaling from the smallest embedded device to any smartphone, laptop, server, operating system. In combination with WIP multi-node features, we expect QVAC to become the standard for local on-device single/multi node inference and fine-tuning. Imagine a horde of AI agents being able to learn from your emails, notes and messages, maintaining 100℅ privacy, leveraging only your own devices. A P2P private intelligence network at your fingertips. That's how QVAC personal assistant will look like. Future of Intelligence won't be centralized. Infinite Ingelligence vibes.

Paolo Ardoino 🤖

142,902 次观看 • 11 个月前

i gave a 9B model an agentic OS and told it to build Octopus Invaders from scratch. Qwen 3.5 9B running through Hermes Agent with full tool access on a single 3060 12GB. browser, file system, code execution, terminal. it reads the prompt, plans the architecture, writes every file, and serves the game. one prompt, zero steering, 11 minutes. i've been going back and forth with this model for days now. text, code, reasoning. it handles small chats, batching, bash scripting. at 9B it's fast enough to feel interactive and genuinely useful for local workflows where you'd normally reach for an API. but this is the real test. can it architect and build a full game autonomously without a single correction? it's coding right now. results coming next.

i gave a 9B model an agentic OS and told it to build Octopus Invaders from scratch. Qwen 3.5 9B running through Hermes Agent with full tool access on a single 3060 12GB. browser, file system, code execution, terminal. it reads the prompt, plans the architecture, writes every file, and serves the game. one prompt, zero steering, 11 minutes. i've been going back and forth with this model for days now. text, code, reasoning. it handles small chats, batching, bash scripting. at 9B it's fast enough to feel interactive and genuinely useful for local workflows where you'd normally reach for an API. but this is the real test. can it architect and build a full game autonomously without a single correction? it's coding right now. results coming next.

Sudo su

65,570 次观看 • 7 个月前

2. Build interactive tools and apps without code Describe your idea and Grok scaffolds the UI, logic, and deployment steps no coding required. Prompt: "You are a senior software architect who excels at building no-code and low-code systems with HTML, JavaScript, and Web APIs. Task Design a fully working interactive tool based on this idea: [DESCRIBE YOUR TOOL IDEA] Deliverables 1. A plain-English explanation of how the tool works. 2. A step-by-step plan to implement it with Grok or a no-code platform. 3. Exact HTML/CSS/JS code if relevant. 4. UX and design improvement tips. 5. Instructions to generate a shareable, embed-ready version. Imagine you are shipping an MVP for a startup demo. "

2. Build interactive tools and apps without code Describe your idea and Grok scaffolds the UI, logic, and deployment steps no coding required. Prompt: "You are a senior software architect who excels at building no-code and low-code systems with HTML, JavaScript, and Web APIs. Task Design a fully working interactive tool based on this idea: [DESCRIBE YOUR TOOL IDEA] Deliverables 1. A plain-English explanation of how the tool works. 2. A step-by-step plan to implement it with Grok or a no-code platform. 3. Exact HTML/CSS/JS code if relevant. 4. UX and design improvement tips. 5. Instructions to generate a shareable, embed-ready version. Imagine you are shipping an MVP for a startup demo. "

God of Prompt

316,672 次观看 • 1 年前

Wow Microsoft just announced Copilot Actions for local files 🤯 You can basically ask Copilot to perform any task on your machine autonomously. Yes. - Open Copilot in Windows - Assign a goal in natural language - Copilot launches a contained environment - It works autonomously for you Copilot can use both your desktop and any web app. And you can take back control at any time. This allows you to do whatever you want while the AI performs your tasks... or even do nothing at all. The entire OS becomes agentic. Impressive.

Wow Microsoft just announced Copilot Actions for local files 🤯 You can basically ask Copilot to perform any task on your machine autonomously. Yes. - Open Copilot in Windows - Assign a goal in natural language - Copilot launches a contained environment - It works autonomously for you Copilot can use both your desktop and any web app. And you can take back control at any time. This allows you to do whatever you want while the AI performs your tasks... or even do nothing at all. The entire OS becomes agentic. Impressive.

