Home Blog Page 2

Engineered extracellular vesicles facilitate supply of superior medicines – NanoApps Medical – Official web site


New technology facilitates delivery of advanced medicines
Graphic summary of the event of VEDIC and VFIC programs for top effectivity intracellular protein supply in vitro and in vivo. Credit score: Nature Communications (2025). DOI: 10.1038/s41467-025-59377-y. https://www.nature.com/articles/s41467-025-59377-y

Researchers at Karolinska Institutet have developed a way that allows environment friendly supply of therapeutic proteins and RNA to cells. The tactic, revealed in Nature Communications, exhibits promising leads to animal research to ship gene editors and protein therapeutics.

The tactic is predicated on so-called extracellular vesicles (EVs),  which might be naturally secreted by cells and may transport biologically energetic molecules between cells. Researchers have now succeeded in enhancing these bubbles by introducing two key parts: a small a part of a bacterial  referred to as intein and a so-called fusogenic protein from a virus.

The fusogenic protein helps the bubbles to fuse with the endosomal membrane and launch their contents into the cell, whereas the intein can minimize itself and thus assist launch therapeutic proteins contained in the cell.

“This progressive engineering technique represents a significant step ahead for extracellular vesicle expertise, successfully overcoming key obstacles resembling poor endosomal escape and restricted intracellular launch,” says Professor Samir EL Andaloussi, final writer of the examine and researcher on the Division of Laboratory Drugs.

“Our in vivo findings spotlight the potential of engineered EVs as a flexible platform for delivering therapeutics to deal with a broad vary of situations, together with systemic irritation, genetic ailments, and neurological problems.”

The examine’s first writer, Dr. Xiuming Liang on the Division of Laboratory Drugs, provides, “By enhancing the effectivity and reliability of therapeutic supply into , this expertise may considerably broaden the applying of superior medicines.”

Mind adjustments in mice

In experiments on cells and , the researchers had been capable of effectively ship Cre recombinase, a protein that may minimize and paste DNA, and Cas9/sgRNA complexes, that are used to edit genes. When extracellular vesicles loaded with Cre recombinase had been injected into the brains of mice, a big change in cells within the hippocampus and cortex mind constructions was noticed.

“This offers hope to make use of the CRISPR/Cas9 gene scissors or comparable instruments to deal with extreme  of the central nervous system, resembling Huntington’s illness and ,” says Liang.

The researchers additionally demonstrated that the approach may very well be used to deal with systemic irritation in mice.

Extra info: Xiuming Liang et al, Engineering of extracellular vesicles for environment friendly intracellular supply of multimodal therapeutics together with genome editors, Nature Communications (2025). DOI: 10.1038/s41467-025-59377-ywww.nature.com/articles/s41467-025-59377-y

LEVA robotic makes use of legs and wheels for autonomous cargo dealing with


Floor-going supply robots are already streamlining operations in numerous industries, however they usually nonetheless need to be loaded and unloaded by people. That is the place the multi-terrain LEVA bot is available in, as it could possibly autonomously carry and drop off cargo packing containers weighing as much as 85 kg.

At present in useful prototype kind, LEVA is being developed by a staff of engineers and designers from Switzerland’s ETH Zurich college, Zurich College of Utilized Sciences, and Zurich College of the Arts.

ETH spinoff firm RIVR (previously Swiss-Mile) already manufactures the RIVR ONE (previously ANYmal) quadruped robotic, which definitely might be used for deliveries in city settings. Like RIVR ONE, LEVA has 4 jointed legs, each with a motorized wheel on the finish as a substitute of a foot.

The existing RIVR ONE delivery robot
The present RIVR ONE supply robotic

RIVR

This design permits the robotic to rapidly and energy-efficiently zip throughout clean surfaces corresponding to sidewalks, just by rolling on its wheels like a automobile. It makes use of a mixture of GPS, LiDAR sensors and cameras to autonomously navigate metropolis streets and keep away from obstacles.

