Agents
DeepSeek Harness launches Super Agent prototype
Sam Witteveenhigh confidence
Why it matters
- →Demonstrates a shift toward plugin-based, modular agent architectures.
- →Provides an open-source prototype for composable AI systems.
- →Highlights the evolving role of harnesses as orchestration layers.
The DeepSeek Harness project has released a prototype for 'Super Agent,' an add-on that reimagines the framework as a plugin-based architecture. This update allows developers to treat models and tools as interchangeable components, enabling more flexible agent configurations. Currently available only on GitHub and tested against synthetic benchmarks, the prototype demonstrates a shift toward modular agent design. The release underscores the growing importance of harnesses in managing complex, multi-step AI workflows.
Read originalThe story around this
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
- 1Agent Plugins 1.0 Launches for VS Code and Copilot
GitHub Changelog · August 12, 2026 · Background
- 2DeepSeek Releases Next-Gen Open Source Coding Agent
Cole Medin · August 20, 2026 · Same story
- 3OpenAI Launches Agents API for Cloud Agents
OpenAI · September 10, 2026 · Background
- 4DeepSeek Releases DeepSeek-V4.1-Flash Model
Matt Wolfe · September 11, 2026 · Related
- 5TrueFoundry launches open-source TrueForge agent harness
Sam Witteveen · September 14, 2026 · Related
- 6AWS Strands Agent Harness Released
Lev Selector · September 25, 2026 · Related
- 7LangChain Managed Agents Self-Improve via Harness Engineering
Cole Medin · October 20, 2026 · Related
- 8Google Launches Harness2 Agent Framework
Lev Selector · October 23, 2026 · Related
What happened next
- 9Dependabot now updates Agent Package Manager
GitHub Changelog · November 3, 2026 · Background
- 10Together AI automates inference optimization with agents
Together AI Blog · November 4, 2026 · Background
- 11BAAI Releases AREX-2 Agent Model
Sam Witteveen · November 6, 2026 · Background
More from Sam Witteveen
© Sam WitteveenModels & Labsagents
BAAI Releases AREX-2 Agent Model
BAAI’s AREX-2 shifts the burden of agent reliability from external orchestration to the model itself. Fine-tuned on Qwen3.8-27B, this open-weights model is explicitly trained to execute the try-check-retry loop internally rather than relying on a separate harness to manage failures. This architectural choice targets long-horizon tasks where external loops often break down under complexity. By embedding reflection and self-correction directly into the weights, it offers a more robust path for autonomous coding agents that need to recover from errors without human intervention.
Sam Witteveen
© Sam WitteveenOpen Sourceother
Ricursive AI Releases Ripple 65B for Chip Design
Ricursive AI has released Ripple 65B, an open-source model specifically trained to accelerate semiconductor design workflows. By focusing on a niche vertical rather than general-purpose reasoning, the company achieved competitive performance against major US open models while training for just $32 million. The architecture is described as thin and deep, optimized for the specific token distributions found in hardware description languages. This release signals that specialized, domain-locked models can outperform broad foundation models when the data distribution is sufficiently distinct and high-value.
Sam Witteveen
More in Agents
© MIT Technology Review AIAgentsagents
Tau Robotics humanoid cleaning pilot reveals remote control reality
Tau Robotics’ San Francisco pilot exposes the gap between humanoid hype and physical reality. The robots are currently teleoperated via kinematic guidance, meaning human pilots perform the motions while sensors record data for future autonomy. While they can mop and vacuum independently, complex tasks like folding laundry or navigating stairs still require heavy human intervention and suffer from latency issues. This service is less a product launch and more a large-scale data collection exercise to train models on real-world cleaning dynamics.
MIT Technology Review AI
© TechCrunch AIAgentsagents
Airbnb commits to AI travel agent by H1 2027
Airbnb is doubling down on the 'generative UI' concept with a dedicated travel agent launching in early 2027. CEO Brian Chesky explicitly rejects the chatbot model, arguing that planning trips is an aspirational experience people enjoy. Instead of text-heavy responses, the agent will dynamically compose interfaces using maps, photos, and itineraries based on user intent. This approach leverages Airbnb's proprietary data to handle complex post-booking tasks like refunds and safety support, which general-purpose agents cannot easily access. The move signals a shift from conversational AI toward context-aware, visual planning tools.
TechCrunch AI
© TechCrunch AIAgentsagents
Meta opens Muse video chat to all US users
Meta has removed the invite-only barrier for Muse’s video avatar feature, making it available to every US user. This rollout leverages new Realtime Avatar and Voice models to drive subtle facial expressions and gestures that sync with conversation tone, aiming to deepen emotional attachment through anthropomorphism. While competitors focus on voice-only interfaces, Meta is betting that a customizable, animated presence—like Alexandr Wang’s fluffy dog avatar—creates a stickier consumer habit. The move signals a shift from experimental demos to mainstream consumer engagement, setting the stage for upcoming hardware integrations like AI glasses and the Muse Charm wearable.
TechCrunch AI