Paul Couvert

87,452 次观看 • 1 年前

[3/5] Introducing Agent Mode ✨ Agent Mode is a simple way to bootstrap an automation from scratch with Notte. Prompt a browser agent in natural language, let it execute the task step by step, then automatically convert its execution into deterministic & editable code. Open it in Automation Studio and continue iterating from there!

[3/5] Introducing Agent Mode ✨ Agent Mode is a simple way to bootstrap an automation from scratch with Notte. Prompt a browser agent in natural language, let it execute the task step by step, then automatically convert its execution into deterministic & editable code. Open it in Automation Studio and continue iterating from there!

Notte

12,101 次观看 • 9 个月前

The "prompt-to-production" flow is already a reality. I hope that in the foreseeable future, we will be able to "prompt-to-real estate" as well, with robots building a complete building from the ground up. For now, it is still too futuristic, but Drafted has already automated the "prompt-to-home-design" step. > Drafted AI has a tool that lets you generate your dream house for FREE and export its detailed rendering. > Users can remix existing designs from the explore section or start from scratch. All the details can be customized. I need a new house now 👀

The "prompt-to-production" flow is already a reality. I hope that in the foreseeable future, we will be able to "prompt-to-real estate" as well, with robots building a complete building from the ground up. For now, it is still too futuristic, but Drafted has already automated the "prompt-to-home-design" step. > Drafted AI has a tool that lets you generate your dream house for FREE and export its detailed rendering. > Users can remix existing designs from the explore section or start from scratch. All the details can be customized. I need a new house now 👀

🚨 AI News | TestingCatalog

24,348 次观看 • 5 个月前

With Google Flow Tools, you can now use natural language to create bespoke tools and workflows. In this video, you'll see how you can: ➡️ Explore and try the various Tools within Google Flow ➡️ Remix a tool to make it even more relevant for your unique workflow ➡️Create your own tool from scratch with a natural language prompt ➡️ Share your tools with a friend for them to try out Whether you’re looking for an image editor, video resizer or custom shaders, you can now create them in Google Flow, no coding experience required.

With Google Flow Tools, you can now use natural language to create bespoke tools and workflows. In this video, you'll see how you can: ➡️ Explore and try the various Tools within Google Flow ➡️ Remix a tool to make it even more relevant for your unique workflow ➡️Create your own tool from scratch with a natural language prompt ➡️ Share your tools with a friend for them to try out Whether you’re looking for an image editor, video resizer or custom shaders, you can now create them in Google Flow, no coding experience required.

Google Flow

20,037 次观看 • 3 个月前

Someone built a tool where you paste an OpenAPI spec and get a fully stateful sandbox that actually behaves like production No local setup No real API costs It's called FetchSandbox The workflow does this : → explores Stripe APIs → runs workflows → triggers webhooks → verifies integrations end to end Works with GitHub, Stripe, Twilio, Shopify, ElevenLabs, Clerk, Paddle, and 45+ more APIs Now it supports an MCP server, letting Cursor & Claude explore APIs & run workflows directly from your IDE Perfect for AI agents, rapid prototyping & testing complex API workflows

Someone built a tool where you paste an OpenAPI spec and get a fully stateful sandbox that actually behaves like production No local setup No real API costs It's called FetchSandbox The workflow does this : → explores Stripe APIs → runs workflows → triggers webhooks → verifies integrations end to end Works with GitHub, Stripe, Twilio, Shopify, ElevenLabs, Clerk, Paddle, and 45+ more APIs Now it supports an MCP server, letting Cursor & Claude explore APIs & run workflows directly from your IDE Perfect for AI agents, rapid prototyping & testing complex API workflows

Aditi

35,840 次观看 • 4 个月前

$And here’s an app extension replacing Siri AI’s server model with GPT-5.6 Terra. It uses the Inference Providing protocol in Model Manager Services. GPT-5.6 receives Apple’s native Siri planner prompt and tool definitions. It can make tool calls that perform system actions, and receive tool call results with personal data. Finally, GPT-5.6 can return a response through Siri AI’s UI and voice, including interactive elements. Siri AI has a local “Skimmer” planner with limited capabilities that hands off most tasks to Apple’s server models, but that’s for another day. Watch until the end to see requests being logged on OpenAI’s platform. Uses a private ExtensionKit extension point: com\.apple.modelmanager.inferenceprovider.$

