Shadow59
Regular
You beat me to it. I'll delete mine
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
wilzy123
Founding Member
Hi rob,
I usually don't respond to what might be called would-be Cassandras, but you have taken some time to elucidate your thoughts.
I thought that when I bought in 5 years ago, the SP was about to explode because the company was giving a geeks presentation in the US, and I still believe the SP is about to explode. It's just taken a bit longer, or maybe it's a slow explosion ...
Starting with your 4 questions:
a response to Q1 would be the sort of thing for which the ASX would demand further and better particulars, which would put us in conflict with all the NDAs, a la Ford;
Similarly for Q2;
Q3 is reasonable, but I think this has been answered with the degree of specificity allowable under the NDAs - automotive, health, mechanical maintenance, aerospace, ... . As I said 5 years ago, it would be easier to list the areas where Akida cannot be used (it's a very short list).
Q4, see Q1.
Taking your example of mobile phones, the development cycle is several years, so, while we missed the recent releases and those shortly upcoming, there is nothing to suggest we are not being considered for future releases. Our association with Prophesee gives us a strong link into the process, especially given Synsense's admission of its low level applicability.
Remember BRN is on a steep learning curve. We need to discover what our customers need, and they don't know because they don't know how revolutionary Akida is. Last year we were talking about the recently invented LSTM functionality for "attention", and now we have bypassed that for ViT/TeNNs in Akida 2. This is not to say Akida 1 is obsolete. We have developed Akida 1500, which is Akida 1 el cheapo, or at least stripped of the ARM Cortex processor, available from 2 nodes (8 NPUs ) up to 20 nodes, whether you want an always-on wake-word watching device, a vibration monitor, or a full video processing chip.
The IP licensing business is a long haul project, much longer than selling chips.
You know that Renesas and Megachips are bringing out processors incorporating Akida 1. Renesas has its own in-house DRP-AI but has found applications where Akida is more relevant.
I find it encouraging to remember:
Prophesee loves us;
nViso loves us;
NASA is a fan;
Mercedes is impressed;
Valeo is a partner;
we are compatible with all ARM processors;
SiFive, ARM's up and coming rival, is also a partner;
...
It is true that the company has had a few major business model changes, and this has inevitably shifted expectations. We've been a software company, a chip company, and now an IP licensing company. Software provides only a pale shadow of Akida's capabilities. Chips would have produced a quicker return, but again, without maximizing the benefits of the Akida design. Would we produce different sizes for different applications. Putting all our eggs in the IP licensing basket makes it a harder sell because of the up-front costs to the customer. But I am trusting that the superior performance of Akida will win over customers like Mercedes, and partners like Prophesee and SiFive, and marquee clients like NASA/DoD/DoE (DOE - that's right, we do do cybresecurity too).
The SP is down now, but within a few (short?) months ...
[Not financial advice, just part of an online discussion - if you have questions about investing in Brainchip, seek professional advice or, better still, do your own research and avoid the variegated foolish professional nay-sayers]
Fullmoonfever
Top 20
Just on SiFive D.
We know NASA has also been looking at the X280.
I noticed on that recent presentation table that the X280 is also slated for 2025 with the RISC-V cores TBD.
To be designed? To be decided? To be developed?
Given we are also on the table stand alone and also have the SiFive relationship, do we have the capability to also be in the X280 cores mix possibly?
Schnitzel lover
Regular
Think you may have to direct that message to another few dozen posters.
Thanks but no thanks Social media memester.
chapman89
Founding Member
Well said my friend
Everybody keeps forgetting VALEO, Brainchip in my opinion is behind Scala3 which has already taken more than $1 billion euros in pre orders, and that was at the back end of last year, so maybe they’ve taken more pre orders?
Regardless, scala3 is hitting the market at the end of this year/very early 2024 in 2024 model vehicles.
Who here remembers the Valeo premier back in 2021? I certainly do. If you need a refresher, watch the video again.
Also, as @Diogenese can attest, automotive takes 3 years minimum, we (Brainchip) have a joint development with Valeo where we get paid on undisclosed milestones, that was announced back in June 2020.
I’ve posted this before, but on the CTO of Valeo interview post on LinkedIn about 2 months ago, I asked if they’ve achieved Scala3 by neuromorphic, this was the response I got from a Valeo employee below.
