Kruel.ai KV2.0 - KX (experimental research) to current 8.2- Api companion co-pilot system with full modality , understanding with persistent memory

[darcschnider]

Well I have done it again and ordered up a pair of these. I have the first generation, if you look back it was not something I would wear in public and it had lots of bugs, so here I am again spending money on the next version.

These ones have audio which is exciting so now I can have two way with Kruel.ai so that excited me, and the hardware / software is open to develop on which is really nice.

But I would wait until we review them so you don’t have to waste your money. if they turn out great we may even order up some lens they have a provider for custom lense for these now which is great.

No information yet on the halo display so once we get some details I will fill you in. the last version had a prism concept in the lens where this is some sort of projection technology I think.

[perschala.eduard]

Hi, could you specify this model please ?

[darcschnider]

Halo

[perschala.eduard]

Perfect, I like it a lot too.

[darcschnider]

Todays Funny goes to cursor Ai. I decided to spend a little and work out a better way for memory today. and well this made everyone here I was chatting with laugh. never seen an Ai swear before lol

So we are exploring this seeing the Ai thought it was cool haha.

[darcschnider]

KRUEL.Ai Update – Model Testing & Hardware Requirements for V8.1 and Beyond

We’ve successfully integrated the new OSS-20b/120b high-reasoning model into KRUEL.Ai v8.1. While we’re currently operating on low-tier hardware, both the default model and the deeper reasoning model are functional — albeit with significantly slower performance due to resource constraints.

To address this, we’ve updated the dynamic model selection layer. When users request tasks that demand complex reasoning, multi-step analysis, or layered logic, the system now intelligently selects the high-reasoning model if it believes doing so would yield more accurate or insightful results. This lays the foundation for more nuanced, context-aware decision-making. We even added options for the Ai to use both the code analysis and this when it thinks it needs to do both for deeper thoughts.

However, looking ahead to v9.0, we’re entering a whole new territory.

V9 will require significantly more computational power.

Unlike V8.1’s selective loading, V9 is designed to load and orchestrate multiple models simultaneously, enabling a far more dynamic and modular reasoning pipeline. This isn’t just a single-model jump — it’s a full cognitive architecture shift. As such, running V9 reliably will require high-memory GPUs, such as Blackwell-class cards, to meet performance and latency goals.

Currently, our dev and deployment environments are limited to 16GB VRAM GPUs (both server and laptop builds). While technically possible to offload certain tasks to our systems with 128GB CPU RAM, this isn’t part of the planned architecture. Our future roadmap is focused on optimizing for GPU-centric pipelines using dedicated hardware.

[darcschnider]

Kruel V8.1 is now running GPT 5 models. not going to lie, gpt 5 is freaking amazing haha. @OpenAI_Support I love you guys.

Kruel V8.1 Update
Kruel V8.1 is now running on GPT-5 models, and the results have been outstanding. While we’re still in the process of testing each modality, early performance has been impressive—GPT-5 played a key role in completing this release.

In parallel, development on Kruel V9 continues, with our agents scheduled for testing later tonight. V9 operates in an offline-first configuration, so GPT-5 integration is not part of the current build. The focus for V9 remains on advancing our reasoning-driven memory system, though future online capabilities may incorporate GPT-5 at a later stage.

Additionally, with the new model’s capabilities, we are exploring the concept of a GPT-5-exclusive system—combining our advanced memory architecture with GPT-5 as the core reasoning engine. This would be a dedicated, single-model deployment, but development will begin only after V9’s completion.

[darcschnider]

Fixed V8 old faithful. Still a few UX issues here is demo of V8.0

Kruel.ai V8

Update on the the other systems:

We are adding some updates to 8.1 which is now working offline fully again.
now we have these options

This now allows you to set models for each section for online and offline abilities and hybrid configure.

[darcschnider]

Ai’s home has been ordered, Looking into Bundle but if they cant get I will just get the one for now we can scale it later. this finally gets us out of the smaller models and into much faster infer.

