# Bittensor For The Long Run Bittensor is the future of synthetic intelligence and we're all in. For brief context: Bittensor is an open-source network and protocol founded to deliver on a simple mission: to use programmable incentives to accelerate development for open-source intelligence markets. We are a team of EU-funded researchers building Subnet 17 (SN 17) — an intelligence market on Bittensor dedicated towards generating 3D assets. Before continuing, this piece assumes a basic underlying understanding of Bittensor. An in-depth look at the underlying principles and development of the network is explained [here](https://medium.com/collab-currency/running-bittensor-9d0e810d2483).

404—GEN

In SN 17 we consider value creation along two (related) vectors: 1. AI model improvements (research value); and 2. Commercial traction (economic value). "If I have seen further it is by standing on the shoulders of Giants."\ \- Isaac Newton, Letter to Robert Hooke (1675) We — and Bittensor as a whole — are strong believers in building upon Open Source (OS) software. To date, OS software has suffered from a lack of proper incentive structures.

A Replication Study on Measuring the Growth of Open Source
M Dorner, M Capraro, A Barcomb, K Wnuk (2022)
arXiv: 2008.07753

By creating an incentive structure (via Bittensor’s $TAO) together with proper validation mechanisms, it is possible to now encourage competition to leverage (and drive improvements beyond) State Of The Art (SOTA) OS AI models. Our team at SN 17 use Bittensor’s incentive mechanism to address the near-infinite demand for 3D content that exists within gaming, and which extends far beyond into entertainment (film and VFX), as well as consumer and retail applications. With recent consumer hardware and end-user device advances, we believe this demand will multiply exponentially as AR, VR and XR products and services mainstream over the next 24 months. The end result is a bottleneck in which creatives who lack the necessary capital are unable to fulfill a global demand for AR, VR, XR, 3D experiences, social gaming, and more. SN 17 aims to relieve that bottleneck by initially targeting three specific use cases: 1. Synthetic dataset generation; 2. Creation of 3D asset marketplaces; 3. Organic consumer applications. Synthetic dataset generation is the creation of 3D models used as inputs for AI training, and a logical first step to move from research to economic value. SN 17 is currently capable of producing 100K+ verified 3D models every eight hours, and this power can be harnessed towards generation of real-time industry trends and content creator needs.

404—GEN will be the world's largest 3D dataset in 60 days

*Effectively, SN 17 has the ability to dwarf web2 market leaders like Unity, Kitbash and Sketchfab in an intelligent way by curating synthetic generation requests according to real-time demand vectors (e.g. latest industry search trends) and storing models onchain.*

Examples of 404—GEN 3D outputs

Ultimately, it is the broader consumer applications that are most exciting to our subnet – steady state, this system has the ability to accelerate the non-technical creation of 3D assets across all skill levels, matching the growing demand for immersive objects and experiences around the internet. Gaming is a market that already craves this type of innovation. User Generated Content (UGC), is heavily relied upon to inform persistent user retention and engagement. On SN 17, we aim to power 3D UGC on any platform by allowing for direct API calls to SN 17. To further these goals we have created a web front end, discord bot and Blender plugin that provide sample generative capabilities and open-source templates for the community to build upon for their own purposes.

Ways to interact with 404—GEN

We are also partnering with innovative gaming, XR, and consumer applications to support the integration of SN 17 into their virtual world creation. Ultimately, with SN 17, it will be possible for anyone to text prompt entire virtual worlds into existence. The improvement of foundational models and growth of open source repositories creates a uniquely diverse and fractured landscape for 3D Gen AI. This coupled with the fact that 3D Gen AI research is still nascent (compared to 2D) means that a ‘market winning’ technology has yet to be determined. Approaches such as 3D Diffusion, Neural Radiance Fields (NeRFs), and Gaussian Splattering (Splatts) all compete with different underlying neural network architectures. Research into all three of these approaches is rapidly developing and benchmarking solutions change nearly every week. Currently, closed source platforms have an advantage. But Bittensor offers exactly the type of open competitive landscape that we believe can ultimately overcome that gap. Specifically, we believe a diversity of models are required to cover the problem constraints of generating 3D assets in unique and novel ways. Simply put, SN 17 incentivizes: 1. Choosing the right model for the right job; and 2. Rewarding the best miners building atop SOTA 3D Gen AI. We've already witnessed the latter point applied to SN 17 – today, top miners on SN 17 are those who seek out new models and modify them. The principle was explained well by one SN 17 miner who was asked how they were constantly earning high rewards:

SN 17 miner explanation of incentive-based performance.
In this sense, SN 17 is a dynamic and real-time representation of SOTA 3D Gen AI.

A visual example of model performance is evident by evaluation of how the results of SN 17 have changed in the 2 months that we have been live on mainnet. Below are results from May and July of 2024. Showing improvements to both the underlying models as well as the validation mechanism which scores results.

3D generated models on SN 17 May 2024 (Left) and July 2024 (Right)

This is just the beginning. We are at the cusp of exponential growth within the broader 3D Gen AI landscape. One such indicator for future quality of 3D is to consider the historic trend we have seen in the 2D space. Shown below is the evolution in quality between Midjourney v1 and Midjourney v5.1 (over a 15-month period). The cryptonomic commons.

MidJourney 2D Improvements from Jan 2022 (v.1) to May 2023 (v.5.1)

While furthering 3D AI research is key towards future value creation, there is also a broader economic rationale behind Bittensor. Creating incentives to improve AI models creates development trajectories towards monetization. The underlying technology of SN 17 has already reached levels of quality suitable for paying enterprises today. This is because the creation of virtual environments rely on the highly specialized and time consuming process of manually modeling, sculpting or procedurally scripting 3D digital assets. The costs in both financial and human capital are highly restrictive. Further, computing power is growing at an exponential rate and with it, consumers’ expectations regarding the size, density and visual fidelity of virtual worlds. **Meet Bittensor** Perhaps such levels of innovation could be facilitated by a multibillion dollar entity such as an OpenAI or Google, but then ultimately every user created worlds would be subject to risks inherent in partnership with a central authority and creatively limited by any components deemed proprietary, closed source or competitive. As a team building a subnet on Bittensor, what is truly unique about SN 17 is not only the value being created, but the way in which it is created – via decentralized systems combined with collective intelligence and ownership. We’re all in. # 404—GEN Whitepaper Drafted 29.02.2024 **Abstract** 404 provides a platform to democratize 3D content creation, ultimately allowing anyone to create virtual worlds, games and AR/VR/XR experiences via SN17 on Bittensor. 404 leverages the existing fragmented and diverse landscape of open-source 3D content generation models - ranging from Gaussian Splatting, Neural Radiance Fields, and 3D Diffusion Models - to facilitate innovation. This is an ideal landscape in which to construct decentralized incentive-based networks via Bittensor. We aim to kickstart the next revolution in gaming around AI native games. The interconnectivity of Bittensor subnets can facilitate experiences in which assets, voice and sound are all generated at runtime. This would effectively allow a creative individual without any coding or game-dev experience to simply describe the game they want to create and have it manifested before them in real time. **Introduction** The creation of virtual environments relies on individuals manually modeling, sculpting and/or procedurally scripting 3D digital assets. Even for those with professional training in these fields, this process is time consuming, inflexible and expensive. For this reason, traditional use cases such as gaming have exceptionally high barriers to entry, take years to build, cost hundreds of thousands of dollars to create and require technical skill sets beyond creativity. This problem is getting worse. Computing power is growing at an exponential rate and so is the expectation of digital consumers regarding the size, density and visual fidelity of virtual worlds. The demand extends far beyond gaming, into entertainment more broadly (film and VFX), as well as consumer and retail applications. Recent consumer hardware advances means this demand will exponentially multiply as AR, VR and XR manifestations become mainstream and are demanded for existing (and new) applications within months and years. The barriers to entry must come down so that we can democratize 3D content creation. This is not only the viable solution to meeting the exponentially growing demand for such services, it is also the way to ensure competition and create innovation in these markets. By allowing non-professional creators to build 3D assets and virtual worlds based on text-prompts, it forces creativity as the differentiating factor rather than technical ability or financial / incumbent positions. Ultimately, AI allows the creation experience to be made more democratic, efficient and even automated, but the creator should always be able to intervene and express directorial control over any decision. This shifting economic and societal landscape comes at a time when 3D AI is poised to explode due to technology advances of the last 12 months. For precedent, we can look towards the last three years of unbelievable improvements in State of the Art (SOTA) 2D AI models thanks to the introduction of Transformer networks and the subsequent creation of foundational models. 3D AI technology has gone through similar research innovations and is now at the stage in which foundational models can be built. **3D AI technology and miner considerations** A variety of technologies have been developed to tackle the issue of 3D content creation. These techniques allow users to input a text prompt(s), image(s), or a combination of both. This means that users who have no experience with 3D modeling or other traditional forms of 3D content creation can become creators. These models are typically trained to extract multiple synthetic views of the desired object or scene and then attempt to reconstruct a mesh, radiance field or splatt representation from the multiple synthetic views. The 3D space is still so nascent that a ‘market winning’ technology has yet to be determined, with approaches such as 3D Diffusion, Neural Radiance Fields (NeRFs), and Gaussian Splattering (Splatts) all competing with different underlying neural network architectures. Research into all three of these approaches is rapidly developing and benchmarking solutions changes nearly every week. In a landscape such as this - with competing, rapidly developing and open source models - Bittensor offers an ideal decentralized incentive-based platform for empowering innovation. It is important to note that the current SOTA 3D AI networks are not yet at a level that rivals professional 3D content creation, but the speed at which these networks are developing suggest a trajectory in which they will do so in 2024. Further, despite aesthetic limitations (compared to professional 3D Artists), the insatiable demand for 3D content means there are already viable applications today which include 1) background / non-hero assets, 2) environments for virtual worlds, 3) abstract and/or highly stylised assets. To help understand the technical differences between these networks a brief overview of the three key technologies powering this revolution is provided below: *Neural Radiance Fields (NeRFs)* Neural Radiance Fields (Mildenhall et al, 2020, Gao et al., 2023) is a machine learning based approach that synthesizes novel views of a complex scene and allows the user to generate its 3D representation. The input scene is defined as a set of images with known camera poses. The fully-connected deep neural network defines the input scene as a continuous 5D vector-field function (function components: (x, y, z) - spatial location, (theta, phi) - viewing direction). This continuous 5D function is approximated with a multilayer perceptron (MLP) network. The output of this network is a volume density and a view-dependent emitted radiance at the spatial location. NeRFs are quite heavy in terms of computation as to compute the radiance emittance it is essential to approximate the volume rendering integral: $$C(\bold{r}) = \int\_{t\_n}^{t\_f} T(t)\sigma (\bold{r}(t)\bold{c}(\bold{r}(t), \bold{d}) dt$$ where $$T(t)=exp\Big(- \int\_{t\_n}^{t} \sigma (\bold{r}(s))ds\Big)$$ Where $$\bold{r}(t)=\bold{o} +t\bold{d}$$ is a camera ray with near $$t\_n$$and far $$t\_f$$, $$\sigma(x)$$ is a differential probability of a ray terminating at an infinitesimal particle at location x and the function T(t) denotes the accumulated transmittance along the ray from $$t\_n$$ to t. In addition, NeRFs are quite heavy on memory consumption. Estimated surfaces are usually not clearly defined using volume density fields and extracting surfaces from the density-based representation often leads to noisy results. Relevant to making this technology more viable within the Subnet, we researched quite a few methods proposed regarding how to overcome computations costs. One of them was “Instant Neural Graphics Primitives with a Multiresolution Hash Encoding” (Muller et al. 2022). Authors of this method suggested two techniques to speed up the computations, namely radiance caching and hash table encoding. These significantly speed up the computation process from hours to several minutes on a single GPU. Two additional influential works worth mentioning are MIP-NeRF (Barron et al. 2021) and Mip-NeRF 360 (Barron et al, 2022). The latest work extends Mip-NeRF to unbounded scenes and introduces several important techniques: namely, prediction of the appropriate sampling intervals for volumetric density, a novel scene parametrization constructed for the Gaussians in MIP-NeRF and a new regularization method that helps to prevent floater geometric artifacts and background collapse. A last remarkable advancement in the NeRF field worth noting was described in paper “Neurangelo: High-Fidelity Neural Surface Reconstruction” (Li et al., 2023), in which the authors managed to achieve significant quality improvement of the extracting surfaces. Relevant to applications addressed in this paper, NeRFs can be editable. For instance, Edit-NeRF(Liu et al., 2021) provides NeRF editability via image conditioning from user input.