It is happening, believe it, we will get paid and that will be a recurring income and will only grow as Scala3 becomes standard for even more vehicles.
Of course my opinion but based on the facts out there, who else can supply Valeo with their neuromorphic IP?? There is nobody. We’ve been through this before!
equanimous
Norse clairvoyant shapeshifter goddess
Im glad you verified thatWell said my friend
Everybody keeps forgetting VALEO, Brainchip in my opinion is behind Scala3 which has already taken more than $1 billion euros in pre orders, and that was at the back end of last year, so maybe they’ve taken more pre orders?
Regardless, scala3 is hitting the market at the end of this year/very early 2024 in 2024 model vehicles.
Who here remembers the Valeo premier back in 2021? I certainly do. If you need a refresher, watch the video again.
Also, as @Diogenese can attest, automotive takes 3 years minimum, we (Brainchip) have a joint development with Valeo where we get paid on undisclosed milestones, that was announced back in June 2020.
I’ve posted this before, but on the CTO of Valeo interview post on LinkedIn about 2 months ago, I asked if they’ve achieved Scala3 by neuromorphic, this was the response I got from a Valeo employee below.
View attachment 42507
View attachment 42508
It is happening, believe it, we will get paid and that will be a recurring income and will only grow as Scala3 becomes standard for even more vehicles.
Of course my opinion but based on the facts out there, who else can supply Valeo with their neuromorphic IP?? There is nobody. We’ve been through this before!
HopalongPetrovski
I'm Spartacus!
This is the bee's knees straight from the Diogenese.
Way to calm the horses O sitter of barrels.
And so Factual.
It's almost as if you were channeling a beloved ghosty.
But from whichever noggin this sprang, it's gold.
Thank you for this salving contribution to our collective angst.
Bring it, BrainChip!
robsmark
Regular
Hey Dio,
Thank you for the detailed response.
You’re a valued contributor here on a technical front, and I appreciate your input.
I’m not sure I appreciate your consideration of me being a “Cassandra“. I don’t think it’s fair to play this off as an emotional post, as would stand by it even if the SP shot off to $5 on open tomorrow.
Starting with your 4 questions:
a response to Q1 would be the sort of thing for which the ASX would demand further and better particulars, which would put us in conflict with all the NDAs, a la Ford;
I‘m pleased you brought this up. Are we still engaged with Ford? It’s been over three years now (25/5/2020) since Brainchip announced that Ford were ”evaluating“ Akida and since then we have heard nothing. This is my point Dio, there is no transparency. Is three years‘ an unreasonable duration of time for a shareholder update? I think not. For all we know, Ford have gone quiet and the company have decided not to disclose it due to a technicality - perhaps they never formally terminated the agreement. I’m not saying this has happened, but it’s certainly plausible after this duration.
Similarly for Q2;
Q3 is reasonable, but I think this has been answered with the degree of specificity allowable under the NDAs - automotive, health, mechanical maintenance, aerospace, ... . As I said 5 years ago, it would be easier to list the areas where Akida cannot be used (it's a very short list).
It hasn’t been answered at all. The company have alluded to working with these sectors. And to state it’s easier to state where Akida cannot be used is besides the point. I understand the use-cases are practically infinite, but that doesn’t mean we are engaged with infinite “customers”, and nobody knows, as we aren’t updated of any specifics.
Q4, see Q1.
Taking your example of mobile phones, the development cycle is several years, so, while we missed the recent releases and those shortly upcoming, there is nothing to suggest we are not being considered for future releases. Our association with Prophesee gives us a strong link into the process, especially given Synsense's admission of its low level applicability.
There is nothing to support that we are being considered for future releases either. Absolutely nothing.
The fact is that Brainchip hasn’t mentioned the EAP is a very long time. Nothing about the participants, nothing about progress, nothing about the potential. Simply nothing. Woulds, coulds, and maybes won’t change this. The company needs to update the market on the status of this program without breaking contractual obligations. It shouldn’t be hard, they’ve have years to figure this out, unless of course there’s nothing to update?
Remember BRN is on a steep learning curve. We need to discover what our customers need, and they don't know because they don't know how revolutionary Akida is. Last year we were talking about the recently invented LSTM functionality for "attention", and now we have bypassed that for ViT/TeNNs in Akida 2. This is not to say Akida 1 is obsolete. We have developed Akida 1500, which is Akida 1 el cheapo, or at least stripped of the ARM Cortex processor, available from 2 nodes (8 NPUs ) up to 20 nodes, whether you want an always-on wake-word watching device, a vibration monitor, or a full video processing chip.