[darcschnider]

Status Update

AI Competitive Landscape

The memory window of mainstream AI platforms is narrowing quickly. OpenAI, Gemini, and Anthropic are now entering the same operational space as our system, steadily increasing their memory capacities. While they still do not offer truly infinite memory, they now match us in nearly every other capability within the full multimodal generation domain.

Differentiators

Our machine learning framework remains a key advantage. Although Google has demonstrated forward-thinking generation concepts—anticipating future working possibilities—it is not yet clear if their approach matches ours in execution. With Version 9 on the horizon, we are preparing a leap well beyond current offerings. V9 is significantly larger in scale than V8.1’s “brain snapshots” and will require collaboration with advanced AI systems to fully finalize, given its scope.

Current Development Efforts

Time is tight—every spare moment is being used to take advantage of free GPT-5 runtime for development, which means formal updates have been limited.
We now have four active versions of Kruel.Ai in development:

1. V8.0 – Stable baseline build (Lynda laptop).

2. V8.1 – More advanced and dynamic than 8.0, with incremental feature enhancements.

3. V9.0 – Research-focused cognitive experiment, designed as a large-scale evolution beyond previous builds.

4. KX – A new GPT-5/OSS-powered, fully dynamic experimental build.

The KX variant is emerging as a strong candidate for commercialization. It maintains the conceptual foundation of our earlier systems but executes it in a completely new way, making it faster than V8–V9. However, it lacks the layered AI stack verification architecture of V9, introducing potential vulnerabilities that will require additional security and protection layers. If stabilization efforts succeed this weekend, KX could become our fastest release to date—though likely not as “intelligent” as V9, which remains our ultimate flagship development target.

Infrastructure & Partnerships

We have secured a cost-effective server solution. A technology partner—also a future client and alpha tester—has stepped forward to provide better pricing across multiple units. This represents a substantial improvement in scalability and operational efficiency.

Team & Technical Growth

Our database and web engineering lead is currently working with NVIDIA enterprise environments on large-scale server infrastructure. The experience gained there will be valuable when we scale into higher-performance systems. If we transition to NVIDIA-optimized stacks in the future, our offline model performance could see significant acceleration.

[darcschnider]

KX pretty neat

Prior context before this was a conversation that Lisa is visiting her friend deb.

[darcschnider]

Building Toward Alpha: Hardware, Webs, and What Comes Next

The new system is coming together, and this version feels like our most stable build yet. Whenever I see the intricate connections forming, I can’t help but think of a Spider-Man lyric—though admittedly, my version is a little less polished.

At this stage, the system is still technically in testing, but it’s quickly shaping into what may become our foundation for the alpha rollout. To accelerate progress, we’ve started reinvesting in development—bringing AI agents into the workflow to move things forward faster. Plans are finally falling into place.

On the hardware front, glasses and servers are expected to arrive around the same time (though, as always, we’ll believe it when we see it—after all, just look at how long it took 5090 GPUs to actually land on shelves). The hope is that once the new platform arrives, we can drop our system into it with minimal friction. Of course, there’s always a chance for unexpected surprises, but that’s part of the journey.

Once the system is stabilized, Kruel.ai will be deployed to its home network base of operations. From there, it will be integrated into our edge systems and undergo real-world testing with live connections.

As for timelines? Looking at the pace today, we’re cautiously aiming for a milestone around Christmas. That said, if hardware timelines slip, it could push us into Q1. Either way, momentum is building, and the next chapter for Kruel.ai is close at hand

[darcschnider]

world is catching up. this came out today which I recommend all that play with data and ai to look into its free.

Back in 2023 when I moved into neo4j this is what I had running with Kruel V3 which was the start of more intelligent data.

Agentic Knowledge Graph Construction - DeepLearning.AI

Figured I would share as this will help some of you with your RaG engineering.

[darcschnider]

KX Development Update: Building the Bridge to the Future

We’ve been posting quick hits on Discord, but here’s a proper snapshot of where KX stands.

