Open Source
Meta open sources Muse AI agent SDK for custom hardware
The Verge AIhigh confidence
Why it matters
- →Developers can now build custom hardware interfaces for Meta's AI agent, decentralizing access beyond standard apps.
- →The open-source SDK lowers the barrier to entry for creating tangible, local AI experiences on cheap microcontrollers.
- →Physical integration expands Muse's utility into smart home automation and dedicated display devices.
Meta has open-sourced the software development kits (SDKs) required to run its Muse AI agent on custom hardware, including ESP32 microcontrollers and Raspberry Pi boards. The company encourages developers to build their own gadgets, such as E Ink reminder displays or HDMI-connected screens, while cautioning users to proceed at their own risk. Additionally, Meta is distributing 5,000 pre-built 'Muse Home Link' devices that allow users to control smart home appliances like lights and TVs using community-built skills. These physical units are available via a waitlist with shipping expected later this month.
Read originalThe story around this
TopicMeta Muse AI Agent
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1Meta AI Launches Muse Spark Agent
Matt Wolfe · July 31, 2026 · Same story
- 2Meta Introduces Muse Code for Large Code Bases
TechCrunch AI · August 5, 2026 · Same story
- 3Meta returns to open-source with Muse Glimmer
The Rundown AI · August 11, 2026 · Same story
- 4Meta Launches AI Agent Muse to Compete in AI Race
The Verge AI · September 8, 2026 · Same story
- 5Meta Launches Muse, a Personal AI Agent
WIRED AI · September 8, 2026 · Same story
- 6Meta Launches 'Muse' AI Agent for Consumer Use
Matt Wolfe · September 15, 2026 · Same story
- 7Meta admits Muse inspired by OpenClaw
TechCrunch AI · September 22, 2026 · Same story
- 8Meta unveils standalone Muse Charm AI hardware
The Verge AI · September 24, 2026 · Same story
What happened next
- 9Meta's Muse Agent Builds Detailed Profiles of Your Social Circle
WIRED AI · October 3, 2026 · Related
Follow this story
Open the full story →Meta Muse now exposes full cloud VM filesystem
2 developments
- Sep 25 · The Verge AI
Meta Muse now exposes full cloud VM filesystem - Oct 2 · The Verge AI
Meta open sources Muse AI agent SDK for custom hardware (This article)↳ Meta open sources the Muse AI agent SDK to enable running it on custom hardware like ESP32 and Raspberry Pi boards.
More from The Verge AI
© The Verge AIGeneral AIother
Capcom plans AI-integrated RE Engine
Capcom is quietly pivoting from its strict no-AI-assets stance to integrating generative tools directly into the RE Engine workflow. This isn't about replacing artists; it's about solving the crushing time costs of AAA production by letting developers co-create with the engine. The shift signals a pragmatic industry realization: if AI can accelerate iteration, studios will adopt it regardless of previous ethical red lines. We are moving from 'AI in games' to 'AI for making games.'
The Verge AI
© The Verge AIGeneral AIother
OpenAI safety lead quits, calls culture broken
David Robinson’s departure from OpenAI marks a significant shift in the internal narrative around AI safety. As the former author of safety reports for major model releases, his public critique carries weight beyond typical employee grievances. He argues that the industry's 'move-fast' culture is fundamentally incompatible with managing existential risks, advocating for nuclear-level safeguards instead. This aligns with a growing trend of insiders leaving firms like Anthropic and Google DeepMind to voice similar concerns. The real story here isn't just one resignation, but the erosion of trust in self-regulation from within the labs themselves.
The Verge AI
© The Verge AIGeneral AIother
AI Hallucinations Disrupt Customer Service
Generative AI is actively degrading frontline service interactions as customers blindly trust hallucinated facts over human expertise. From diners ignoring allergen warnings based on ChatGPT to parents dismissing pediatric advice for AI sleep schedules, the phenomenon reveals a dangerous erosion of professional authority. This isn't just about bad data; it's about users delegating critical judgment to models that prioritize confidence over accuracy. The trend is accelerating with agentic AI, where automated requests further disconnect human context from service delivery.
The Verge AI
More in Open Source
© Hugging Face BlogOpen Sourcewriting
Ai2 open-sources AstaBrief 8B for scientific reports
Allen Institute for AI has released AstaBrief 8B, an open-weight model designed specifically for generating cited scientific literature reviews. Built on Qwen3-8B and trained with supervised fine-tuning and direct preference optimization, it prioritizes speed and grounding over complex multi-step reasoning. The model generates full reports in a single pass, cutting generation time to roughly 51 seconds compared to the 178 seconds required by proprietary alternatives like Claude. This release offers researchers a faster, locally deployable option for synthesizing evidence without relying on external APIs.
Hugging Face Blog
Open Sourceother
llama.cpp adds IBM ZDNN backend support
IBM is finally getting first-class CI support in llama.cpp with the addition of the ZDNN backend for s390x architecture. This isn't just a minor tweak; it enables efficient inference on mainframe hardware, bridging a gap for enterprise environments that rely on IBM Z systems. While currently limited to build pipelines without automated testing, this signals a serious commitment to supporting non-x86/ARM infrastructure in the local LLM ecosystem. It’s a quiet but necessary expansion for anyone running models on legacy or specialized enterprise silicon.
llama.cpp Releases
© Lev SelectorOpen Sourcemodels
Bonsai 2 and Needle: Tiny Local AI Models
New tiny local models Bonsai 2 and Needle (8-29 MB) demonstrate that small, offline-capable AI can make fast, useful decisions.
Lev Selector