And here’s an app extension replacing Siri AI’s server model with GPT-5.6 Terra. It uses the Inference Providing protocol in Model Manager Services. GPT-5.6 receives Apple’s native Siri planner prompt and tool definitions. It can make tool calls that perform system actions, and receive tool call results with personal data. Finally, GPT-5.6 can return a response through Siri AI’s UI and voice, including interactive elements. Siri AI has a local “Skimmer” planner with limited capabilities that hands off most tasks to Apple’s server models, but that’s for another day. Watch until the end to see requests being logged on OpenAI’s platform. Uses a private ExtensionKit extension point: com\.apple.modelmanager.inferenceprovider.

pdfu

46,714 次观看 • 29 天前

now doing gemma 4 moe. it'a crazy how usable this is. just a little web search tool (wiggles its head on tool calls), and it can serve as a mostly capable, local voice assistant. tomorrow i shall replace google auto crap with this in my car. just need to give it access to spotify.

now doing gemma 4 moe. it'a crazy how usable this is. just a little web search tool (wiggles its head on tool calls), and it can serve as a mostly capable, local voice assistant. tomorrow i shall replace google auto crap with this in my car. just need to give it access to spotify.

Mario Zechner

27,323 次观看 • 4 个月前

Markdown was doomed from the start. It's just a format with low information density. HTML is better for humans, and agents can now consume and produce it without issues. But nobody wants to type HTML, so here is an alternative: This is an open-source tool for generating dashboards from data without writing a single HTML tag. You define your dashboard in YAML or TSK, and the tool will serve the HTML file for you. It comes with skills for Claude Code and Codex, so they know how to build these dashboards. And you can connect this to Postgres, MySQL, Snowflake, BigQuery, Redshift, Databricks, and many other databases. Repo link below.

Markdown was doomed from the start. It's just a format with low information density. HTML is better for humans, and agents can now consume and produce it without issues. But nobody wants to type HTML, so here is an alternative: This is an open-source tool for generating dashboards from data without writing a single HTML tag. You define your dashboard in YAML or TSK, and the tool will serve the HTML file for you. It comes with skills for Claude Code and Codex, so they know how to build these dashboards. And you can connect this to Postgres, MySQL, Snowflake, BigQuery, Redshift, Databricks, and many other databases. Repo link below.

Santiago

86,565 次观看 • 5 个月前

Insane progress for small language models! MiniCPM5-2B is a dense 2B-parameter model by OpenBMB from China that's built for reasoning, coding, and tool use on resource-constrained hardware. The model specifically excels at coding and tool calling, two capabilities central to the shift from on-device LLMs to on-device agents. Instead of only answering prompts, it can use tools, generate code, carry information between steps, and complete multi-step tasks. I ran it 100% locally and connected it to a small investigation agent with one request: > Revenue dropped last week. Investigate what happened, quantify the impact, identify the likely cause, and produce an incident report with supporting evidence. The evidence was spread across orders, traffic, payments, refunds, and deployment logs. The model inspected the files, wrote its own queries, analyzed the intermediate results, and decided what to investigate next. Each tool result informed the next action, so the final report depended on the model maintaining a coherent investigation across the complete trajectory. The recording shows the actual task from beginning to end. It starts with the revenue question, follows the tool calls and supporting evidence, and ends with a quantified diagnosis and incident report. The data, tool execution, and model inference all remained on my machine. These capabilities were optimized through Agentic Pre-training, SFT, and large-scale RL. They do not come entirely from an application-level agent framework. MiniCPM5-2B supports SGLang, vLLM, llama(.)cpp, Ollama, iOS, Android, and HarmonyOS. OpenBMB has also released the model weights and parts of the training recipes and data resources behind it. Download MiniCPM5-2B: A 2B model can now maintain enough state to coordinate tools and complete a useful investigation on local hardware.