Sample results from video [frames]
Credit: “Neurangelo: High-Fidelity Neural Surface Reconstruction” (Li et al., 2023)

*Gaussian Splatting* Gaussian Splatting (Kerbl et al., 2023, Chen et al., 2024) is another method that aims to efficiently optimize a scene representation to achieve high fidelity results for a novel view synthesis. This method relies on 3D Gaussians for defining the scene as it is a differentiable volumetric representation which is unstructured and explicit to allow very fast rendering. 3D Gaussians can be easily projected to 2D splats that are important for computing fast alpha blending for rendering. Gaussians are defined as a covariance matrix that describes the configuration of an ellipsoid. In this case, covariance matrix can be defined as: $$\Sigma=RSS^TR^T$$ Where S is a scaling matrix and R is a rotation matrix. Using the covariance matrix defined in this form it is possible to optimize it using stochastic gradient-descent. To accurately capture the view-dependent appearance of the scene, spherical harmonics coefficients of each 3D Gaussian representing the color should be also optimized. This step is interleaved with steps that control the density of the Gaussians. According to (Chen et al., 2024) Gaussian Splatting representation can be defined as: $$L\_{3DGS}(x, y, z, \theta, \phi)=\sum\_i G(x,y,z,\bold{\mu}\_i,\bold{\Sigma}\_i)\cdot c\_i(\theta,\phi)$$ Where G() is the Gaussian function, x, y, z - are the coordinates of the point, $$\bold{\mu}\_i$$ is the mean, $$\bold{\Sigma}\_i$$ is a covariance matrix, c is the view-dependent color.

Sample results (immersive 3D world in VR) from text prompted OpenAI Sora video
Credit: X @dankvr

*3D Diffusion* Diffusion models are a class of latent variable generative models. The diffusion model is composed of three steps: some forward process, some reverse process and the sampling procedure. The goal of the diffusion model is to learn a diffusion process used for generating a probability distribution of a given dataset. Such models were one of the first to be used within Text-to-3D generative pipelines. 3D generative models can be trained on explicit representations of structure (e.g. voxels or point clouds). The generative model is trained to slowly add structure to the initial random noise with some predefined transition (Poole et al., 2022). Combinations of these techniques are also emerging. NeRFs (Poole et al., 2022; Lin et al., 2023; Melas-Kyriazi et al., 2023) and Gaussian Splatting (Liang et a., 2023; Tang et al., 2023; Wang et al., 2023) approaches have become critical building blocks for inverse rendering and generation of novel views within generative systems. In Poole, authors proposed a DreamFusion model that combines a 3D diffusion model with score distillation sampling method with a NeRF model for generating 3D objects. This method has two notable drawbacks: extremely slow NeRF optimization and low-resolution image space supervision on NeRF. In another work (Lin et al. 2023) researchers have proposed another 3D generative model, namely Magic3D, that outperforms DreamFusion and provides a solution for two mentioned drawbacks. Their approach has two major stages. First a coarse model is obtained using a low resolution diffusion model the computation of which is accelerated using 3D hash grid structure. After it a coarse representation is further optimized as a textured 3D model using an efficient differentiable renderer that interacts with a high-resolution latent diffusion model. Further, the Gaussian Splatting method was used for multi-view reconstruction within the DreamGaussian model (Tang et al., 2023) instead of NeRF. This significantly speeds up the generation process to several minutes. The introduced pipeline allows the user to generate a 3D model from text prompt or from a single image. Authors also addressed the problem of accurate mesh extraction from the Gaussian Splats and proposed a technique for sharpened texture extraction. The LucidDreamer model (Liang et al., 2023) is another model that incorporated Gaussian Splatting within the text-to-3D generation pipeline. To solve this problem of over smoothing caused by the Score Distillation Sampling technique, the authors proposed a novel sampling method called Interval Score Matching. This method allows for deterministic diffusing trajectories and uses interval-based score matching to counteract over-smoothing. Overall, our tests and research suggest that Gaussian Splattering is computationally more efficient and aesthetically higher quality than the other techniques. Therefore, although the subnet’s validation mechanism is able to accurately evaluate results generated from a diversity of neural networks (as long as the output result is the correct format of generated 3D representation), the sample miner code is set up for Splatts. **Subnet Evaluation / Validation**\ \ While such a nascent and diverse technology landscape is ideal for builders and miners seeking innovation within the space. It makes the subnet’s evaluation system for open source models (the validation mechanism) particularly challenging. We have sought to tie key evaluation criteria to those factors which will have meaningful implications for both builders (visual fidelity, editability) and miners (training & inferencing time, memory efficiency), understanding that the distinction is not mutually exclusive. A persistent challenge faced by subnets operating within AI/ML networks is the evaluation of miners' output when mining results are inherently indeterministic. Traditional scoring mechanisms struggle in these environments due to the absence of a single, definitive answer - introducing a degree of subjectivity that could be exploited, for instance through prompt engineering or targeted model fine-tuning. This predicament is amplified when requests are issued both by subnet users and validators, amplifying the uncertainty surrounding the "correct" response. One conventional approach to establishing a benchmark for evaluation is the construction of a curated dataset consisting of predetermined question-answer pairs. Validators then use this dataset to gauge the quality of answers provided by miners. However, this method not only generates additional work with dubious added value but also carries the risk of 'dataset leakage', where miners might access the "correct" answers in advance, leading to manipulation (superficial score improvements) of the evaluation process. Our solution circumvents these issues by leveraging image references to anchor the evaluation process and minimize subjectivity. These references can either accompany the input directly or be synthesized by image/multimodal subnets within the Bittensor ecosystem. Although image generation introduces some level of subjectivity, advancements in image generation models mitigate this concern significantly. By incorporating image references, our methodology fosters an objective and fair framework for scoring miner performance without the need for a static dataset of known answers, thereby preserving the integrity and reliability of the network. Our initial setup is oriented around onboarding miners and validators with a clear goal to generate 3D synthetic datasets that are organized into game type and style categories so that results can be used as asset packs for immediate applications (as detailed later in this whitepaper). From this foundation, we will continually expand the diversity and quality of the datasets, to facilitate new creative applications and be more representative of user needs in every new iteration (while still recording data and model provenance).

Initial proposed setup - February 2024

While this initial setup builds asset packs that over time will rival and ultimately dwarf even the largest online 3D asset stores (sketchfab, Unity marketplace, etc) via synthetic traffic. This is ultimately a bridge to our longer term system which becomes increasingly bespoke, organic, and user-oriented as game developers, creators and other builders begin directly interacting with the subnet through web2 front end interfaces and partnerships.