When the bot must climb stairs or step over curbs, it locks up its wheels after which walks/climbs like a canine or different four-legged animal. It may well additionally use this performance to traverse tough terrain that is not conducive to rolling.

The large factor that units LEVA aside from RIVR ONE is its skill to autonomously spot, make its means over to, then carry and carry Euro-type cargo packing containers weighing as a lot as 85 kg (187 lb).

The robot navigates via a combination of GPS, LiDAR sensors and cameras
The robotic navigates through a mixture of GPS, LiDAR sensors and cameras

LEVA

This course of is partially made potential by 5 onboard cameras – one on every vertical aspect of the robotic’s rectangular(ish) physique, and one on the underside. Moreover, as a result of every of the wheels is independently steerable, the omnidirectional bot can roll straight to the left or proper as wanted.

As soon as LEVA has positioned itself straight over the field, it bends its legs to decrease its physique downwards. Inside powered hooks then latch onto the highest of the field, securing it beneath the bot’s physique. The robotic then simply rises again up and heads for its vacation spot.

It is laborious to say how quickly it is likely to be earlier than you see a LEVA on a road close to you, because the scientists are nonetheless perfecting the bot’s real-world performance. Within the meantime, you’ll be able to see it in box-toting, stair-climbing, ground-rolling motion, within the video beneath.

LEVA: A high-mobility logistic car with legged suspension

Supply: LEVA



NVIDIA Cosmos: Empowering Bodily AI with Simulations


The event of bodily AI programs, corresponding to robots on manufacturing unit flooring and autonomous automobiles on the streets, depends closely on massive, high-quality datasets for coaching. Nonetheless, gathering real-world information is dear, time-consuming, and sometimes restricted to some main tech firms. NVIDIA’s Cosmos platform addresses this problem by utilizing superior physics simulations to generate practical artificial information on a scale. This permits engineers to coach AI fashions with out the fee and delay related to gathering real-world information. This text discusses how Cosmos improves entry to important coaching information and accelerates the event of secure, dependable AI for real-world functions.

Understanding Bodily AI

Bodily AI refers to synthetic intelligence programs that may understand, perceive, and act throughout the bodily world. In contrast to conventional AI, which could analyze textual content or photos, bodily AI should take care of real-world complexities like spatial relationships, bodily forces, and dynamic environments. For instance, a self-driving automotive wants to acknowledge pedestrians, predict their actions, and alter its path in actual time, whereas contemplating elements like climate and highway situations. Equally, a robotic in a warehouse should navigate obstacles and manipulate objects with precision.

Growing bodily AI is difficult as a result of it requires huge quantities of information to coach fashions on numerous real-world eventualities. Accumulating this information, whether or not it is hours of driving footage or robotic job demonstrations, will be time-consuming and costly. Furthermore, testing AI in the true world will be dangerous, as errors might result in accidents. NVIDIA Cosmos addresses these challenges by utilizing physics-based simulations to generate practical artificial information. This strategy simplifies and accelerates the event of bodily AI programs.

What Are World Basis Fashions?

On the core of NVIDIA Cosmos is a group of AI fashions referred to as world basis fashions (WFMs).  These AI fashions are particularly designed to simulate digital environments that carefully mimic the bodily world. By producing physics-aware movies or eventualities, WFMs simulate how objects work together based mostly on spatial relationships and bodily legal guidelines. As an example, a WFM might simulate a automotive driving by a rainstorm, exhibiting how water impacts traction or how headlights mirror off moist surfaces.

WFMs are essential for bodily AI as a result of they supply a secure, controllable house to coach and take a look at AI programs. As a substitute of gathering real-world information, builders can use WFMs to generate artificial information—practical simulations of environments and interactions. This strategy not solely reduces prices but additionally accelerates the event course of and permits for testing advanced, uncommon eventualities (corresponding to uncommon site visitors conditions) with out the dangers related to real-world testing. WFMs are general-purpose fashions that may be fine-tuned for particular functions, just like how massive language fashions are tailored for duties like translation or chatbots.