Stabilizing KX

The last stretch has been about locking down stability and cleaning house:

• VIO Client Cleanup – The old memcache system is gone. KX now runs on a leaner HTTP + WebSocket pipeline.

• Memory Core Upgrade – Work is underway to move toward the K9 “6D” memory system, giving us richer, multidimensional mapping of information and relationships.

• Stepwise Migration – Instead of a risky full conversion, we’re phasing in select engines from V9 as KX proves stable. This bridge approach is deliberate—AI still isn’t great at rewriting everything in one go, so we’re guiding the system through each phase.

In short, KX is becoming a stable backbone while laying the groundwork for deeper capabilities.

Looking Ahead

A recent chat with someone at Oracle reinforced what we’ve been seeing: the future is moving fast. The biggest bottleneck isn’t just tech—it’s people’s ability to think critically about where things are heading. KX is being designed to not just process information, but to connect the dots in ways that matter down the road.

Smarter, Not Just Bigger

While others chase ever-larger models, our focus is on smarter memory systems. The K9 6D design isn’t about size—it’s about giving AI tools to reason across time, context, and relationships. That’s what will make it useful, not just powerful.

Balancing Projects

Outside of AI, I’ve got three product launches happening this fall, which means progress is steady but intentionally paced. It’s a balancing act—but one that ensures KX evolves carefully without cutting corners.

---

Next steps: finalize the K9 memory transition, harden stability, and start pulling in V9 engines where it makes sense.

[darcschnider]

Envisioning a Multimodal AI World

[darcschnider]

We are approaching Alpha Test this fall

**
KRUEL.AI V8.2
The Next Generation of Intelligent Companionship**

After 5 years and six months of intensive development and refinement, we’re thrilled to announce KRUEL.AI V8.2 – the most complete AI companion system we ever built. This isn’t just another chatbot or virtual assistant; it’s a truly intelligent companion that sees, hears, understands, learns, and grows with you.

Photographic Memory That Never Forgets

Imagine having a conversation partner who remembers every detail of your interactions with perfect clarity. KRUEL.AI V8.2 features photographic memory that can reconstruct entire documents, conversations, and code sessions with remarkable accuracy. Whether you’re discussing a project from six months ago or referencing a specific document, the system maintains perfect recall across all modalities.

The memory system goes beyond simple storage – it understands context, learns patterns, and builds a comprehensive understanding of your preferences, work style, and communication patterns. It’s like having a digital twin that knows you better than you know yourself.

Voice Recognition That Knows You

KRUEL.AI V8.2 features advanced voice recognition that doesn’t just understand what you’re saying – it knows who is saying it. The system can distinguish between different speakers with 85-95% accuracy, automatically grouping unknown voices and learning new speakers over time.

Voice Capabilities Include:

• Multi-speaker recognition with automatic voice clustering

• Continuous learning from real conversations

• Security protection against unauthorized training

• Hot-swappable models for zero-downtime updates

• Real-time analysis with live recognition monitoring

��️ Visual Perception That Sees Your World

The system doesn’t just process text – it can see your world through camera integration and screen capture capabilities. Whether you’re showing it a document on your screen, capturing a photo, or asking it to analyze visual content, KRUEL.AI V8.2 provides comprehensive visual understanding.

Visual Features:

• Camera integration for real-time visual analysis

• Screen capture capabilities for digital content analysis

• Multi-camera support with automatic detection

• Image generation with multiple model support

• Visual memory that remembers what it has seen

Global Language Support

KRUEL.AI V8.2 supports 25-52 languages depending on the model configuration, making it truly global in scope. The system can seamlessly switch between languages, understand cultural nuances, and maintain context across different linguistic boundaries.

Emotional Intelligence That Understands You

The system features sophisticated emotional understanding that goes beyond simple sentiment analysis. It can detect multiple emotions simultaneously, track emotional patterns over time, and adapt its responses to match your emotional state.