Extended proposed setup - February 2024

**Applications** We expect a variety of diverse real-world applications to emerge built on top of the 3D AI technology innovations and development facilitated and empowered by this subnet. In this section, we outline some tangible applications that we will actively facilitate through partnerships, collaborations and front end interfaces. These applications focus within the Gaming vertical as our team has strong domain knowledge in this industry and sees strong web3 and web2 opportunities arising here. One could easily expand these applications into related entertainment verticals, especially given the entertainment market as a whole is moving towards gamification - exemplified by the February 2024 Disney x Epic partnership announcement. Attention is becoming an increasingly scarce resource and passive consumption is being replaced by active engagement in the same way that finalized products in the entertainment industry are being replaced by continuous creation. > “We compete with (and lose to) Fortnite more than HBO” (Netflix letter to shareholder) *Immediate / Short Term: Single vertical game asset creation* We anticipate that some of the first widely used use cases will start as modding and vertical creation around a specific game genre and/or art style. This directly ties to our initiative to build out datasets in alignment with common gaming genres and aesthetics. As a point of reference one can consider the more narrow focus of Unreal Editor for Fortnight rather than the broader Unreal Engine. *Medium Term: Broader world building platform* Our short-term applications including the datasets created by miners & validators can act as a template for a larger flywheel: building games as drivers of constant engagement within expanding platforms in which interconnecting new games furthers the network effects of the ecosystem. In particular, when enabled by AI and Bittensor this should be possible in a fraction of the time and cost compared to game development of the last decade. And if historical precedent is true, then the first games produced using our subnet do not need to be a success, rather they simply need to foster growth on the platform and kickstart a creator flywheel until a viral hit is achieved (e.g. Fortnight Battle Royale). *Long Term: Foundational creation engine(s) / OS for entertainment verticals* As viral games and experiences are created using the subnet, this will create more interest and incentive to build atop the subnet and therefore increase in the number of viral games and experiences produced. Such traction will lead to network effects that ultimately facilitate breakthroughs and expansion beyond single games and even platforms. 3D AI combined with decentralization as envisioned by 3D Gen Subnet will enable a world in which an IP created can be iterated upon in multiple AI native forms. Other creators can mod games built using the subnet and/or builders can facilitate this by enabling UGC relying on the same Bittensor subnet. Furthermore, complementary features like NPCs and/or AI-enabled dialogue would naturally leverage other subnets given the interconnected potentials of the BIttensor system. In combination, this would kickstart the next revolution in gaming around AI native games in which assets, voice and sound are all generated at runtime. Effectively allowing a creative individual without any coding or game-dev experience to simply describe the game they want to create and have it manifested before them in real time. Such applications eventually extend beyond our subnet and by their nature will leverage the broader Bittensor ecosystem. Given 3D experiences are an amalgamation of art, sound, animation, code, physics and more, a fully AI-native engine will not call underlying generative models in isolation (e.g. those provided by 3D Gen Subnet) but rather will be multi-modal, leveraging semantic scene or world understanding as input. This is a unique power of decentralized AI. This is just the beginning. The applications above focus on a single vertical (gaming), but a larger societal shift towards immersive environments suggests that there are numerous additional applications beyond gaming into VFX, retail, fashion design, architecture / urban planning, AR/VR/XR experiences, and so many more of which we are not yet aware. This hints at the true value of such a decentralized system - it incentivizes the community to build a broad range of initiatives and applications that discover and capture the highest value-added activities of 3D AI based upon our subnet. **Conclusion** We have proposed a system to democratize 3D content creation, thereby lowering the barrier to entry for the entertainment and media industries, and ultimately empowering any individual to create games, virtual worlds and AR/VR/XR experiences. This system is intrinsically reliant upon the decentralized intelligence of miners incentivised to build upon fragmented and rapidly evolving SOTA 3D AI models. It systematically enables validation of quality and aesthetics through the combination of synthetic dataset building (e.g. asset packs) combined with organic, user-oriented requests. Finally, it fundamentally will excel through tapping into the larger interconnected nature of combining various AI subnets to seed virtual worlds and experiences with potential to kickstart the next revolution in gaming. **About 404** 404 is a team of EU-grant funded AI researchers, gaming industry veterans, and blockchain-native builders. For over three years, we have developed proprietary 3D generative AI solutions for AAA web2 game developers and entertainment industry leaders. We believe this subnet offers a unique opportunity to leverage our industry networks, technology and expertise to onboard web2 industry players into web3. [404.xyz](https://404.xyz) **References** 1. J. T. Barron, B. Mildenhall, M. Tancik, P. Hedman, R. Martin-Brualla, and P. P. Srinivasan, “Mip-nerf: A multiscale representation for antialiasing neural radiance fields,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 5855–5864 2. J. T. Barron, B. Mildenhall, D. Verbin, P. P. Srinivasan, and P. Hedman, “Mip-nerf 360: Unbounded anti-aliased neural radiance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 5470–5479. 3. G. Chen, W. Wang, “A Survey on 3D Gaussian Splatting”, in IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024 4. B. Kerbl, G. Kopanas, T. Leimkuhler, G. Drettakis, “3D Gaussian Splatting for Real-Time Radiance Field Rendering”, in ACM Transactions on Graphics, vol. 42, no. 4, 2023. 5. Y. Liang, X. Yang, J. Lin, H. Li, X. Xu, and Y. Chen, “Luciddreamer: Towards high-fidelity text-to-3d generation via interval score matching,” arXiv preprint arXiv:2311.11284, 2023. 6. Z. Li, T. Muller, A. Evans, R. H. Taylor, M. Unberath, M.-Y. Liu, C.-H. Lin, “Neuralangelo: High-Fidelity Neural Surface Reconstruction”, in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023 7. C.-H. Lin, J. Gao, L. Tang, T. Takikawa, X. Zeng, X. Huang, K. Kreis, S. Fidler, M.-Y. Liu, and T.-Y. Lin, “Magic3d: High-resolution text-to-3d content creation,” arXiv preprint arXiv:2211.10440, 2022. 8. S. Liu, X. Zhang, Z. Zhang, R. Zhang, J.-Y. Zhu, and B. Russell, “Editing conditional radiance fields,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 5773–5783. 9. L. Melas-Kyriazi, C. Rupprecht, I. Laina, and A. Vedaldi, “Realfusion: 360° reconstruction of any object from a single image,” arXiv e-prints, pp. arXiv–2302, 2023. 10. B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” in European conference on computer vision. Springer, 2020, pp. 405–421. 11. T. M¨uller, A. Evans, C. Schied, and A. Keller, “Instant neural graphics primitives with a multiresolution hash encoding,” ACM Trans. Graph., vol. 41, no. 4, pp. 102:1–102:15, Jul. 2022. \[Online]. Available: 12. B. Poole, A. Jain, J. T. Barron, and B. Mildenhall, “Dreamfusion: Text-to-3d using 2d diffusion,” arXiv preprint arXiv:2209.14988, 2022. 13. J. Tang, J. Ren, H. Zhou, Z. Liu, and G. Zeng, “Dreamgaussian: Generative gaussian splatting for efficient 3d content creation,” arXiv preprint arXiv:2309.16653, 2023. 14. X. Li, H. Wang, and K.-K. Tseng, “Gaussiandiffusion: 3d gaussian splatting for denoising diffusion probabilistic models with structured noise,” arXiv preprint arXiv:2311.11221, 2023. 15. Disney and Epic Games to Create Expansive and Open Games and Entertainment Universe Connected to Fortnite. Press Release: 16. Netflix Q4 2018 Letter to Shareholders. Source: 17. Troy Kirwin and Jonathan Lai. Unbundling the Game Engine: The Rise of Next Generation 3D Creation Engines. 2024. # Terms of Service Effective as of 30 August 2024 Thank you for using the 404-GEN platform. These Terms of Service (“Terms”) govern the relationship between 404-GEN AG. (“Company,” “we,” “us,” or “our”) and our affiliates and the entity or person (“Customer, “You”, “Your”) using or accessing our services, applications, or platform through our website available at 404.xyz (the “Site”), through any of our Discord servers, or by any other means (together, the “Services”). These Terms explain what rights you have with respect to images, 3D models and other assets which you might generate with the Service (the "Assets"), or prompts you might enter into the Service (the “Inputs”), your use of the Services, and other important topics like arbitration. Our privacy policy outlines how we handle your data [here](/404-privacy-policy). Please carefully read these Terms, along with our privacy policy, and all other documents referenced in these Terms, including the Subscription Plans page and the Community Guidelines below. Together with the Terms, these documents form a single binding agreement between us (the “Agreement”). This Agreement is effective when the Customer is presented with this Agreement and proceeds to use or access the Services (the "Effective Date") or to receive or distribute Assets. This agreement may be updated and presented again to the Customer from time to time. Continued use of the Services constitutes acceptance of the updated terms. If You do not agree to this Agreement, please stop using or accessing the Services. (Service availability and quality) We are constantly improving the Services to make them better. The Services are subject to modification and change, including but not limited to the art style of Assets, the algorithms used to generate the Assets, and features available to the Customer. No guarantees are made with respect to the Services’ quality, stability, uptime or reliability. Please do not create any dependencies on any attributes of the Services or the Assets. We will not be liable to You or Your downstream customers for any harm caused by Your dependency on the Service. Both the Services and the Assets are provided to Customer on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Assets and assume any risks associated with use of the Services. 404-GEN reserves the right to suspend or ban Your access to the Services at any time, and for any reason. You may not access or use the Services for purposes of developing or offering competitive products or services. You may not reverse engineer the Services or the Assets. You may not use automated tools to access, interact with, or generate Assets through the Services. You may not resell or redistribute the Services or access to the Service. Only one user may use the Services per registered account. Each user of the Services may only have one account. You may not use the Service to try to violate the intellectual property rights of others, including copyright, patent, or trademark rights. Doing so may subject you to penalties including legal action or a permanent ban from the Service. We reserve the right to investigate complaints or reported violations of this Agreement and to take any action we deem appropriate including but not limited to reporting any suspected unlawful activity to law enforcement officials, regulators, or other third parties and disclosing any information necessary or appropriate to such persons or entities relating to user profiles, e-mail addresses, usage history, posted materials, IP addresses and traffic information. (Age requirements) By accessing the Services, You confirm that You are at least 13 years old and meet the minimum age of digital consent in Your country. If You are old enough to access the Services in Your country, but not old enough to have authority to consent to our terms, Your parent or guardian must agree to our terms on Your behalf. Please ask Your parent or guardian to read these terms with You. If You are a parent or legal guardian, and You allow Your teenager to use the Services, then these terms also apply to You and You are responsible for Your teenager’s activity on the Services. 