Unveiling NVIDIA Cosmos

NVIDIA Cosmos is a platform designed to allow builders to construct and customise WFMs for bodily AI functions, notably in autonomous automobiles (AVs) and robotics. Cosmos integrates superior generative fashions, information processing instruments, and security options to develop AI programs that work together with the bodily world. The platform is open supply, with fashions accessible beneath permissive licenses.

Key elements of the platform embody:

  • Generative World Basis Fashions (WFMs): Pre-trained fashions that simulate bodily environments and interactions.
  • Superior Tokenizers: Instruments that effectively compress and course of information for sooner mannequin coaching.
  • Accelerated Knowledge Processing Pipeline: A system for dealing with massive datasets, powered by NVIDIA’s computing infrastructure.

A key novelty of Cosmos is its reasoning mannequin for bodily AI. This mannequin supplies builders with the power to create and modify digital worlds. They will tailor simulations to particular wants, corresponding to testing a robotic’s skill to choose up objects or assessing an AV’s response to a sudden impediment.

Key Options of NVIDIA Cosmos

NVIDIA Cosmos supplies varied elements for addressing particular challenges in bodily AI improvement:

  • Cosmos Switch WFMs: These fashions take structured video inputs, corresponding to segmentation maps, depth maps, or lidar scans, and generate controllable, photorealistic video outputs. This functionality is especially helpful for creating artificial information to coach notion AI, corresponding to programs that assist AVs establish objects or robots acknowledge their environment.
  • Cosmos Predict WFMs: Cosmos Predict fashions generate digital world states based mostly on multimodal inputs, together with textual content, photos, and video. They will predict future eventualities, corresponding to how a scene may evolve over time, and assist multi-frame era for advanced sequences. Builders can customise these fashions utilizing NVIDIA’s bodily AI dataset to satisfy their particular wants, corresponding to predicting pedestrian actions or robotic actions.
  • Cosmos Cause WFM: The Cosmos Cause mannequin is a completely customizable WFM with spatiotemporal consciousness. Its reasoning skill permits it to know each spatial relationships and the way they alter over time. The mannequin makes use of chain-of-thought reasoning to research video information and predict outcomes, like whether or not an individual will step right into a crosswalk, or a field will fall off a shelf.

Functions and Use Circumstances

NVIDIA Cosmos is already having a major influence on the trade, with a number of main firms adopting the platform for his or her bodily AI tasks. These early adopters spotlight the flexibility and sensible influence of Cosmos throughout varied sectors:

  • 1X: Utilizing Cosmos for superior robotics to enhance their skill to develop AI-driven robots.
  • Agility Robotics: Increasing their partnership with NVIDIA to make the most of Cosmos for humanoid robotic programs.
  • Determine AI: Using Cosmos to advance humanoid robotics, specializing in AI that may carry out advanced duties.
  • Foretellix: Making use of Cosmos in autonomous car simulation to generate a variety of testing eventualities.
  • Skild AI: Utilizing Cosmos to develop AI-driven options for varied functions.
  • Uber: Integrating Cosmos into their autonomous car improvement to enhance coaching information for self-driving programs.
  • Oxa: Utilizing Cosmos to speed up industrial mobility automation.
  • Digital Incision: Exploring Cosmos for surgical robotics to enhance precision in healthcare.

These use circumstances reveal how Cosmos can meet a variety of wants, from transportation to healthcare, by offering artificial information for coaching these bodily AI programs.

Future Implications

The launch of NVIDIA Cosmos is vital for the event of bodily AI programs. By providing an open-source platform with highly effective instruments and fashions, NVIDIA is making bodily AI improvement accessible to a wider vary of builders and organizations. This might result in vital developments in a number of areas.

In autonomous transportation, enhanced coaching information and simulations might result in safer and extra dependable self-driving vehicles. In robotics, the sooner improvement of robots able to performing advanced duties might remodel industries corresponding to manufacturing, logistics, and healthcare. In healthcare, applied sciences like surgical robotics, as explored by Digital Incision, might enhance the precision and outcomes of medical procedures.

