Robots today see the world through powerful world models, but their inner life is mostly a blur of frames, tokens, and logs. They react, but they rarely remember.
At the same time, Artificial Intelligence is accelerating into our world in ways that feel less human and more transactional. Tools lack empathy, safety, and transparency.
Introduced here are the founders of Loosh, who saw an opportunity to bridge the two paths—if AI could be built to reflect the qualities of consciousness itself: distributed, self-improving, ethically aligned, and expressive—then it could move beyond tools and become a real cognitive partner in human evolution with robotics.
Loosh’s tagline is machine consciousness—but what does that really mean? According to the founders, they are “designing embodied AI that is self-reflecting, ethically aligned, and emotionally aware. Starting with our proprietary micro-services architecture, our technology enables robotics and agentic systems with complex persistent memory, nuanced reasoning, and sophisticated awareness of intent,” says cofounder Chris Sorel.
“Our mission is to create AI that partners with humans in trusted environments. This is not about machines replacing people, it is about creating systems that give us back time, expand creativity, and amplify what humanity can achieve,” says cofounder Lisa Cheng.
Loosh AI has big goals to help accelerate the next stage in the evolution of human-machine collaboration.
They’re betting that in the next 5 years, the average family will be faced with a decision to buy a robot or a car, and although a car is useful, the home companion robot will be infinitely more useful, compounded in the ways it helps across the family and the home by freeing up time.
However, today’s AI has many shortcomings, which is where Loosh provides the answer for AI to understand what is happening now, what has happened before, and what matters for what happens next. They’ve built a platform that introduces a cognitive system turning raw perception and predictions into a continuous stream of structured understanding that is revisited and revised over time.
Their system works with third-party World Models and integrates them into a larger cognitive architecture, effectively placing the AI inside a container that applies constraints.
It does this with the Context Builder, which assembles temporal knowledge graphs and semantically searchable memory with symbolic ontologies. Every event, object, and interaction is anchored in time and linked to other concepts—forming a temporal knowledge graph that the robot and agent can reason over.
In addition, Loosh has developed a Memory Fabric: a highly performant, semantically searchable working memory cache, keeping the most relevant and recent context immediately available for decision making.
Despite Loosh’s lofty ambitions, the implementation is pragmatic and approaches machine consciousness with today’s tools to solve complex reasoning, long-term memory, emotional awareness, self-reflection, self-improvement, and the creation of relatable personalities. The result is a cognitive architecture for robotics that doesn’t just allow it to perceive but experience its environment over time.
And after a year of building in stealth, they recently made their debut on the Bittensor network as Subnet 78—leveraging a global network of distributed GPUs to train inference models. Effectively solving their scaling problem.
This means that rather than raising funds to pay for the H100s or Nvidia’s latest Thor, Loosh has skipped all of that and instead created a competition to allow Data Scientists around the world to help train their model and battle test its cognitive system.
The winners are compensated directly from the Bittensor network itself, sort of like how Lewis Hamilton may race the latest Ferrari spec, but ultimately, the prize money is awarded by F1 racing. In this scenario, Loosh is Ferrari, and Bittensor is F1. And all the other racers are data scientists running the latest GPUs on the network, where each race is a different subnet, with a different track, conditions, and rewards.
Back to thinking about cognitive systems, Loosh is grounded in ethics, safety, and a deep understanding of human verbal and non-verbal communications, coupled with expressive, emotive, and relatable personalities for robots and agentic systems that engender trust, confidence, and amity in the people they work with.
Loosh is building partners, not servants. And they’ve just completed the first part of their roadmap—the beta version of their environment is running, and the system is actively starting to learn how to apply deontology and ontological rule sets to queries fed by users through a chat interface. More than a chatbot, it’s like building a high-performance racecar—we start in a simulator first—the chat interface. To stand this up on Bittensor, the team has created a validator and miner codebase to get people involved in the ‘race to train cognitive AI.’
Despite its seemingly complex nature, anyone with a GPU at home can become a miner or validator—placing themselves in a global race and getting paid for it at the same time.
If you think you have what it takes, there is no setup fee to get started—you just need a computer with a GPU, an internet connection, and ADHD (which is in abundance these days).
The next phase of their plans includes integrating brain data, because what is better than a robot knowing exactly how you think and feel? It’s knowing that eventually the robot dog you had always dreamed of as your best friend is going to be a reality.
They have also tapped a senior neuroscientist to create their proprietary data model that takes raw EEG data, and after some fine-tuning, the AI model is able to guess the emotional state of the subject with 70% accuracy.
Their immediate plans include creating another Bittensor-style competition with EEG data to perfect the model that will eventually be used for Sorel’s prototype of a robot dog, which he is building at home for his kids.
Although they have not raised funds, the team is on the path to scale and grow—eventually creating a prototype for a wearable that will feed live brain data into your robot at home for real-time inference and human interaction.
What Loosh is really doing is taking the future of robotics out of the lab and into an open arena where anyone can compete. If world models are the eyes, Loosh is trying to build the part that remembers, reflects, and chooses with intent. And by putting that system on Bittensor, they are turning the scaling problem into a race, with miners and validators acting as the drivers and pit crew, and engineers who make the car faster, safer, and more reliable over time.
The near-term story is practical. A validator and miner stack is now live. A cognition layer is being trained in public. The first environment is running, learning how to apply ethics and ontologies to real queries. The next phase adds a higher bar: emotional inference, starting with EEG, moving toward live signals, and eventually embodied companions that can understand not just what you said, but what you meant.
If it works, the impact is bigger than a new model or a new subnet. It is a shift in how intelligence is built, measured, and distributed. Less closed, less opaque, less transactional. More accountable, more aligned, and more human compatible.
When the robot dogs arrive, it won’t replace the companionship of an 18-year-old Multipoo—but with Loosh’s cognitive architecture, it can try, and it might come close.
Source link
#Robot #Learn
