
Inside DaniLabs (SN2): The Bittensor Subnet Focused on Speed, Cost Goals, and Privacy
DaniLabs (SN2) is building a distributed AI inference network on Bittensor focused on faster responses, lower costs, and privacy-conscious AI services. […]

DaniLabs (SN2) is building a distributed AI inference network on Bittensor focused on faster responses, lower costs, and privacy-conscious AI services. […]
<p class="wp-block-paragraph">Imagine using an AI assistant that responds quickly, costs less to run, and does not retain your prompts after processing them. For developers building AI-powered applications, these are practical needs that can influence how much their products cost and how users trust them.</p> <p class="wp-block-paragraph">DaniLabs (SN2) is working toward this kind of infrastructure on Bittensor. Its goal is to distribute AI inference across a network of miners, allowing applications to access model responses while prioritizing speed, low costs, and data privacy.</p> <p class="wp-block-paragraph">The subnet is still in active development, but its validator infrastructure is already running as the team works toward launching its broader inference service.</p> <h2 class="wp-block-heading">What Is DaniLabs?</h2> <p class="wp-block-paragraph"><a href="https://danilabs.ai/">DaniLabs</a> stands for <strong>Distributed Accelerated Neural Intelligence</strong>. It is a Bittensor subnet focused on building a distributed AI inference network, with the tagline “Intelligence, distributed.”</p> <p class="wp-block-paragraph">The project aims to make AI inference more accessible by allowing developers and applications to send requests to the network and receive responses from models running across distributed compute providers.</p> <figure class="wp-block-image size-large is-resized"><img loading="lazy" decoding="async" width="1024" height="281" src="https://taodaily.io/wp-content/uploads/2026/11/image-97-1024x281.png" alt="" class="wp-image-26307" style="aspect-ratio:3.6491228070175437;width:624px;height:auto" srcset="https://taodaily.io/wp-content/uploads/2026/11/image-97-1024x281.png 1024w, https://taodaily.io/wp-content/uploads/2026/11/image-97-300x82.png 300w, https://taodaily.io/wp-content/uploads/2026/11/image-97-765x210.png 765w, https://taodaily.io/wp-content/uploads/2026/11/image-97.png 1272w" sizes="(max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption"><a href="https://danilabs.ai/">DaniLabs Core Priorities</a></figcaption></figure> <p class="wp-block-paragraph">Its three core priorities are:</p> <p class="wp-block-paragraph">1) <strong>No data retention:</strong> Prompts and responses are processed without being stored after a request.</p> <p class="wp-block-paragraph">2) <strong>Speed:</strong> Reduce the time users wait for a usable response, including time to first token and overall generation speed.</p> <p class="wp-block-paragraph">3) <strong>Low cost:</strong> Make AI inference more affordable so developers can integrate useful AI features into everyday applications and standard plans.</p> <p class="wp-block-paragraph">These priorities shape the service DaniLabs is developing. The team has already recorded more than <strong>22,000 requests</strong> during its early testing and has been speaking with potential customers who are routing inference workloads through the network.</p> <figure class="wp-block-embed is-type-rich is-provider-x wp-block-embed-x"><div class="wp-block-embed__wrapper"> <blockquote class="twitter-tweet" data-width="500" data-dnt="true"><p lang="en" dir="ltr">We’re building AI inference on Bittensor. So far, we’ve recorded over 22,000 requests.<br><br>Take a look at the usage behind that number.<a href="https://t.co/qH3GWygNWW">https://t.co/qH3GWygNWW</a> <a href="https://t.co/GQRbH771nF">pic.twitter.com/GQRbH771nF</a></p>— DaniLabs (@danilabsai) <a href="https://x.com/danilabsai/status/2120423156985016766?ref_src=twsrc%5Etfw">November 11, 2026</a></blockquote><script async src="https://platform.x.com/widgets.js" charset="utf-8"></script> </div></figure> <h2 class="wp-block-heading">From DSperse to DaniLabs</h2> <p class="wp-block-paragraph">DaniLabs did not begin with a completely new subnet. <strong>Subnet 2</strong> previously operated as <a href="https://taodaily.io/dsperse-subnet-2-building-the-trust-layer-for-verifiable-ai-on-bittensor/"><strong>DSperse</strong></a> under <strong>Inference Labs</strong>, where the focus was on zero-knowledge machine learning and cryptographic proofs that verified AI computations.