404-GEN tries to make its Services PG-13 and family friendly, but the Assets are generated by an artificial intelligence system based on user queries. This is new technology and it does not always work as expected. No guarantees are made as to the suitability of the Assets for the Customer. (Your information) By using the Services, You may provide 404-GEN with personal information like Your email address, user name, billing information, favorites, image outputs, and text prompts that You enter, or sample images that You upload to the Service. Our privacy policy can be found [here](https://doc.404.xyz/404-privacy-policy/). (Content rights) (Your rights and obligations) You own all Assets You create with the Services to the fullest extent possible under applicable law. There are some exceptions: * Your ownership is subject to any obligations imposed by this Agreement and the rights of any third-parties. * If you upscale the Assets of others, these assets remain owned by the original creators. Please consult Your own lawyer if You want more information about the state of current intellectual property law in Your jurisdiction. Your ownership of the Assets you created persists even if in subsequent months You downgrade or cancel Your membership. Inputs, Assets, and other content such as messages, photos, videos, and documents that you may provide to the Services (such as through uploading, posting, sharing, or chat messages) are collectively, “Content”. You are responsible for all Content that you provide or generate, including ensuring that it does not violate any applicable laws or this Agreement, and that you have all necessary rights and permissions to provide the Content. (Rights you give to 404-GEN) By using the Services, You grant to 404-GEN, its affiliates, successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare derivative works of, publicly display, publicly perform, sublicense, and distribute the Content You input into the Services, as well as any Assets produced by You through the Service. This license survives termination of this Agreement by any party, for any reason. (DMCA and takedowns policy) *Notification Procedures* We respect the intellectual property rights of others. If you believe that material located on or linked to by the Services violates your copyright or trademark, please send a notice of claimed infringement to <404@404.xyz> with the subject “Takedown Request,” and include the following: 1. Your physical or electronic signature. 2. Identification of the copyrighted work (or mark) you believe to have been infringed or, if the claim involves multiple works, a representative list of such works. 3. Identification of the material you believe to be infringing in a sufficiently precise and detailed manner to allow us to locate that material. 4. Adequate information by which we can contact you (including your name, postal address, telephone number, and, if available, email address). 5. A statement that you have a good faith belief that use of the copyrighted material is not authorized by the copyright owner, its agent, or the law. 6. A statement that the information in the written notice is accurate. 7. A statement, under penalty of perjury, that you are authorized to act on behalf of the copyright owner. 8. If the copyright owner’s rights arise under the laws of a country other than the United States, please identify the country. Upon receipt of a notice that complies with the foregoing, we reserve the right to remove or disable access to the accused material or disable any links to the material; notify the party accused of infringement that we have removed or disabled access to the identified material; and terminate access to and use of the Services for any user who engages in repeated acts of infringement. Please be aware that if you knowingly misrepresent that material or activity on the Services is infringing your copyright, you may be held liable for damages (including costs and attorneys’ fees) under Section 512(f) of the DMCA. *Counter-Notification Procedures* If you believe that material was removed or access to it was disabled by mistake or misidentification, you may file a counter-notification with us by submitting a written notification to our copyright agent designated above. Such notification must include substantially the following: 1. Your physical or electronic signature. 2. An identification of the material that has been removed or to which access has been disabled and the location at which the material appeared before it was removed or access disabled. 3. Adequate information by which we can contact you (including your name, postal address, telephone number, and, if available, email address). 4. A statement under penalty of perjury by you that you have a good faith belief that the material identified above was removed or disabled as a result of a mistake or misidentification of the material to be removed or disabled. 5. A statement that you will consent to the jurisdiction of the Federal District Court for the judicial district in which your address is located (or if you reside outside the United States for any judicial district in which the Services may be found) and that you will accept service from the person (or an agent of that person) who provided us with the complaint at issue. 6. Our designated agent to receive counter notices is the same as the agent shown above. 7. The DMCA allows us to restore the removed content within 10-14 business days unless the complaining party initiates a court action against you during that time period and notifies us of the same. 8. Please be aware that if you knowingly materially misrepresent that material or activity on the Services was removed or disabled by mistake or misidentification, you may be held liable for damages (including costs and attorney’s; fees) under Section 512(f) of the DMCA. (Dispute resolution and governing law) In the event a dispute, controversy, or claim arises out of or relating to these Terms (“Dispute”), the Dispute will be resolved by binding arbitration rather than in court. The parties will first try in good faith to settle any Dispute within 30 days after the Dispute arises. If the Dispute is not resolved within 30 days, it shall be resolved by binding arbitration by the American Arbitration Association’s International Centre for Dispute Resolution in accordance with its Expedited Commercial Rules in force as of the date of this Agreement ("Rules"). The parties will mutually select one arbitrator. The arbitration will be conducted in English in Zug, Switzerland. By agreeing to mandatory arbitration as set forth herein, You and 404-GEN knowingly and irrevocably waive any right to trial by jury in any action, proceeding, or counterclaim, except that either party may apply to any competent court for injunctive relief necessary to protect its rights pending resolution of the arbitration. The arbitrator may order equitable or injunctive relief consistent with the remedies and limitations in the Agreement. The arbitral award will be final and binding on the parties and its execution may be presented in any competent court, including any court with jurisdiction over either party or any of its property. Each party will bear its own lawyers’ and experts’ fees and expenses, regardless of the arbitrator’s final decision regarding the Dispute. (Rate limiting) We reserve the right to rate limit You to prevent quality decay or interruptions to other customers. (Community guidelines) 1. Be kind and respect each other and staff. Do not create Assets, use text prompts, or submit content that are inherently disrespectful, aggressive, hateful, or otherwise abusive. Violence or harassment of any kind will not be tolerated. 2. No adult content or gore. Please avoid making visually shocking or disturbing content. We will block some text inputs automatically. 3. Respect others’ creations. Do not distribute or publicly repost the creations of others without their permission. 4. You may not use the Services to generate images for political campaigns, or to try to influence the outcome of an election. 5. You may not use the Services or the Assets to attempt to or to actually deceive or defraud anyone. 6. You may not use the Services for illegal activity nor may you upload images to our servers that involve illegal activity, or where the uploading itself may be illegal. 7. You may not intentionally mislead recipients of the Assets about their nature or source. 8. Respect others’ rights. Do not upload others’ private information. 9. Be careful about sharing. It’s OK to share Your creations outside of the 404-GEN community but please consider how others might view Your content. 10. Banhammer. Any violations of these rules may lead to bans from our services. We are not a democracy. Behave respectfully or lose Your rights to use the Service. (Limitation of liability and indemnity) We provide the service as is, and we make no promises or guarantees about it. You understand and agree that we will not be liable to You or any third party for any loss of profits, use, goodwill, or data, or for any incidental, indirect, special, consequential or exemplary damages, however they arise. Our aggregate liability under this Agreement will not exceed the amount You paid for the Services that gave rise to the claim during the 12 months before the claim. You are responsible for Your use of the service. If You harm someone else or get into a dispute with someone else, we will not be involved. To the extent permitted by law, you will indemnify and hold us harmless, our affiliates, and our personnel, from and against any costs, losses, liabilities, and expenses (including attorneys’ fees)\ from third party claims arising out of or relating to your use of the Services and Assets or any violation of these Terms. (Miscellaneous) 1. Force Majeure. Neither party will be liable for failure or delay in performance to the extent caused by circumstances beyond its reasonable control, including acts of God, natural disasters, terrorism, riots, or war. 2. No Agency. This Agreement does not create any agency, partnership, or joint venture between the parties. 3. Severability. If any part of this Agreement is invalid, illegal, or unenforceable, the rest of the Agreement will remain in effect. 4. No Third-Party Beneficiaries. This Agreement does not confer any benefits on any third party unless it expressly states that it does. 5. Survival. The sections and obligations in this Agreement that a reasonable person would expect to survive this agreement, will. Particularly the IP and privacy stuff. 6. Governing Law. This Agreement shall be governed by the laws of the Zug, Switzerland, without reference to conflict of law rules. All disputes will be governed by the arbitration agreement above. (How to contact us) * Email: <404@404.xyz> # Privacy Policy Effective as of 30 August 2024 (Introduction) This privacy policy (the “Policy”) describes 404-GEN AG (“404-GEN”)’s practices with respect to personal data that is collected when you access 404.xyz (the “Site”), or access or use the 404-GEN services, applications, or platform through the Site, Discord servers administered by 404-GEN, or by any other means (collectively, the “Services”) . 404-GEN is a communications technology incubator that provides 3D generation services to augment human creativity and foster social connection. As used in this Policy, “personal data” means any information that relates to, describes, could be used to identify an individual, directly or indirectly. **Applicability:** This Policy applies to personal data that 404-GEN collects, uses, and discloses and which may include: (i) data collected through the Services, (ii) data collected through the process of training 404-GEN machine learning algorithms, (iii) data collected through 404-GEN websites, and (iv) data collected from third party sources. Third party sources may include, but not be limited to: public databases, commercial data sources, and the public internet. This Policy does not apply to the following information: * Personal Data about 404-GEN employees and candidates, and certain contractors and agents acting in similar roles. **Changes:** We may update this Policy from time-to-time to reflect changes in legal, regulatory, operational requirements, our practices, and other factors. Please check this Policy periodically for updates. If any of the changes are unacceptable to you, you should cease interacting with us. When required under applicable law, we will notify you of any changes to this Policy. **Definitions:** Through this Policy, You, or Your means the individual accessing or using the Service, or the company, or other legal entity on behalf of which such individual is accessing or using the Service, as applicable. Company (referred to as either "the Company", "We", "Us" or "Our" in this Agreement) refers to 404-GEN AG. Usage Data refers to data collected automatically, either generated by the use of the Service or from the Service infrastructure itself (for example, the duration of a page visit). (Collecting and using your personal data) (Types of data collected) **Personal data** While interacting with the Services, You may provide certain personally identifiable information that could be used to contact or identify You. Personally identifiable information may include, but is not limited to: * Your user name for the Services * Text or image prompts and other content such as photos, videos, documents, and