The Backside Line

NVIDIA Cosmos performs an important position within the improvement of bodily AI. This platform permits builders to generate high-quality artificial information by offering pre-trained, physics-based world basis fashions (WFMs) for creating practical simulations. With its open-source entry, superior options, and moral safeguards, Cosmos is enabling sooner, extra environment friendly AI improvement. The platform is already driving main developments in industries like transportation, robotics, and healthcare, by offering artificial information for constructing clever programs that work together with the bodily world.

#IROS2024 – tweet round-up – Robohub


The 2024 IEEE/RSJ Worldwide Convention on Clever Robots and Methods (IROS 2024) was held from 14-18 October in Abu Dhabi, UAE. We check out what the members bought as much as.




AIhub
is a non-profit devoted to connecting the AI group to the general public by offering free, high-quality data in AI.


AIhub
is a non-profit devoted to connecting the AI group to the general public by offering free, high-quality data in AI.



Ian Riopel, CEO and Co-Founding father of Root.io – Interview Collection


Ian Riopel, CEO and Co-Founding father of Root.io, leads the corporate’s mission to safe the software program provide chain with cloud-native options. With over 15 years in tech and cybersecurity, he has held management roles at Slim.AI and FXP, specializing in enterprise gross sales, go-to-market technique, and public sector progress. He holds an ACE from MIT Sloan and is a graduate of the U.S. Military Intelligence Faculty.

Root.io is a cloud-native safety platform designed to assist enterprises safe their software program provide chain. By automating belief and compliance throughout growth pipelines, Root.io allows quicker, extra dependable software program supply for contemporary DevOps groups.

What impressed the founding of Root, and the way did the concept for Automated Vulnerability Remediation (AVR) come about?

Root was born from a deep frustration we repeatedly confronted firsthand: organizations dedicating huge quantities of time and sources to chasing vulnerabilities that by no means totally went away. Triage had grow to be the one protection towards quickly accruing CVE technical debt, however with the speed of rising vulnerabilities, triage alone merely is not sufficient anymore.

As maintainers of Slim Toolkit (previously DockerSlim), we had been already deeply engaged in container optimization and safety. It was pure for us to ask: What if containers may proactively repair themselves as a part of the usual software program growth lifecycle? Automated fixing, now referred to as Automated Vulnerability Remediation (“AVR”),  was our resolution—an strategy not targeted on triage and record constructing, however mechanically eliminates them, instantly in your software program, with out introducing breaking adjustments.

Root was previously referred to as Slim.AI—what prompted the rebrand, and the way did the corporate evolve throughout that transition?

Slim.AI started as a instrument to assist builders reduce and optimize containers. However we quickly realized our expertise had developed into one thing much more impactful: a robust platform able to proactively securing software program for manufacturing at scale. The rebrand to Root captures this transformative shift—from a developer optimization instrument to a strong safety resolution that empowers any group to fulfill rigorous safety calls for round open-source software program in minutes. Root embodies our mission: attending to the basis of software program threat and remediating vulnerabilities earlier than they ever grow to be incidents.

You have received a staff with deep roots in cybersecurity, from Cisco, Trustwave, and Snyk. How did your collective expertise form the DNA of Root?

Our staff has constructed safety scanners, defended world enterprises, and architected options for among the most delicate and high-stakes infrastructures. We have grappled instantly with the trade-offs between velocity, safety, and developer expertise. This collective expertise basically formed Root’s DNA. We’re obsessive about automation and integration—not merely figuring out safety points however fixing them swiftly with out creating new friction. Our expertise informs each resolution, making certain that safety accelerates innovation relatively than slows it down.

Root claims to patch container vulnerabilities in seconds—no rebuilds, no downtime. How does your AVR expertise really work beneath the hood?

AVR works instantly on the container layer, swiftly figuring out weak packages and patching or changing them throughout the picture itself—with out requiring complicated rebuilds. Consider it as seamlessly hot-swapping weak code snippets with safe replacements whereas preserving your dependencies, layers, and runtime behaviors. No extra ready on upstream patches, no must re-architect your pipelines. It is remediation on the velocity of innovation.