The IP licensing business is a long haul project, much longer than selling chips. - If I recall correctly, Renesas came out of nowhere, the company were surprised by this licensing agreement, so I find your argument weak here. It’s been years since our last licensing agreement, and the Company itself has admitted that this is taking longer than expected, and that was a while ago! Where are they??? Why aren’t they licensing a groundbreaking innovation like Akida???
You know that Renesas and Megachips are bringing out processors incorporating Akida 1. Renesas has its own in-house DRP-AI but has found applications where Akida is more relevant.
I find it encouraging to remember:
Prophesee loves us; Where is the licence?
nViso loves us; Where is the licence?
NASA is a fan;
Mercedes is impressed; Why isn’t it in their cars, and licensed?
Valeo is a partner; Where is the licence?
we are compatible with all ARM processors; Then why hasn’t anyone licensed it?
SiFive, ARM's up and coming rival, is also a partner; Where is the licence?
...My intention isn’t to be objective for the sake of being objective. The point is it’s all hearsay until someone signs and we generate revenue, for some reason that hasn’t happened, and I want to know what the feedback is as to why that is.
We cannot survive on partnerships, we aren’t a university or a funded research centre, we are a company that needs revenue and answers to shareholders.
It is true that the company has had a few major business model changes, and this has inevitably shifted expectations. We've been a software company, a chip company, and now an IP licensing company. Software provides only a pale shadow of Akida's capabilities. Chips would have produced a quicker return, but again, without maximizing the benefits of the Akida design. Would we produce different sizes for different applications. Putting all our eggs in the IP licensing basket makes it a harder sell because of the up-front costs to the customer. But I am trusting that the superior performance of Akida will win over customers like Mercedes, and partners like Prophesee and SiFive, and marquee clients like NASA/DoD/DoE (DOE - that's right, we do do cybresecurity too).
We don’t do cyber secunity. We currently do nothing. We currently have our IP in zero products. This is a fact Dio. As evident by the lack of revenue. Why???
I’m not saying we’ve failed yet, I believe the tech is strong, but for some reason people aren’t buying it, and I’d like to see some detail as to why.
The SP is down now, but within a few (short?) months ...
Come on mate, we’ve been in a death spiral for 30 months!
Last edited:
Bravo
If ARM was an arm, BRN would be its biceps💪!
Do we have the capability to also be in the X280 cores mix possibly?
I'm going to go with "YES" on that one.
Last edited:
wilzy123
Founding Member
Well said my friend
Everybody keeps forgetting VALEO, Brainchip in my opinion is behind Scala3 which has already taken more than $1 billion euros in pre orders, and that was at the back end of last year, so maybe they’ve taken more pre orders?
Regardless, scala3 is hitting the market at the end of this year/very early 2024 in 2024 model vehicles.
Who here remembers the Valeo premier back in 2021? I certainly do. If you need a refresher, watch the video again.
Also, as @Diogenese can attest, automotive takes 3 years minimum, we (Brainchip) have a joint development with Valeo where we get paid on undisclosed milestones, that was announced back in June 2020.
I’ve posted this before, but on the CTO of Valeo interview post on LinkedIn about 2 months ago, I asked if they’ve achieved Scala3 by neuromorphic, this was the response I got from a Valeo employee below.
View attachment 42507
View attachment 42508
It is happening, believe it, we will get paid and that will be a recurring income and will only grow as Scala3 becomes standard for even more vehicles.
Of course my opinion but based on the facts out there, who else can supply Valeo with their neuromorphic IP?? There is nobody. We’ve been through this before!
Valeo (global leader in ADAS solutions) have been PUMPING Mercedes in all their marketing material. Scala3 is coming and Merc are no doubt a healthy portion of their huge pre-orders.