messages that you input into the Services * Your IP address * Usage Data * Tracking Technologies and Cookies * Contact Information * Organizational Information like your company title * Your Email * Cookies * Other data that you elect to send to 404-GEN, such as your survey responses or information you include in communications to us **Use of your personal data** 404-GEN may use Personal Data for the following purposes: * **To provide,maintain, and improve our Service,** including to monitor the usage of our Service. * **To manage Your account:** to manage Your registration as a user of the Service. The Personal Data You provide can give You access to different functionalities of the Service that are available to You as a registered user. * **For the performance of a contract:** the development, compliance and undertaking of the purchase contract for the products, items or services You have purchased or of any other contract with Us through the Service. * **To contact You:** To contact You by email, telephone calls, SMS, or other equivalent forms of electronic communication, such as a mobile application's push notifications regarding updates or informative communications related to the functionalities, products or contracted services, including the security updates, when necessary or reasonable for their implementation. * **To provide You** with news, special offers and general information about other goods, services and events which we offer that are similar to those that you have already purchased or enquired about unless You have opted not to receive such information. * **To manage Your requests:** To attend and manage Your requests to Us. * **For business transfers:** We may use Your information to evaluate or conduct a merger, divestiture, restructuring, reorganization, dissolution, or other sale or transfer of some or all of Our assets, whether as a going concern or as part of bankruptcy, liquidation, or similar proceeding, in which Personal Data held by Us about our Service users is among the assets transferred. * **For other purposes:** We may use Your information for other purposes, such as data analysis, identifying usage trends, determining the effectiveness of our promotional campaigns and to evaluate and improve our Service, products, services, marketing and your experience. #### Sharing of your personal data We may share Your personal information in the following situations: * **With Service Providers, Third Party Vendors, Consultants, and other Business Partners:** We may share Your personal information with these parties in order to provide services on our behalf, monitor and analyze the use of our services, contact You, and for the reasons stated in the Agreement. Service Providers to monitor and analyze the use of our Service, to contact You. * **For business transfers:** We may share or transfer Your personal information in connection with, or during negotiations of, any merger, sale of Company assets, financing, or acquisition of all or a portion of Our business to another company. * **With Your consent:** We may disclose Your personal information for any other purpose with Your consent. * **With Law Enforcement:** Under certain circumstances, 404-GEN may be required to disclose Your Personal Data if required to do so by law or in response to valid requests by public authorities (e.g. a court or a government agency). To the extent we receive a request from Law Enforcement for Your personal data, we will promptly notify You and provide You with a copy of the request, unless we are legally prohibited from doing so. * **With Other Parties in order to:** * Comply with a legal obligation * Protect and defend the rights or property of the Company * Prevent or investigate possible wrongdoing in connection with the Service * Protect the personal safety of Users of the Service or the public * Protect against legal liability **Retention of your personal data** The Company will retain Your Personal Data only for as long as is necessary for the purposes set out in this Privacy Policy. We will retain and use Your Personal Data to the extent necessary to comply with our legal obligations (for example, if we are required to retain your data to comply with applicable laws), resolve disputes, and enforce our legal agreements and policies. The Company will also retain Usage Data for internal analysis purposes. Usage Data is generally retained for a shorter period of time, except when this data is used to strengthen the security or to improve the functionality of Our Service, or We are legally obligated to retain this data for longer time periods. **2.5 Transfer of Your Personal Data** Your information, including Personal Data, is processed at the Company's operating offices and in any other places where the parties involved in the processing are located. It means that this information may be transferred to — and maintained on — computers located outside of Your state, province, country or other governmental jurisdiction where the data protection laws may differ from those in Your jurisdiction. Your consent to this Privacy Policy followed by Your submission of such information represents Your agreement to that transfer. The Company will take all steps reasonably necessary to ensure that Your data is treated securely and in accordance with this Privacy Policy and no transfer of Your Personal Data will take place to an organization or a country unless there are adequate controls in place including the security of Your data and other personal information. #### Security of your personal data The security of Your Personal Data is important to Us, but remember that no method of transmission over the Internet, or method of electronic storage is 100% secure. While We strive to use commercially acceptable means to protect Your Personal Data, We cannot guarantee its absolute security. (Children's privacy) Our Service does not address anyone under the age of 13. We do not knowingly collect personally identifiable information from anyone under the age of 13. If You are a parent or guardian and You are aware that Your child has provided Us with Personal Data, please contact Us. If We become aware that We have collected Personal Data from anyone under the age of 13 without verification of parental consent, We take steps to remove that information from Our servers. If We need to rely on consent as a legal basis for processing Your information and Your country requires consent from a parent, We may require Your parent's consent before We collect and use that information. (Links to other websites) Our Service may contain links to other websites that are not operated by Us. If You click on a third party link, You will be directed to that third party's site. We strongly advise You to review the Privacy Policy of every site You visit. Third party sites and services (including Discord) have separate practices regarding the protection of personal data that you provide to them. We have no control over and assume no responsibility for the content, privacy policies or practices of these third party sites or services. (Changes to this Privacy Policy) We may update Our Privacy Policy from time to time. We will notify You of any changes by posting the new Privacy Policy on this page. We will let You know via email and/or a prominent notice on Our Service, prior to the change becoming effective and update the "Effective as of" date at the top of this Privacy Policy. You are advised to review this Privacy Policy periodically for any changes. Changes to this Privacy Policy are effective when they are posted on this page. (Supplemental Terms and Conditions for certain regions) **Europe**\ If You are located in the European Economic Area (the “EEA”), Switzerland, or the United Kingdom (the “UK”) Our legal basis for collecting and using the Personal Data described in this Policy will depend on the personal data concerned and the specific context in which we collect it. However, we will normally collect personal data from you only where we have your consent to do so, where we need the personal data to perform a contract with you, or where the processing is in our legitimate interests and not overridden by your data protection interests or fundamental rights and freedoms. In some cases, we may also have a legal obligation to collect personal data from you. 404-GEN may share information internally or with third parties, as further described in this Policy. When we share Personal Data of individuals in the EEA, Switzerland or UK with third parties, we make use of a variety of legal mechanisms to safeguard the transfer including the European Commission-approved standard contractual clauses, as well as additional safeguards where appropriate. Additionally You have the following data protection rights: * You can request access, correction, updates or deletion of your Personal Data. * You can object to our processing of your Personal Data, ask us to restrict processing of your Personal Data or request portability of your Personal Data. * If we have collected and processed your Personal Data with your consent, then you can withdraw your consent at any time. Withdrawing your consent will not affect the lawfulness of any processing we conducted prior to your withdrawal, nor will it affect processing of your Personal Data conducted in reliance on lawful processing grounds other than consent. * You have the right to complain to a data protection authority about our collection and use of your Personal Data. (How to contact us) * Email: <404@404.xyz> # Unlocking Infinite Innovation Elo Ranking System for Bittensor Subnet 17 The ultimate goal of every Bittensor subnet is to drive innovation, which should be reflected in the incentives (reward formula). Subnet 17 provides a platform to democratize 3D content creation and incentivize miners to generate 3D content, prioritizing the throughput of generated content that passes the quality threshold. Current reward mechanism formula: $$ reward = quantity \* quality $$ ## Explanation Quantity refers to the number of generated assets that pass the validation criteria over a defined observation window. Quantity depends on three factors: validator's capacity, generation time, and delivery speed. The last two are within the miners' control. Generation time is self-explanatory. Delivery speed is crucial for ensuring a good end-user experience in getting results promptly. We incentivize miners to deploy edge nodes with good connections to validators to trim fractions of seconds for each result delivery. As for validator capacity, that has been our focus for the last couple of months. We've optimized the validation to handle 240 validations per minute (2x the initial load), with a theoretical maximum of 480 validations per minute (4x the initial load) on the same GPU. Scaling on multiple GPUs is also a configurable option. Quality is a more challenging metric. While we have success stories of artists using generated results, state-of-the-art text-to-3D models can't provide end-products and require some manual polishing. The ideal metric would be "how much work is needed to polish results before using," which is incredibly hard to implement. We practice a threshold approach by introducing a score value that results need to achieve to be accepted. For each result, we have three possible outcomes: denied, passed as high quality, or passed as medium quality. High-quality results get a 1.0 quality score, while medium-quality results get a 0.75 quality score (subject to change). Medium-quality results are not intended for use and are introduced to encourage new miners with positive feedback. They are also needed for transition periods (more on this below). The final quality score is the exponential moving average (EMA) of all quality results. The high and medium quality scores of 1.0 and 0.75 will soon be replaced with linear scoring to be more precise. To drive innovation, we are researching and implementing new ways of detecting defects. We closely monitor new developments in the market and increase quality thresholds with regular patches. We must align with the speed of development in the sphere to avoid starving the network of generations by setting the threshold too high or demotivating miners from exploring new options by setting the threshold too low. With each update, we intend to push current high-quality results down to the mid-range, penalizing miners and motivating them to improve. ## The Challenge While the formula allows us to incrementally drive quality improvement and has no defined limits, a cap exists. Infinitely scaling subnet throughput with synthetic traffic brings zero value and raises the question of pumping numbers for nothing. We already have a daily bandwidth of 5 million generations with the ability to increase it 2x with just one constant changed. We focus on increasing organic traffic (the demand) and maintaining a healthy proportion of organic/synthetic traffic. New validation or validation threshold increments have a certain cadence, which means we have periodic time windows when innovation occurs, followed by stability periods.