</p> <p class="wp-block-paragraph">That work generated billions of proofs through developer and community activity. In mid-October 2026, Inference Labs handed the subnet over to DaniLabs, led by <strong>Dan Ivanov</strong>, known as <a href="https://x.com/opendansor">@opendansor</a>, who had contributed to scaling the previous project.</p> <p class="wp-block-paragraph"><strong>≫</strong> See our earlier coverage on this transition, <a href="https://taodaily.io/danilabs-takes-over-dsperse-sn2-with-plans-for-faster-cheaper-ai-inference/">HERE</a>.</p> <h2 class="wp-block-heading">How DaniLabs Works</h2> <figure class="wp-block-image size-large is-resized"><img loading="lazy" decoding="async" width="1024" height="516" src="https://taodaily.io/wp-content/uploads/2026/11/image-98-1024x516.png" alt="" class="wp-image-26308" style="aspect-ratio:1.980952380952381;width:624px;height:auto" srcset="https://taodaily.io/wp-content/uploads/2026/11/image-98-1024x516.png 1024w, https://taodaily.io/wp-content/uploads/2026/11/image-98-300x151.png 300w, https://taodaily.io/wp-content/uploads/2026/11/image-98-767x387.png 767w, https://taodaily.io/wp-content/uploads/2026/11/image-98.png 1271w" sizes="(max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption">Building The Future Of AI Inference On Bittensor </figcaption></figure> <p class="wp-block-paragraph">DaniLabs follows Bittensor’s standard miner-validator structure on Subnet 2 (<strong>Netuid 2</strong>).</p> <p class="wp-block-paragraph">Miners provide the compute resources that run AI models and generate responses to inference requests while validators assess miner performance and submit on-chain weights, which determine how subnet emissions are distributed.</p> <p class="wp-block-paragraph">The validator system evaluates several operational metrics:</p> <p class="wp-block-paragraph">a) <strong>Decode speed:</strong> How quickly a miner generates tokens, with this metric weighted heavily.</p> <p class="wp-block-paragraph">b) <strong>Time to first token:</strong> How long it takes for a response to begin.</p> <p class="wp-block-paragraph">c) <strong>Completion reliability:</strong> Whether inference requests complete successfully.</p> <p class="wp-block-paragraph">d) <strong>Gateway-to-miner latency:</strong> The round-trip time between the service gateway and a miner.</p> <p class="wp-block-paragraph">The validator code is publicly available in the <a href="https://github.com/DaniLabsAI">DaniLabs subnet GitHub repository</a>. It verifies signed performance reports from miners against finalized chain state, helping ensure that reported measurements can be checked.</p> <h2 class="wp-block-heading">What Comes Next for DaniLabs?</h2> <figure class="wp-block-image size-large is-resized"><img loading="lazy" decoding="async" width="1024" height="576" src="https://taodaily.io/wp-content/uploads/2026/11/image-99-1024x576.png" alt="" class="wp-image-26309" style="aspect-ratio:1.7777777777777777;width:624px;height:auto" srcset="https://taodaily.io/wp-content/uploads/2026/11/image-99-1024x576.png 1024w, https://taodaily.io/wp-content/uploads/2026/11/image-99-300x169.png 300w, https://taodaily.io/wp-content/uploads/2026/11/image-99-768x432.png 768w, https://taodaily.io/wp-content/uploads/2026/11/image-99-678x381.png 678w, https://taodaily.io/wp-content/uploads/2026/11/image-99.png 1672w" sizes="(max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption">Building AI Inference On Bittensor</figcaption></figure> <p class="wp-block-paragraph">DaniLabs is still early in its development. Its validator infrastructure is running, performance scoring is being tested, and the team is measuring usage while preparing the inference service