Mercedes also haven't lost their appetite for Neuromorphic, having advertised the position "Development engineer in the field of artificial intelligence with a focus on neuromorphic computing and ADAS (f/m/x)" (https://jobs.mercedes-benz.com/job/...on-neuromorphic-computing-and-adas-f-m-x.html) with these goals:
- Applied research in the field of neuromorphic computing and spiking neural networks
- Development of innovative algorithms for automotive applications
- Implementation of neuromorphic algorithms for automotive applications
- Researching and transferring current neuromorphic computing trends for automotive applications with a special focus on ADAS
- Conceptual implementation of ideas and development of demonstrators
- Close cooperation with colleagues from the project and the specialist departments
- Involvement in the development of new ideas and in the development of the strategic orientation of the project
- Close cooperation with universities, research institutions, technology companies and start-ups
- Supervision of internships, bachelor and master thesis and doctoral dissertations
Here's Valeo's view on the role of AI in their state of the art Scala 3 presented at the last CES.
gilti
Regular
try i week short of 35 months low.
everybody who has brought since 28/9/2020 is underwater on that purchase
this is total bollocks and ignores all progress for nearly 3 years
cosors
👀
"Neuromorphic Chips-Top Five Important Things You Need To Know
Neuromorphic Chips: Unveiling the Future of Cognitive Computing
In the realm of cutting-edge technology, the pursuit of mimicking the extraordinary capabilities of the human brain has led to the development of neuromorphic chips. These innovative chips represent a paradigm shift in the field of artificial intelligence and computing, striving to emulate the intricate neural networks and cognitive functions of the human brain. Unlike traditional digital processors that rely on binary logic and sequential operations, neuromorphic chips leverage a biologically inspired approach, harnessing the power of parallelism and adaptability to revolutionize various domains, including machine learning, robotics, and sensory processing. By emulating the brain’s efficient neural connections, these chips hold the potential to unlock unprecedented levels of energy efficiency, speed, and versatility, ultimately paving the way for a new era of cognitive computing.
Neuromorphic chips, at their core, embody the convergence of neuroscience and computer engineering. The term “neuromorphic” stems from the fusion of “neuro,” referring to neurons and neural networks, and “morph,” suggesting a form or structure. Essentially, these chips are engineered to replicate the neural architecture and behavior of the brain, aiming to bridge the gap between biological cognition and artificial intelligence. The human brain’s astounding capabilities, such as pattern recognition, sensory processing, and adaptive learning, have served as a wellspring of inspiration for scientists and engineers in their quest to create intelligent machines that can simulate human-like reasoning.
Unlike conventional processors that operate based on the binary on-off switches of transistors, neuromorphic chips employ a vastly different approach. They consist of networks of artificial neurons and synapses that communicate and process information in a manner akin to biological neural circuits. These neurons, often simplified abstractions of real neurons, communicate through analog signals, allowing for continuous variations in their output. This analog nature enables neuromorphic chips to process information with remarkable efficiency, as they can simultaneously process multiple inputs and perform complex computations in parallel. The result is a significant reduction in power consumption and a potential leap in computational speed, making them highly attractive for various applications where energy efficiency and real-time processing are crucial.
The development of neuromorphic chips has also rekindled interest in hardware acceleration for artificial intelligence tasks. While traditional computing paradigms, such as the von Neumann architecture, have propelled technological progress for decades, they are increasingly encountering limitations when faced with the demands of complex AI algorithms, particularly those related to deep learning. Neuromorphic chips, however, are tailor-made for these types of algorithms due to their neural network-inspired design. This has led to a growing recognition of their potential to usher in a new era of efficient and high-performance AI processing, catalyzing research and investment in this innovative field.
Intriguingly, one of the remarkable aspects of neuromorphic chips is their ability to learn and adapt to new information, similar to the plasticity of biological brains. This phenomenon, known as synaptic plasticity, allows the strength of connections between artificial neurons to be modified based on the patterns of input they receive. This intrinsic learning capability opens the door to unsupervised learning, where machines can autonomously extract meaningful patterns from data without explicit programming. Such self-learning characteristics are especially valuable in scenarios where traditional rule-based programming becomes impractical due to the complexity and variability of the data.
In essence, neuromorphic chips hold the potential to redefine the landscape of computing and artificial intelligence. Their emergence marks a departure from the rigid, rule-based processing of traditional digital systems, ushering in an era of adaptable, parallel, and energy-efficient computing inspired by the neural elegance of the human brain. As researchers delve deeper into the intricacies of neuromorphic design and capitalize on their promising characteristics, these chips could unlock a multitude of transformative applications, from brain-inspired robotics to intricate real-time sensory processing. While challenges remain, including refining the accuracy and scalability of these systems, the trajectory of neuromorphic chips underscores their potential to shape the future of cognitive computing and propel us into an era where machines not only compute but truly comprehend.