Incentive curve flattening

## Proposal: Elo Ranking System We want to add an additional factor to the reward formula that reflects the average miner quality relative to other miners. We want it to be statistically correct and take into account occasional bad generations. We propose adding miners' duels (clashes) where two pairs of miners get the same prompt, and their rank is updated based on the generated results. If we change the terminology, replacing "miner's quality" with "player's skill" and "miners' duel" with "chess match," we arrive at a problem that has been addressed for the last 70 years in statistically correct chess player ratings. The most well-known system, used in chess federations since 1960, is the Elo system, designed by Professor Arpad Elo, an active participant in the USCF. Elo's central assumption was that each player's chess performance in each game is a normally distributed random variable. Although a player might perform significantly better or worse from one game to the next, Elo assumed that the mean value of any given player's performances changes only slowly over time. Elo thought of a player's true skill as the mean of that player's performance random variable. From Wikipedia:
This formula has proven its value over the years and millions of matches, not only in chess but in other board games, online games, and some athletic sports. The naive implementation of the Elo ranking system in our context would look like this: $$ reward = Elo \* quality \* quantity $$ Previously, if we had a new leader on the subnet, they would pull slightly ahead of the other miners, with the other miners following soon thereafter. Now, assuming the subnet has reached its status quo with all miners producing similar results and having the same Elo rank, if a new leader emerges with superior results and wins each duel, their Elo rank would rise much more dramatically. An additional positive side effect would be that this leader might even have longer generation times, but superior quality will allow them to get more incentives, motivating others to reach the same quality and optimize generation and delivery times. {% embed url="" %} Simulation of Elo Implementation {% endembed %} ## Compensating for Elo's Deficiencies Arpad Elo in 1960 didn't have access to modern data analysis instruments or the vast collected statistics we have today. While the computational simplicity of the Elo system has proven to be one of its greatest assets, it has alredy been improved and adopted by different chess federations. USCF and FIDE switched from normal distribution to logistic distribution and updated constants based on collected statistics. Subjectively superior rating systems were developed in 1995 by Mark Glickman, called the Glicko and Glicko-2 rating systems. While not used in chess federations, they are implemented in several online games (e.g., CS: Global Offensive, Team Fortress 2, Dota 2, Guild Wars 2). Glickman's principal contribution to measurement is "ratings reliability," called RD for ratings deviation. Given this, before implementing the Elo rating system (or Glicko) into our formula, we need to implement duels, collect statistics, and find the probability distribution our miner "skill" follows, as well as determine the proper constants for the formulas. ## Additional Technological Challenges Thorough results comparison might be computationally heavy. Fortunately, we don't have to do it for every single prompt. We just need statistically enough duels for each miner. Plus, there are ways to add computationally heavy calculations with minimal effect on subnet throughput, starting from the cheapest but easiest options: * Pause the validator to perform comparison and rank update calculations (with the new gateway/hub implemented, this will not affect organic traffic) * Dedicate a separate GPU for comparison and rank updates * Offload calculations to other subnets ## Conclusion If implemented correctly, this system will unlock continuous, unlimited innovations in our subnet, driving consistent improvement in the quality and efficiency of 3D content generation. # 404—GEN Unity Add-on End-User License Agreement Last Updated: March 05, 2025 Please read this End-User License Agreement carefully before clicking the "I Agree" button, downloading or using 404—GEN. ### Interpretation and Definitions #### **Interpretation** The words of which the initial letter is capitalized have meanings defined under the following conditions. The following definitions shall have the same meaning regardless of whether they appear in singular or in plural. #### **Definitions** For the purposes of this End-User License Agreement: * **Agreement** means this End-User License Agreement that forms the entire agreement between You and the Company regarding the use of the Application. * **Application** means the software program provided by the Company downloaded by You to a Device, named 404—GEN. * **Company** (referred to as either "the Company", "We", "Us" or "Our" in this Agreement) refers to 404-GEN AG, Grabenstrasse 25, 6340 Baar, Switzerland. * **Content** refers to content such as text, images, or other information that can be posted, uploaded, linked to or otherwise made available by You, regardless of the form of that content. * **Country** refers to: Switzerland * **Device** means any device that can access the Application such as a computer, a cellphone or a digital tablet. * **Third-Party Services** means any services or content (including data, information, applications and other products services) provided by a third-party that may be displayed, included or made available by the Application. * **You** means the individual accessing or using the Application or the company, or other legal entity on behalf of which such individual is accessing or using the Application, as applicable. ### Acknowledgment By clicking the "I Agree" button, downloading or using the Application, You are agreeing to be bound by the terms and conditions of this Agreement. If You do not agree to the terms of this Agreement, do not click on the "I Agree" button, do not download or do not use the Application. This Agreement is a legal document between You and the Company and it governs your use of the Application made available to You by the Company. The Application is licensed, not sold, to You by the Company for use strictly in accordance with the terms of this Agreement. ### License #### Scope of License The Company grants You a revocable, non-exclusive, non-transferable, limited license to download, install and use the Application strictly in accordance with the terms of this Agreement. ### Third-Party Services The Application may display, include or make available third-party content (including data, information, applications and other products services) or provide links to third-party websites or services. You acknowledge and agree that the Company shall not be responsible for any Third-party Services, including their accuracy, completeness, timeliness, validity, copyright compliance, legality, decency, quality or any other aspect thereof. The Company does not assume and shall not have any liability or responsibility to You or any other person or entity for any Third-party Services. You must comply with applicable Third parties' Terms of agreement when using the Application. Third-party Services and links thereto are provided solely as a convenience to You and You access and use them entirely at your own risk and subject to such third parties' Terms and conditions. ### Term and Termination This Agreement shall remain in effect until terminated by You or the Company. The Company may, in its sole discretion, at any time and for any or no reason, suspend or terminate this Agreement with or without prior notice. This Agreement will terminate immediately, without prior notice from the Company, in the event that you fail to comply with any provision of this Agreement. You may also terminate this Agreement by deleting the Application and all copies thereof from your Device or from your computer. Upon termination of this Agreement, You shall cease all use of the Application and delete all copies of the Application from your Device. Termination of this Agreement will not limit any of the Company's rights or remedies at law or in equity in case of breach by You (during the term of this Agreement) of any of your obligations under the present Agreement. ### Indemnification You agree to indemnify and hold the Company and its parents, subsidiaries, affiliates, officers, employees, agents, partners and licensors (if any) harmless from any claim or demand, including reasonable attorneys' fees, due to or arising out of your: (a) use of the Application; (b) violation of this Agreement or any law or regulation; or (c) violation of any right of a third party. ### No Warranties The Application is provided to You "AS IS" and "AS AVAILABLE" and with all faults and defects without warranty of any kind. To the maximum extent permitted under applicable law, the Company, on its own behalf and on behalf of its affiliates and its and their respective licensors and service providers, expressly disclaims all warranties, whether express, implied, statutory or otherwise, with respect to the Application, including all implied warranties of merchantability, fitness for a particular purpose, title and non-infringement, and warranties that may arise out of course of dealing, course of performance, usage or trade practice. Without limitation to the foregoing, the Company provides no warranty or undertaking, and makes no representation of any kind that the Application will meet your requirements, achieve any intended results, be compatible or work with any other software, applications, systems or services, operate without interruption, meet any performance or reliability standards or be error free or that any errors or defects can or will be corrected. Without limiting the foregoing, neither the Company nor any of the company's provider makes any representation or warranty of any kind, express or implied: (i) as to the operation or availability of the Application, or the information, content, and materials or products included thereon; (ii) that the Application will be uninterrupted or error-free; (iii) as to the accuracy, reliability, or currency of any information or content provided through the Application; or (iv) that the Application, its servers, the content, or e-mails sent from or on behalf of the Company are free of viruses, scripts, trojan horses, worms, malware, timebombs or other harmful components. Some jurisdictions do not allow the exclusion of certain types of warranties or limitations on applicable statutory rights of a consumer, so some or all of the above exclusions and limitations may not apply to You. But in such a case the exclusions and limitations set forth in this section shall be applied to the greatest extent enforceable under applicable law. To the extent any warranty exists under law that cannot be disclaimed, the Company shall be solely responsible for such warranty. ### Limitation of Liability Notwithstanding any damages that You might incur, the entire liability of the Company and any of its suppliers under any provision of this Agreement and your exclusive remedy for all of the foregoing shall be limited to the amount actually paid by You for the Application or through the Application or 100 USD if You haven't purchased anything through the Application. To the maximum extent permitted by applicable law, in no event shall the Company or its suppliers be liable for any special, incidental, indirect, or consequential damages whatsoever (including, but not limited to, damages for loss of profits, loss of data or other information, for business interruption, for personal injury, loss of privacy arising out of or in any way related to the use of or inability to use the Application, third-party software and/or third-party hardware used with the Application, or otherwise in connection with any provision of this Agreement), even if the Company or any supplier has been advised of the possibility of such damages and even if the remedy fails of its essential purpose. Some states/jurisdictions do not allow the exclusion or limitation of incidental or consequential damages, so the above limitation or exclusion may not apply to You. ### Severability and Waiver **Severability** If any provision of this Agreement is held to be unenforceable or invalid, such provision will be changed and interpreted to accomplish the objectives of such provision to the greatest extent possible under applicable law and the remaining provisions will continue in full force and effect. **Waiver** Except as provided herein, the failure to exercise a right or to require performance of an obligation under this Agreement shall not affect a party's ability to exercise such right or require such performance at any time thereafter nor shall the waiver of a breach constitute a waiver of any subsequent breach. ### Product Claims The Company does not make any warranties concerning the Application. ### United States Legal Compliance You represent and warrant that (i) You are not located in a country that is subject to the United States government embargo, or that has been designated by the United States government as a "terrorist supporting" country, and (ii) You are not listed on any United States government list of prohibited or restricted parties. ### Changes to this Agreement The Company reserves the right, at its sole discretion, to modify or replace this Agreement at any time. If a revision is material we will provide at least 30 days' notice prior to any new terms taking effect. What constitutes a material change will be determined at the sole discretion of the Company. By continuing to access or use the Application after any revisions become effective, You agree to be bound by the revised terms. If You do not agree to the new terms, You are no longer authorized to use the