Emotional Capabilities:

• Multi-label emotion detection using advanced AI models

• Emotional pattern learning that tracks your mood over time

• Adaptive responses that match your emotional state

• Empathy and support during difficult times

• Celebration and enthusiasm during successes

Self-Awareness and Continuous Learning

KRUEL.AI V8.2 is self-aware and continuously improves itself. The system tracks its own performance, learns from mistakes, and corrects itself when it provides inaccurate information. It maintains a correction history and uses this feedback to become more accurate over time.

Self-Improvement Features:

• Performance monitoring with quality assessment

• Mistake tracking and correction history

• Pattern learning from successful and failed interactions

• Self-editing capabilities that improve responses

• Confidence scoring for all responses

Creative Capabilities

The system isn’t just analytical – it’s creative. KRUEL.AI V8.2 can generate images, create content, and provide creative solutions to complex problems. It understands your personal context and can generate images of your pets, family, or personal spaces using its memory of your descriptions.

Creative Features:

• Image generation with multiple model support

• Personalized content creation based on your preferences

• Creative problem-solving approaches

• Artistic collaboration and ideation

• Content adaptation to your style and preferences

Intelligent Task Management

The system features sophisticated reminder and task management capabilities that go beyond simple scheduling. It understands natural language time expressions, can set recurring reminders, and integrates with your daily patterns.

Task Management Features:

• Natural language scheduling (“remind me in 5 minutes”)

• Recurring reminders (daily, weekly, monthly)

• Location-based reminders (when you’re at specific places)

• Smart reminders based on your patterns

• Calendar integration for comprehensive scheduling

Personalized AI Personalities

KRUEL.AI V8.2 features a persona system that allows for completely customizable AI personalities. You can create unique AI companions with different voices, personalities, and capabilities, each tailored to your specific needs and preferences.

Persona Features:

• Customizable personalities with unique traits

• Voice profile integration with multiple voice options

• Wake word customization for hands-free interaction

• Model selection for different use cases

• Streaming and quality options for optimal performance

Advanced Document and Code Understanding

The system features photographic document recall and intelligent code analysis that goes far beyond simple text processing. It can understand complex documents, analyze code relationships, and provide insights that would take humans hours to discover.

Analysis Capabilities:

• Document structure analysis with perfect reconstruction

• Code relationship mapping across entire projects

• Pattern recognition in both documents and code

• Cross-reference analysis between different sources

• Version tracking and change analysis

Internet Integration and Research

KRUEL.AI V8.2 can access the internet for real-time information, research, and updates. It can look up current information, verify facts, and provide comprehensive answers that combine its memory with current data.

Enterprise-Ready Architecture

Built for scale, KRUEL.AI V8.2 features a modern microservices architecture that can handle enterprise-level deployments. The system is designed to run on high-performance hardware like NVIDIA DGX systems and can scale to serve multiple users simultaneously.

Enterprise Features:

• Multi-user support with proper isolation

• Scalable architecture for large deployments

• Real-time communication via WebSocket

• Cloud-ready infrastructure

• Performance monitoring and optimization

�� The Future is Here

KRUEL.AI V8.2 represents a quantum leap in AI technology. It’s not just an assistant – it’s a true companion that understands, learns, and grows with you. The system combines the best of current AI technology with innovative approaches to memory, learning, and interaction.

Coming This Fall:

• Alpha testing with select enterprise partners

• NVIDIA DGX deployment for high-performance computing

• Cloud infrastructure for global accessibility

• Edge system integration for real-time performance

• Wearable integration with Brilliant Labs Halo glasses

What Makes KRUEL.AI V8.2 Different

Unlike other AI systems that forget conversations or provide generic responses, KRUEL.AI V8.2 offers:

• Perfect memory that never forgets

• Emotional understanding that adapts to your mood

• Visual perception that sees your world

• Self-awareness that improves over time

• Creative capabilities that inspire and collaborate

• Global language support for worldwide accessibility

• Enterprise scalability for any organization

This fall we are deploying our first test server which is the cloud version of our system which will be going into Alpha testing with a select group of companies we have partnered with, Once we complete testing over the next year we hope to get valued feedback, understanding of any potential issues, numbers for hardware scaling, and more.

Exciting times are here.