Parallel Processing:
Neuromorphic chips are designed to process information in a massively parallel manner, similar to how the human brain’s billions of neurons work together simultaneously. This parallelism allows for the efficient execution of complex computations, enabling tasks such as pattern recognition and data analysis to be performed with exceptional speed.
Analog Signal Processing:
Unlike traditional digital processors that use discrete on-off signals, neuromorphic chips operate with analog signals. This analog nature enables them to perform continuous computations and process real-world data with a high degree of precision, leading to improved accuracy in tasks that require subtle variations or gradient-based learning.
Adaptive Learning:
One of the defining features of neuromorphic chips is their ability to exhibit synaptic plasticity, akin to the brain’s ability to adapt and learn from experience. The connections between artificial neurons can be dynamically strengthened or weakened based on the input patterns they receive. This intrinsic learning capability allows for unsupervised learning and the extraction of meaningful insights from raw data.
Energy Efficiency:
Neuromorphic chips are designed with energy efficiency in mind. The parallel and analog processing nature reduces the need for excessive power consumption, making them particularly suited for applications where low energy consumption is crucial, such as mobile devices, edge computing, and IoT (Internet of Things) devices.
Real-time Processing:
The combination of parallelism, analog processing, and adaptive learning equips neuromorphic chips with the ability to process information in real-time. This makes them well-suited for tasks that require rapid decision-making and response, such as autonomous vehicles, robotics, and real-time sensor data analysis.
These key features collectively distinguish neuromorphic chips from traditional digital processors and underline their potential to revolutionize various fields, including artificial intelligence, machine learning, robotics, and beyond.
Neuromorphic chips stand as a testament to the relentless pursuit of understanding and emulating the remarkable computational capabilities of the human brain. Rooted in the principles of neuromorphic engineering, these chips embody a fusion of neuroscience, computer science, and engineering prowess. They represent a convergence of disciplines, aiming to bridge the gap between the complex workings of biological neural networks and the creation of artificial systems that can replicate, to some extent, the brain’s intricate functions.
The inception of neuromorphic chips emerged as a response to the limitations of traditional digital processors in tackling certain types of complex computations. While digital processors excel in performing sequential tasks with high precision, they encounter challenges when dealing with tasks that require parallelism, adaptability, and the kind of cognitive flexibility that human brains effortlessly exhibit. Neuromorphic chips arose as a visionary alternative, built upon the inspiration drawn from nature’s most sophisticated computational structure.
One of the captivating aspects of neuromorphic chips lies in their ability to unravel the mysteries of the brain’s information processing mechanisms. By attempting to mimic the neuronal connections and signal propagation observed in the brain, scientists and engineers dive deep into the intricate dance of ions and neurotransmitters that underlie human cognition. These chips provide a tangible platform to experiment with theories about learning, memory, and decision-making. This emulation, while far from a complete replication of the brain’s complexity, offers a unique playground to test hypotheses about neural behavior and explore novel methods of computation.
The journey of developing neuromorphic chips is fraught with challenges, yet every obstacle fuels the flames of innovation. Researchers delve into the fundamental physics of materials, seeking to replicate the behavior of synapses and neurons in silicon and other semiconductor substrates. The quest involves developing models that not only capture the essence of neural connections but also encapsulate the plasticity – the brain’s ability to strengthen or weaken connections based on experience. This endeavor has given rise to intricate algorithms, programming languages, and simulation tools specifically designed to harness the potential of neuromorphic hardware.
In the grand tapestry of technology, the story of neuromorphic chips intertwines with the ambitions of creating machines that possess cognitive capabilities. These chips invite us to ponder questions about the nature of intelligence, the boundaries between biology and technology, and the ethical implications of creating systems that mimic human-like learning and decision-making. The drive to replicate the brain’s marvels is accompanied by a profound curiosity – a desire to understand not only how the brain computes but also how it gives rise to consciousness, emotions, and the rich tapestry of human experience.