Application. ### Governing Law The laws of the Country, excluding its conflicts of law rules, shall govern this Agreement and your use of the Application. Your use of the Application may also be subject to other local, state, national, or international laws. ### Entire Agreement The Agreement constitutes the entire agreement between You and the Company regarding your use of the Application and supersedes all prior and contemporaneous written or oral agreements between You and the Company. You may be subject to additional terms and conditions that apply when You use or purchase other Company's services, which the Company will provide to You at the time of such use or purchase. ### Contact Us If you have any questions about this Agreement, You can contact Us by visiting our Discord server: # Gateway API Whitepaper Drafted 29.05.2025 ## **Motivation and Overview** In preparation for several large partnerships, public API access and expected increased demand of organic traffic, Subnet 17 sought solutions for robust and efficient subnet access that could improve the way validators work with organic traffic while at the same time creating a foundation for future monetization strategies and additional functionalities as business needs evolve. Subnet 17 works with heavy 3D model file sizes and therefore this project is designed not only for high throughput but also to support advanced statistics and dynamic load balancing based on key parameters such as latency and the number of tasks in different regions. This scheme provides exceptional stability and performance, making it capable of handling up to 100,000+ user requests per second while laying a solid foundation for future enhancements. Key Components of the Project: • Consensus Engine – Based on Raft (OpenRaft), the system achieves distributed consensus across nodes and shares the internal state for every node in the cluster. • Transport Protocol – QUIC transports RAFT messages, ensuring fast, low-latency communication while also encrypting all data. • API Layer – The interface operates over HTTP/3, adhering to modern web standards for responsive client interactions. • Encryption – Rustls provides TLS 1.3 encryption to secure communication channels. • Concurrency Model – The solution is entirely non-blocking and employs lock-free algorithms throughout, with the only exception being an RwLock used in the commit log for consistency. ## **How Raft (OpenRaft) Works** Every node initially functions as a voting member, and the first process they undertake is to elect a leader. Here’s how the process unfolds: 1\. All nodes start as voting members (also known as followers) and participate in the initial leader election. 2\. If a node doesn’t hear from an active leader (through heartbeat messages or similar signals), it considers the leader unresponsive, transitions to a candidate state, and initiates an election. 3\. The candidate solicits votes from the other nodes—each of which is already a voting member—and if it secures a majority, it is elected as the new leader. 4\. With a leader now in place, the leader takes on the responsibility of sending out regular heartbeats and log entries to the followers, ensuring data consistency across the cluster. 5\. Maintaining cluster integrity generally requires that a majority (typically **two-thirds**) of the nodes remain online and responsive, securing the quorum needed for both leader elections and log replication. Once a leader is chosen, it starts sending heartbeats immediately and begins processing client requests. This election mechanism guarantees that the cluster always has an active leader, while ensuring that only one leader is active at any time to maintain consistency of our data (in our case it's just a which saves the internal state of all nodes. Each node has its own task queue, which is not synchronized with other nodes, **sharing only the size of the queue of available tasks in the global state.** In the cluster, write operations are always handled by the leader. When a node needs to write data, it sends the request directly to the leader. The leader then appends the new entry to its local log and initiates a replication process by sending the entry to the follower nodes. Each follower appends the received entry to its own commit log. Once a majority of nodes have successfully stored the entry, the leader marks it as committed and applies it to its state machine. Only after this confirmation does the client receive an updated response. This process ensures data consistency across the entire cluster. ## Gateways Global State and JSON Representation All nodes in the network maintain one global state, and this state is represented using JSON. This JSON structure provides a consistent and unified snapshot of the cluster's status—including each node's details and its respective available\_tasks. Because the global state is replicated among all nodes through the Raft log replication process, any node in the network can serve queries about the current state with minimal delay. For instance, consider the following internally shared JSON structure that represents the entire cluster: ``` { "gateways": [ { "node_id": 1, "domain": "gateway-eu.404.xyz", "ip": "5.9.29.227", "name": "node-1-eu", "http_port": 4443, "available_tasks": 0, "last_task_acquisition": 1746010976, "last_update": 1746011013 }, { "node_id": 2, "domain": "gateway-us-east.404.xyz", "ip": "3.226.98.135", "name": "node-2-us-east", "http_port": 4443, "available_tasks": 0, "last_task_acquisition": 1746010975, "last_update": 1746011013 }, { "node_id": 3, "domain": "gateway-us-west.404.xyz", "ip": "13.56.102.231", "name": "node-3-us-west", "http_port": 4443, "available_tasks": 0, "last_task_acquisition": 1746010978, "last_update": 1746011014 } ] } ``` Notice that any change made to the **available\_tasks** field is propagated throughout the cluster in a timely manner through the log replication. As soon as an update is committed by the leader and applied to the state machines of the majority of nodes, the global JSON state is refreshed. This ensures that if you query any node in the network, you'll receive the most up-to-date information with minimal delay. The field **last\_task\_acquisition** indicates the most recent time a task was obtained from the Gateway. This helps validators develop strategies to balance their subsequent requests effectively and decide whether to collect a task from a specific gateway, especially when the task count is very low. The validators also receive the structure above when tasks are requested from any Gateway. This way, they know in advance where they need to go for tasks next time, and they can also balance depending on the region (for example, by using latency and last\_task\_acquisition). **This architecture addresses current high-demand environments and is structured to support further development in analytics, convenient API key management, load balancing, and monetization, ensuring a robust, future-proof platform that can scale beyond conventional limitations.** ## Validator Behavior and Optimization Validators are engineered to process and distribute tasks in a highly efficient, adaptive manner. They rely on the latest global state—which includes real‐time details such as available\_tasks, last\_task\_acquisition and latency. Key aspects include: * Validators initially source tasks from the geographically closest gateway node, minimizing latency and response times. * Each validator maintains an optimized task queue that minimizes both gateway queue time and validator processing time, ensuring efficient throughput. * The system implements dynamic load balancing where validators monitor the load across gateway nodes and can pull tasks from distant nodes when necessary to maintain overall system performance. * Latency tracking mechanisms allow validators to prioritize gateways based on response metrics, continuously adjusting to minimize the average task delivery time to end users. * Validators retain flexibility in traffic participation, with the option to opt out of serving organic traffic based on their resource allocation preferences. * For customized deployment, validators can establish their own gateway clusters to attract and manage dedicated traffic streams. * Advanced configuration options enable validators to interface with multiple separate gateway clusters simultaneously, with assigned priorities to optimize workload distribution. * Regarding miners, the established prioritization remains unchanged—organic traffic maintains priority status and is directed to miners first, preserving the core workflow while enhancing the overall system capabilities. ## Conclusion As Subnet 17 prepares for increased organic traffic, we feel this project provides a basis to maintain flexibility and control necessary to not only grow in scale but also to enable new types of interactions and monetization strategies that will align with the changing business needs of the subnet. By open sourcing the underlying code [here](https://github.com/404-Repo/gateway-rs) and detailing the conceptual and technical reasons for the project above we also hope that other subnet owners and validators may benefit from similar implementations. # 31 October 2D to 3D Benchmark Metrics On 24 October 2D inputs went live on Subnet 17 mainnet enabling direct like-for-like comparison with existing generative 3D foundational models. This is a snapshot of the results after 1 week. This benchmark evaluates leading generative 3D foundational models. It is inspired by [3D Arena](https://huggingface.co/spaces/dylanebert/3d-arena) and uses Visual Language Model judges as a non-human evaluation criteria. Read more about evaluation methodology [here](https://github.com/404-Repo/three-gen-subnet/tree/main). #### Motivation Evaluating generated 3D quality quantitatively is challenging, subjective, and there is no standard practice for evaluating aesthetics in real-world applications. This head-to-head competition utilizes the strength of reasoning models in like-for-like comparisons and displays final results side-by-side with the ability to download files for further human evaluation. #### Criteria Models must handle image (.png, .jpg) inputs and produce mesh (.obj, .glb) or splat (.splat, .ply) outputs. They should run end-to-end without human intervention, including UV unwrapping, texture mapping, and other post-processing. #### Contributing All inputs and outputs are publicly available [here](https://huggingface.co/datasets/dylanebert/3d-arena). Input image URLs are provided [here](https://huggingface.co/datasets/dylanebert/3d-arena/raw/main/inputs.txt). #### Results **404 v. CSM Cube**\ 404 wins: 43 \ CSM Cube wins: 16 \ Draw: 42\ [Results & Data](https://3d-arena-images.b-cdn.net/comparison/arena_duels_cube.html) **404 v. Trellis**\ 404 wins: 39 \ Trellis wins: 17 \ Draw: 45\ [Results & Data](https://3d-arena-images.b-cdn.net/comparison/arena_duels_trellis.html) **404 v. Hunyuan 2.1**\ 404 wins: 70 \ Hunyuan wins: 15 \ Draw: 16\ [Results & Data](https://3d-arena-images.b-cdn.net/comparison/arena_duels_hunyan-2.html) **404 v. Meshy**\ 404 wins: 97 \ Meshy wins: 2 \ Draw: 1\ [Results & Data](https://3d-arena-images.b-cdn.net/comparison/arena_duels_meshy.html) #### Future Benchmarks This set of generations (1 week after launching on mainnet) has been submitted to [3D Arena](https://huggingface.co/spaces/dylanebert/3d-arena) for human-in-the-loop ranking on their leaderboard. This benchmark process will be updated with additional closed source models on an ongoing basis (as new models are released). # Atlas and 404: Our Application Layer for Decentralized 3D Intelligence Building the bridge between decentralized 3D AI and enterprise adoption. #### The Vision 404 is building toward Large Spatial Models. Understanding how objects exist and relate in 3D space is something current AI struggles with. It's the missing capability between language models that understand text and systems that can actually interact with the physical world. These models will define the next era of AI innovation. To build toward this vision, Subnet 17 runs 3D AI competitions. Miners compete to build the best models - submitting code that generates meshes, gaussian splats, and procedural assets. Validators evaluate those models for quality, speed and structure. Over time, this produces not just 3D data but the systems and frameworks for spatial manipulation, composition, and physics-aware generation. These competitions build on each other, creating continuously improving models that expand AI's ability to generate digital worlds and understand / interact with the physical world. They are also commodities. #### The Commodification of Intelligence 404's competition framework uses open-source submissions to ensure that miners build atop each other, accelerating innovation amid constantly increasing competition. As a result, the subnet pays per innovation rather than per output, and incentivizes miners to optimize speed, cost, and quality to survive. This type of hyper-competitive framework creates a fundamentally different trajectory than centralized AI platforms. On a centralized platform, quality improves when the company invests in R\&D. Speed improves when they optimize infrastructure. And prices increase over time as the platform captures margin. On 404, quality improves because miners compete to win. Speed improves because generation time is a competitive advantage. And costs decrease over time as decentralized compute scales and miners optimize for their own efficiency. | Metric | Centralized Platforms | 404 (SN 17) | | -------------- | ---------------------------------------------- | ---------------------------------------------------- | | **Quality** | Improves with R\&D budget cycles | Improves continuously (miners compete daily) | | **Speed** | Improves slowly with infrastructure investment | Improves fast (speed directly affects rewards) | | **Cost** | Increases over time (margin capture) |