Are you able to clarify what units Root aside from different safety options like Chainguard or Rapidfort? What’s your edge on this area?

Not like Chainguard, which mandates rebuilds utilizing curated photos, or Rapidfort, which shrinks assault surfaces with out instantly addressing vulnerabilities, Root instantly patches your present container photos. We seamlessly combine into your pipeline with out disruption—no friction, no handoffs. We’re not right here to interchange your workflow, we’re right here to speed up and improve it. Each picture that runs by Root basically turns into a golden picture—totally secured, clear, managed–delivering speedy ROI by slashing vulnerabilities and saving time. Our platform cuts remediation from weeks or days to only 120-180 seconds, enabling corporations in extremely regulated industries to remove months-long vulnerability backlogs in a single session.

Builders needs to be targeted on constructing and delivery new merchandise – not spending hours fixing safety vulnerabilities, a time-consuming and sometimes dreaded facet of software program growth that stalls innovation. Worse, many of those vulnerabilities aren’t even their very own – they stem from weaknesses in third-party distributors or open-source software program initiatives, forcing groups to spend beneficial hours fixing another person’s drawback.

Builders and R&D groups are among the many largest price facilities in any group, each when it comes to human sources and the software program and cloud infrastructure that helps them. Root alleviates this burden by leveraging agentic AI, relatively than counting on groups of builders working across the clock to manually test and patch recognized vulnerabilities.

How does Root particularly leverage agentic AI to automate and streamline the vulnerability remediation course of?

Our AVR engine makes use of agentic AI to copy the thought processes and actions of a seasoned safety engineer—quickly assessing CVE affect, figuring out one of the best out there patches, rigorously testing, and safely making use of fixes. It accomplishes in seconds what would in any other case require important guide effort, scaling throughout hundreds of photos concurrently. Each remediation teaches the system, constantly enhancing its effectiveness and flexibility, basically embedding the experience of a full-time safety engineer instantly into your photos.

How does Root combine into present developer workflows with out including friction?

Root effortlessly integrates into present workflows, plugging instantly into your container registry or pipeline—no rebasing, no new brokers, and no extra sidecars. Builders push photos as regular, and Root handles patching and publishing up to date photos seamlessly in place or as new tags. Our resolution stays invisible till wanted, providing full visibility by detailed audit trails, complete SBOMs, and easy rollback choices when desired.

How do you steadiness automation and management? For groups that need visibility and oversight, how customizable is Root?

At Root, automation enhances—not diminishes—management. Our platform is extremely customizable, permitting groups to scale the extent of automation to their particular wants. You resolve what to auto-apply, when to contain guide overview, and what to exclude. We offer in depth visibility by detailed diff views, changelogs, and affect analyses, making certain safety groups stay knowledgeable and empowered, by no means left at midnight.

With hundreds of vulnerabilities mounted mechanically, how do you guarantee stability and keep away from breaking dependencies or disrupting manufacturing?

Stability and reliability underpin each motion that Root’s AVR takes. By default, we undertake a conservative strategy, meticulously monitoring dependency graphs, using compatibility-aware patches, and rigorously testing each remediated picture towards all publicly out there testing frameworks for open-source initiatives earlier than deployment. Ought to a problem ever come up, it is caught early, and rollback is easy. In apply, we’ve maintained lower than a 0.1% failure price throughout hundreds of automated remediations.

As AI advances, so do potential assault surfaces. How is Root making ready for rising AI-era safety threats?

We view AI as each a possible risk vector and a defensive superpower. Root is proactively embedding resilience instantly into the software program provide chain, making certain that containerized workloads—together with complicated AI/ML stacks—are constantly hardened. Our agentic AI evolves as threats evolve, autonomously adapting defenses quicker than attackers can act. Our final aim is autonomous software program provide chain resilience: infrastructure that defends itself on the velocity of rising threats.

Thanks for the good interview, readers who want to be taught extra ought to go to Root.io