Roba Labs has launched its first publicly available version of an open, interoperable robotics ecosystem aimed at replacing proprietary, closed-stack robotics development.
Roba Labs, described as “The People’s Robotics Platform” and "the Hugging Face for robotics," promises a simplified workflow: build → train → share → monetize, aimed at disrupting an industry projected to grow to $218B by 2030.
The company emphasizes ownership, openness and community-governance: creators retain full IP/data ownership and monetization rights; the ecosystem is governed by a utility token ($ROBA) and DAO model, with about 30% of token supply allocated to ecosystem rewards.
Keycard raised $38M to build the “federated trust fabric for the agent-native era”. They’re focusing on identity and access management (IAM) specifically for AI agents rather than traditional human-users, for whom the current IAM systems were built.
Keycard’s offering: a platform with ephemeral, identity-bound tokens, federated identity for agents, full auditability of agent actions (who delegated what, when, on whose behalf), and integration with existing user/workload identity systems.
Today we're incredibly excited to introduce Keycard and announce our agent identity and platform is now available in early access. To support our mission, we've raised a $38M combined inception round led by Andreessen Horowitz, Acrew Capital, and boldstart ventures with participation from Mantis Venture Capital, Tapestry VC, Essence Venture Capital, Exceptional Capital, Modern Technical Fund, Vermillion Cliffs Ventures and many incredible angels.
Enable secure, transparent, context-aware AI agent interactions with Keycard. Identity infrastructure for MCP servers and tools, agent workflows, and security teams in the evolving AI landscape
The Web3 educational account Eli5defi has posed a compelling question: was AI capable to prevent, or at least soften the blow of, October 10th's crypto cascade liquidation events? For context, the day now referred to as "Crypto Black Friday" saw $19B crypto market liquidation provoked by Trump's China tariffs, high leverage, and potential manipulation.
The article presents a number of AI-powered solutions that could have alleviated the disastrous outcome, including:
Allora Labs' forecasting solutions, capable of adjusting model weights in real time amid abrupt market shifts and vulnerabilities;
Detecting anomalies and pinpointing manipulation via analytics tools like Nansen, Arkham, and Bubblemaps;
Sentiment and bias analysis via SocialFi platforms like Kaito and Cookie3;
Dynamic risk simulation and management by Theoriq;
Autonomous agents for arbitrage and reviews like the ones provided by Almanak;
Read the full article below to learn how "integrating these solutions could shift cryptocurrency's core trading framework from responsive to anticipatory".
Fhenix, the provider of Confidentiality-as-a-Service through FHE (Fully Homomorphic Encryption), has published a fascinating case study of Fluton Labs. In it, they've described how Fluton achieves confidentiality through Zama's and Fhenix's FHE coprocessors, and anonymity through smart accounts, thus presenting a new standard in privacy-preserving blockchain design.
For users, Fluton enables on-chain trades and bridge transfers that become completely unlinkable to original wallets, while transaction amounts and strategies remain fully confidential throughout execution. This is achieved without any compromise on DeFi composability or user experience, representing a significant advancement over existing privacy solutions that require users to choose between privacy and functionality.
In a world where universal confidentiality encompasses all blockchain networks, and the industry can offer compliance-grade privacy, meeting regulatory requirements and achieving institutional adoption will no longer be unattainable. Fluton and Fhenix demonstrate that we have reached the point where we no longer need to choose between privacy and functionality.
Privacy stops being an optional add-on and becomes the default state of life onchain.
Concluding today's edition with an exciting innovation coming from my native Bulgaria: a team at INSAIT - Institute for Computer Science, Artificial Intelligence and Technology has released an open-source robotics “brain” model called SPEAR‑1, which is the first robotic foundation model built in Europe!
Unlike many existing robot foundation models that rely primarily on 2D image data, SPEAR-1 incorporates 3D data during training, giving it a better spatial awareness of how objects move and exist in real physical space. What's more, it uses both robotic and non-robotic 3D data, thus solving the data bottleneck existing in training foundation robotic models. I've often written about DePIN projects working towards solving this same bottleneck through incentivizing crowdsourced environmental and mobility data.
🧩 What makes it different? SPEAR-1 is a new robotic foundation model which learns from both robotic and non-robotic 3D data, a breakthrough since 3D data is abundant and easy to obtain. SPEAR-1 outperforms or matches the leading foundation models such as OpenVLA, π0-FAST, and π0.5, trained on 20× more robot demonstration data.
On the RoboArena benchmark (which evaluates a robot’s ability to perform tasks like squeezing a ketchup bottle, closing a drawer, stapling paper) SPEAR-1 is reported to perform at a level approaching that of large commercial robotics models.
🧭 SPEAR-1 is open-weight and general-purpose, capable of controlling different robots across multiple tasks directly from human language instructions. This opens the path to flexible, scalable robot learning - the kind that will eventually power robots in homes, factories, and beyond.
🧠 We are releasing SPEAR-1 - a new robotic foundation model! SPEAR-1 achieves state-of-the-art performance with 20x less robotic data, making it not only a step towards solving the data bottleneck in robotics, but also the first robotic foundation model built in Europe! 🔬 Why does this matter?
WIRED argues that similarly to how open-source language models have accelerated AI innovation, open robotics models could accelerate progress in embodied AI. However, it's still a challenge to make models that generalize across different robots, environments, and tasks, and researchers caution that while 3D awareness seems promising, its full importance is not yet settled.
Open source language models are crucial to AI innovation. Can open robotics models do the same for physical machines?
Thank you for reading! The next edition is coming tomorrow.
I invite you to subscribe to The Web3 + AI Newsletter to stay in the loop on the hottest dAI developments.
I'm looking forward to connecting with fellow Crypto x AI enthusiasts, so don't hesitate to reach out to me on social media.
Disclaimer: None of this should or could be considered financial advice. You should not take my words for granted, rather, do your own research (DYOR) and share your thoughts to create a fruitful discussion.
Verifiable AI networks for dispute settlement like Talus Labs, Inc.
Share Dialog
Albena Kostova-Nikolova
Support dialog
157
No comments yet