Neuromorphic chips, however, do not exist in isolation. Their development has sparked collaborations across disciplines, transcending the boundaries of computer science and neuroscience. Researchers from various domains join hands to unravel the mysteries of cognition, bringing together insights from psychology, neurobiology, physics, materials science, and more. This multidisciplinary effort mirrors the interconnectedness of the human brain itself, where diverse regions collaborate to create the symphony of thoughts and actions.
As neuromorphic chips evolve, they have the potential to alter the landscape of artificial intelligence, paving the way for more efficient and brain-inspired computing models. These chips offer promise in applications ranging from medical diagnostics, where they could aid in the interpretation of complex data such as brain scans, to environmental monitoring, where they could analyze vast datasets in real time. Their deployment in robotics might enable machines to perceive and interact with the world in a manner that feels more natural and intuitive.
In the larger context, the journey of neuromorphic chips underscores the remarkable feat of human ingenuity. It illustrates our capacity to draw inspiration from nature’s most intricate designs and transform them into tangible innovations that shape the world. These chips symbolize the unending pursuit of pushing boundaries, not merely for the sake of technology, but to uncover the secrets of existence itself.
In conclusion, the story of neuromorphic chips is a narrative of ambition, curiosity, and interdisciplinary collaboration. It is a testament to humanity’s quest to comprehend the enigmatic workings of the brain and translate that understanding into revolutionary technology. Neuromorphic chips stand at the crossroads of science and imagination, inspiring us to delve deeper into the realms of cognition, computation, and consciousness. As they continue to evolve, they hold the potential to reshape our relationship with technology and guide us towards a future where the boundaries between human and machine become increasingly blurred."
https://dotcommagazine.com/2023/08/neuromorphic-chips-top-five-important-things-you-need-to-know/ .)
A lot of shareholders that have been holding BRN shares for years and years have been patiently waiting for these next 9 months. These are the championship rounds. No time for the wobbles now. Can’t understand the efforts some are making to try and convince others that they should be worried.
JDelekto
Regular
I'll add my $0.02 to this excellent response.
Aside from algorithm development, Akida is used to its full advantage by sensors that provide event-based output. There are many different types of sensors, but those which require no conversion, like Prophesee's event-based vision solution, will provide the best integration.
The algorithms that will come into play will be applications that use Akida's one-shot learning (such as training for a new voice, image, vibration, gustatory or taste, etc.) and inferencing for specific tasks. NVISO's algorithms detecting the emotional state of passengers in a vehicle most likely use several models to detect eye position, mouth shape, etc.
Consider a toy that will learn the voice of the child that owns it and determine their mood based on their current facial expression, and provide seemingly empathic responses. This is where I think the algorithms come into play; some of these solutions would work with other inferencing chips, and the performance and minimal power requirements would make it ideal for toys and wearables while leaving them disconnected from the cloud.
From the SiFive X280 datasheet it is clear that their AI/ML does not include Akida. It would have been on the drawing board before we met.
On the other hand, SiFive have recognized Akida's capabilities.
https://brainchip.com/sifive-and-brainchip-partner-to-demo-ip-compatibility/#:~:text=SiFive and BrainChip have partnered to show their,IP working alongside SiFive’s RISC–V host processor IP.
April 20, 2022
SiFive and BrainChip have partnered to show their IP is compatible in SoC designs for embedded artificial intelligence (AI). The companies have demonstrated BrainChip’s neuromorphic processing unit (NPU) IP working alongside SiFive’s RISC–V host processor IP.
Brainchip’s NPU processor IP, the basis for its Akida chip, is a neuromorphic processor designed to accelerate spiking neural networks. This IP can be used to analyze inputs from most sensor types, including cameras, to provide ultra–low power analysis in real–time applications. A recent BrainChip demo showed its Akida chip in a vehicle, detecting the driver, recognizing the driver’s face, and identifying their voice simultaneously. Keyword spotting required 600 µW, facial recognition needed 22 mW, and the visual wake–word inference used to detect the driver was 6–8 mW.
https://brainchip.com/akida2-0/
Phil Dworsky, Global Head of Strategic Alliances, SiFive
"more complex applications including object detection, robotics, and more can take advantage of SiFive X280 Intelligence™ AI Dataflow Processors tightly integrated with BrainChip’s Akida-S or Akida-P neural processors"
... so it's there for the future.
Attachments
Last edited:
Azzluvstrades
Member
Space Cadet
Regular
Similar threads