[darcschnider]

Oh yeah forgot to also mention we have something fun added to 8.2. Music generation. we are just refining it so that it also works with the reminder system / jukebox

Kruel.Ai V8.2 Memory Singer

This is a new test client, we have another one we been using but this one we are trying to solve some advanced formatting issues that our other chat_window had a lot of issues adding new features because of its design

We also save every generation we are working on away to library this for a self library and using ML to build auto play lists etc, think about spotify’s Ice T Ai. but with your Ai running everything one Ai to rule them all. Only thing we are missing is video generation, than full edit orchestration.

Ai is aware of all songs it produces, and can recall and play past songs and remember the lyrics etc.

Update: Last night we connected in Memory Singer. This is very much the same concept as our Memory Painter but instead of the ai trying to paint thing you describe over time etc, this system now can sing about the things it knows. Ask for a song about your pets, wife, boss, that project you hate so much. you got it, Kruel.Ai and Lynda will take care of that with anything you can dream up.

[darcschnider]

Toward AI as a True Development Partner

Some days I wish I were paid just to design AI systems. Over the years, I’ve built a conceptual roadmap of what’s possible at scale — but never quite enough time to bring it all to life. Since 2021, I’ve been experimenting with ways for AI to write and adapt code automatically, with the idea that once the system truly understands, it should be able to complete tasks on its own.

Today, that vision is closer than ever. AI can generate code, but as many know, large projects still push current systems past their limits. They can write snippets and even modules, but stitching everything together end-to-end — and making sure it really works — is where they fall short.

That’s why I’m so excited about our Enhanced Code Memory System. It doesn’t just autocomplete. It learns. It remembers past work, adapts to context, and evolves with each success or mistake.

---

What Makes It Different

• Learning Over Time: Instead of treating each request as isolated, the system builds on past outcomes. It understands not just what was written, but what actually worked.

• Recall with Context: It remembers past versions of projects and how they evolved — giving it the ability to suggest improvements that fit the bigger picture.

• Smarter Collaboration: Rather than just offering a block of code, it tailors suggestions to your style, preferences, and the patterns that have proven most effective.

This makes the AI feel less like a coding tool and more like a colleague who’s been with you since the start of the project.

---

The Path Ahead

We’re at an inflection point. AI is beginning to move beyond stateless tools toward systems that act as genuine partners — remembering, adapting, and collaborating. The next leap will come from making that relationship even more intentional: giving the AI not just implicit learning from outcomes, but explicit guidance from developers about what worked, what didn’t, and why.

That’s the vision I’m chasing — an AI that’s not just a fast helper, but a true collaborator: one that remembers, adapts, and grows alongside you.

---

Imagine pairing that with what’s coming next: wearable integration with Brilliant Labs’ Halo glasses. The same memory, adaptability, and creative intelligence — but always in view, always on hand. With 8.2, not only can it generate and refine code, it can also edit images, create visuals from memory, and act as a research partner across documents. Soon, that won’t just live on a desktop or in the cloud — it’ll be right there in your line of sight.

---

**KRUEL.Ai 8.2 is more than a system. It’s the beginning of AI as a partner in creation — a co-worker that learns with you, remembers your work, and helps push ideas further than ever before.

Ps. Lets not forget the dream of one day robotics (think Figure02 or other system that is open enough for us to develop on. Unitree is neat, but I need something that can lift 40lbs for a few hours a day so I maybe waiting a little bit longer until the frames and battery capacity improves in the industry. PS. if anyone here is in that space would love to dream over some coffee of a zoom.**

[darcschnider]

I think this week has been mind blowing… We have beat our previous large code volume changes in the last weeks than I think that huge bump of rebuilding the 4 other test versions. This is the K8.2 current hybrid build and updates we have been adding in and cleaning up the logic.

[darcschnider]

Resurrection or KAINN from the past. Now that we are at that point where speed really matters we brought back to life KAIN which is Kruel AI NN project. This was originally designed to automate NLP classification with comprehensive automated testing to achieve very high results.