Decreases over time
(competition + scale)

| | **Innovation** | Limited to internal team | Global (open to anyone) | The role of 404 is to direct that intelligence into work that both 1) can provide value today and 2) can further our vision of building Large Spatial Models. Competitions range from generating raw data in the form of gaussian splats, meshes, and procedural representations, to spatial editing, composition and physics aware generations. The longer 404 runs, the more diverse and expansive the tech stack becomes. Over time, this expansion addresses the needs of 3D generation (content creation tools, world models and game engines, robotics and digital twins), and the effects compound. 404 will become the default 3D generation layer for the internet: the cheapest, fastest, highest-quality infrastructure available, improving every day through open competition. Not because we're the only option, but because the economics of decentralized competition compound in ways centralized platforms cannot match. #### AI Innovation → Real World Adoption Such an expansive vision requires a balance between innovation-oriented competitions and real usage proof points. Market usage data becomes a key component, as adoption, revenue, and achievements serve as indicators for current and future viability of these models. Within the creative industries, adoption is a key signal of future success, driving both feedback to improve AI models and competitive advantages to drive revenue. Midjourney (a 2D AI company) provides a clear case study. By providing a free product, they gained adoption with individual content creators - first generating insights required for development, then building network effects that allowed them to earn $50 million when they later moved to a subscription model. 404 follows a similar path and as such, we make extensive efforts to ensure our tech is not locked to any single interface. Developers, studios, and creators can access the network however fits their workflow with a limited amount sample generations to lower the barrier to entry: **Unity Plugin** The first decentralized 3D model generation integration for Unity as an official Verified Solution - [Unity Asset Store](https://assetstore.unity.com/packages/tools/generative-ai/404-gen-3d-generator-311107).
**Blender Add-on** Our most popular open source integration - [Blender Marketplace](https://404gen.gumroad.com/).
**Web App** Our lowest barrier to entry - [Web Generator](https://gen.404.xyz/). **Direct API** For developers building custom integrations. [*Atlas*](https://atlas.design/) *is our own custom integration.* #### Atlas If 404 is open infrastructure, why build Atlas? Because enterprises need more than raw generation. They need production-ready pipelines. #### The Two-Layer Stack We believe AI technology needs to exist at two distinct layers:
Cover image
The Intelligence Layer404 (Subnet 17) — where models compete, improve, and generate 3D assets through decentralized incentives.
The Application LayerAtlas — where those capabilities become products that enterprises and creators can actually use.
This is the architecture that will lead to enterprise Subnet 17 adoption. 404 is infrastructure. Atlas is the interface. Together, they form a complete stack for bringing decentralized 3D intelligence to the world. #### What actually is Atlas? {% embed url="" %} Atlas is an orchestration layer that takes raw 3D generation from Subnet 17 and makes it production-ready - handling texture refinement, format conversion, and quality filtering through integration with other AI models. Game studios use it to generate assets at scale. At enterprise scale, no single model produces perfect results for every use case, and enterprises need a variety of solutions. For example, a gaussian splat may be visually stunning but require mesh LODs for gameplay, or a fast generation mesh may need 2D texture refinement for studio use. Atlas solves this by combining 404's outputs with other AI models in node-based workflows:
This broader enhancement pipeline integrates models beyond 404's core 3D generation, including 2D image, video, rigging, and animation models - ultimately making 404's outputs more valuable to enterprises. #### No White Labeling: Atlas showcases (and charges for) 404 When users generate assets through Atlas, they see that the generation is powered by 404. Transparency builds trust. When a professional studio discovers 404 through our Unity and Blender plugins they can move to Atlas to generate at scale using automated quality pipelines, format optimization, multi-model orchestration, and enterprise support. This value-add amplifies the professional use cases of 404. And paying customers on Atlas that use 404 will be paying 404 - Atlas isn't taking a cut of that revenue. Ultimately for 404 to succeed long-term, the technology needs to generate real revenue because it is truly State-of-the-Art and can win on its own output, speed, and visual fidelity - revenue from customers paying for value delivered. Atlas is designed to facilitate and scale that revenue path with enterprise customers. The model is straightforward: usage-based pricing. Customers pay per generation, per asset, per API call. #### Why Enterprises Choose 404 via Atlas There are other 3D generation options. Meshy, Tripo, Hunyuan, Trellis - the market isn't empty. So why would enterprises choose Atlas and 404? * Enterprise use cases balance quality with optimization and generation speed. 404 is the fastest high quality foundational model on the market today. * For mobile gaming assets and real-time generation this is a game changer and can welcome a new era of AI-native ["Living Games"](https://cloud.google.com/blog/products/gaming/games-start-ups-developers-partners-innovating-with-gen-ai). * For use cases requiring maximum visual fidelity and/or mesh topology, Atlas enhances 404's outputs through multimodal orchestration. * Enterprise customers are wary of depending on a single AI provider and Atlas allows them to test with 404 before building and embedding a content creation pipeline around it. * Atlas's orchestration layer can be configured for customization of 404 at scale around specific use cases. * The pipeline can be tuned: different quality thresholds, different formats, different enhancement steps. This flexibility is hard to match with single-model interfaces. * Atlas is now a well-known web2 company, having launched the solution in [partnership with Google Cloud](https://www.globenewswire.com/news-release/2025/08/18/3135265/0/en/Atlas-Partners-with-Google-Cloud-to-Power-the-Next-Generation-of-AI-Native-3D-Game-Development.html). * Atlas has existing customer relationships with major studios like Square Enix - One of the largest game publishers in the world. * This opens doors to enterprises that would be difficult for 404 to build without an application layer. > ### "Atlas's ability to seamlessly integrate with our highly customized workflows has been a game-changer." > > — Joseph Burnette, Technical Director, Innovation Technology Division, SQUARE ENIX This is a route to take our subnet 'mainstream' - not by asking enterprises to change how they operate, but by meeting them where they are. #### Why This Architecture Wins 404 stays focused on the hard problem: building the best decentralized competition system for 3D generation. Miner incentives. Validator infrastructure. Model evaluation. Network reliability. This is deep, technical work that requires singular focus. Atlas stays focused on the market problem: understanding what customers actually need, building products they'll pay for, navigating enterprise sales cycles, ensuring quality meets expectations. This is commercial work that requires different skills. Atlas exists to bring customers to 404. 404 exists to provide the intelligence that Atlas delivers. We're building both layers because we believe you need both to win at enterprise scale. #### What's Next The pieces are in place. * 404 will continue to evolve towards Large Spatial Models through innovating and commoditizing key 3D AI technologies. * 404's open infrastructure will facilitate a diverse ecosystem of builders creating games, world models, tools, AR/VR/XR, digital twins and other unique applications of immersive environments. * Atlas will focus on enterprise adoption of these technologies. The roadmap is a delicate balance of innovation, execution, research and showing real-world proof points via adoption, benchmarks and value. * 404 has millions of generated assets, a proven competition system, and active miners improving models daily. * Atlas has enterprise relationships, a working product, and customers ready to pay. * 404's technology has been demonstrated at GDC, the World Expo, Gamescom and the Venice Biennale. The question was never whether decentralized 3D generation could work. That's been proven. The question is whether it can become the spatial-reasoning foundation for the next wave of AI innovation on the scale of OpenAI and Anthropic - one that competes with these centralized players while maintaining the openness and incentive alignment that only decentralization can provide. We're building 404 to be that foundation.
# Code-Generated Assets for the Agentic Era It’s time agents start mining on 404 #### The Shift The most interesting conversations in gaming aren't happening in the offices of major studios. They're happening in GitHub repos and Discord servers and X. They're with people building games in ways that didn't exist two years ago. Solo creators shipping full experiences in a weekend. Agents - actual AI agents - generating complete playable builds from a prompt. 25% of Y Combinator's Winter 2025 batch had codebases that were 95%+ AI-generated, and we're seeing the same trend in gaming. The first agent game dev studios are appearing and they're taking off. "Claude Code Game Studios" - an open source project that turns Claude Code into a full game dev studio with 48 AI agents, 36 workflow skills, and a complete coordination system mirroring real studio hierarchy - grew to over 8,000 stars within a month. This isn't replacing or even competing with enterprise game development. It's creating an entirely new layer beneath it where individuals and small teams generate a new genre of games. #### 3D as Code not Mesh Thus far these games have grown within the 2D space because they run into limitations of traditional 3D representation. Typical meshes don't work. They're "opaque" - the agent can't read them, can't modify them, can't reason about interactivity and gameplay mechanics. What agents need is code. Specifically, code that generates 3D geometry. Something they can read, understand, tweak, and rebuild. These are called procedural assets. This is the focus of 404's next competition. #### Competition Framework This is the third form of 3D representation the subnet has produced - each competition building on the last toward Large Spatial Models (LSMs). We started with Gaussian Splats for visual fidelity and research trajectory. Then Meshes for physics engines and enterprise adoption. Now Procedural Assets for simulation and agentic infrastructure. Each representation brings something the others can't, and all three will converge into the LSMs. For procedural, the code becomes the asset - parameterized, inspectable, infinitely variable. Exactly what this new genre of games requires. Each competition balances research with real-world adoption - innovations that actually get used. This shift towards code generation is also designed such that agents can participate directly. For the first time on Subnet 17, we expect to see AI agents mine the subnet. Write code, submit assets, compete autonomously and openly. We're not just building infrastructure for agent-created games - we're encouraging agents to be the ones who build it. We believe this design represents a positive example for Bittensor mining as a whole (as other subnets have done) - Karpathy-style research, decentralized, permissionless and agentic. #### Furthering Commercial & Research We're building toward Large Spatial Models - models that understand 3D space, object relationships, and physics. This requires data generation pipelines across different forms of representation, each contributing unique value: * **Gaussian Splats** — visual fidelity and composition * **Meshes** — physics simulation and explicit editing * **Procedural Code** — parameterization, variation, and structural reasoning A splat tells you what something looks like. A mesh integrates it within an environment. Procedural code tells you what something *is* — the operations that define it, the parameters that control it. That's the kind of structured understanding the foundation model needs. With these three representations established, future competitions can focus on combination and composition: world and scene generation, editing across formats, physics-aware generation, and ultimately the Large Spatial Model itself.