Insane progress for small language models! MiniCPM5-2B is a dense 2B-parameter model by OpenBMB from China that's built for reasoning, coding, and tool use on resource-constrained hardware. The model specifically excels at coding and tool calling, two capabilities central to the shift from on-device LLMs to on-device agents. Instead of only answering prompts, it can use tools, generate code, carry information between steps, and complete multi-step tasks. I ran it 100% locally and connected it to a small investigation agent with one request: > Revenue dropped last week. Investigate what happened, quantify the impact, identify the likely cause, and produce an incident report with supporting evidence. The evidence was spread across orders, traffic, payments, refunds, and deployment logs. The model inspected the files, wrote its own queries, analyzed the intermediate results, and decided what to investigate next. Each tool result informed the next action, so the final report depended on the model maintaining a coherent investigation across the complete trajectory. The recording shows the actual task from beginning to end. It starts with the revenue question, follows the tool calls and supporting evidence, and ends with a quantified diagnosis and incident report. The data, tool execution, and model inference all remained on my machine. These capabilities were optimized through Agentic Pre-training, SFT, and large-scale RL. They do not come entirely from an application-level agent framework. MiniCPM5-2B supports SGLang, vLLM, llama(.)cpp, Ollama, iOS, Android, and HarmonyOS. OpenBMB has also released the model weights and parts of the training recipes and data resources behind it. Download MiniCPM5-2B: A 2B model can now maintain enough state to coordinate tools and complete a useful investigation on local hardware.

Avi Chawla

14,272 次观看 • 1 个月前

Open Design 0.8.0 now works inside Cursor Composer 2.5. Composer 2.5 is built for this: long tasks, tool use, and iteration in one run. So we asked Composer to control Open Design through chat and build an animated HTML website. What happened 2.5 minutes later is insane: - landing page - smooth motion via GSAP gsap-skills - responsive HTML - ready to iterate Here is a quick demo:

Open Design 0.8.0 now works inside Cursor Composer 2.5. Composer 2.5 is built for this: long tasks, tool use, and iteration in one run. So we asked Composer to control Open Design through chat and build an animated HTML website. What happened 2.5 minutes later is insane: - landing page - smooth motion via GSAP gsap-skills - responsive HTML - ready to iterate Here is a quick demo:

Open Design

16,926 次观看 • 4 个月前

LFM2.5-2.6B is live! At only 2.6B, it shouldn't be this good at agentic tasks. Here you can see how a 100% local agent is cleaning up my Desktop. I wrote a step-by-step guide on how to set it up yourself: • Model: LFM2.5-2.6B • Server of choice: llama.cpp, LMStudio, and more • Harness of choice: OpenClaw, Hermes Agent, Pi

LFM2.5-2.6B is live! At only 2.6B, it shouldn't be this good at agentic tasks. Here you can see how a 100% local agent is cleaning up my Desktop. I wrote a step-by-step guide on how to set it up yourself: • Model: LFM2.5-2.6B • Server of choice: llama.cpp, LMStudio, and more • Harness of choice: OpenClaw, Hermes Agent, Pi

Leonie

64,025 次观看 • 2 个月前

🚀 The Google DeepMind team just added Gemini 3.1 to the Live API, so we built a small demo showing how Gemini voice agents can plug directly into the document processing ecosystem powered by LlamaIndex. 🔥 In this example, we integrate LiteParse to enable fast, fully-local document parsing. With our TUI-based voice assistant, you can literally talk to your terminal: - Speak commands - Trigger live document parsing via tool calls - Hear the agent read back results in real time 🔊 The assistant can extract content from single files or entire folders, leveraging the lightning-fast local parsing that LiteParse provides ⚡ Take a look at the demo👇 👩‍💻 GitHub repo 📚 LiteParse docs

🚀 The Google DeepMind team just added Gemini 3.1 to the Live API, so we built a small demo showing how Gemini voice agents can plug directly into the document processing ecosystem powered by LlamaIndex. 🔥 In this example, we integrate LiteParse to enable fast, fully-local document parsing. With our TUI-based voice assistant, you can literally talk to your terminal: - Speak commands - Trigger live document parsing via tool calls - Hear the agent read back results in real time 🔊 The assistant can extract content from single files or entire folders, leveraging the lightning-fast local parsing that LiteParse provides ⚡ Take a look at the demo👇 👩‍💻 GitHub repo 📚 LiteParse docs

LlamaIndex 🦙

14,766 次观看 • 6 个月前