Explore how TPN (SN65) pivoted from a profitable VPN to AI model compression on Bittensor, helping language models fit users’ hardware while preserving essential capabilities. […]
<p class="wp-block-paragraph">A profitable product is usually a reason to keep building. For the team behind <a href="https://taodaily.io/tpn-sn65-to-pivot-from-privacy-infrastructure-to-ai-optimization/">True Performance Network (TPN)</a>, it prompted a closer examination of the opportunity ahead.</p> <p class="wp-block-paragraph">Speaking on Episode 106 of the <a href="https://taodaily.io/djinn-sn103-on-the-ventura-labs-podcast-selling-the-signal-without-losing-the-secret/">Ventura Labs Podcast</a>, co-founder <a href="https://taodaily.io/meet-mentor-the-tpn-co-founder-who-turned-a-vpn-into-a-llm-compression-subnet/">Mentor Palokaj</a> explained why the team moved away from a working decentralized VPN to focus on compressing large language models. The decision centred on whether the business could deliver the scale of growth they sought.</p> <figure class="wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio"><div class="wp-block-embed__wrapper"> <iframe title="Mentor Palokaj: AI Model Compression & Quantization | TPN, Bittensor Subnet 65 | Ep. 106" width="500" height="281" src="https://www.youtube.com/embed/y96hZTVjZaI?feature=oembed" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe> </div></figure> <p class="wp-block-paragraph"></p> <p class="wp-block-paragraph">The team’s first venture was a fully compliant TAO-to-EVM bridge developed in the Netherlands. After that product became obsolete, it built Taofu, a fundraising platform for tokenizing subnet ownership stakes. The concept attracted considerable discussion, but actual usage remained limited.</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">The first Subnet Seeds launch is coming on Taofu. <br><br>🗓 Sale opens on Friday, June 6th 2025 <br>🖊️Register now: <a href="https://t.co/G3znBgLYEK">https://t.co/G3znBgLYEK</a><a href="https://x.com/sn65_tpn?ref_src=twsrc%5Etfw">@SN65_TPN</a> leads the way. <br><br>Full sale details on Discord: <a href="https://t.co/tADeqyK2q5">https://t.co/tADeqyK2q5</a> <a href="https://t.co/MIehkmH9ov">pic.twitter.com/MIehkmH9ov</a></p>— τaofu (τ, τ) (@taofuxyz) <a href="https://x.com/taofuxyz/status/1929572739918025079?ref_src=twsrc%5Etfw">June 2, 2025</a></blockquote><script async src="https://platform.x.com/widgets.js" charset="utf-8"></script> </div></figure> <p class="wp-block-paragraph">The decentralized VPN that followed achieved stronger traction. It reached <a href="https://taodaily.io/tpn-vpn-app-launches-on-android-expanding-access-to-decentralized-privacy-globally/">app stores</a>, attracted reliable miners, secured paying users, and became profitable. Nevertheless, during a six-hour review around May, the founders concluded that AI model compression better matched their expertise and ambitions.</p> <p class="wp-block-paragraph">Mentor described the willingness to pivot as a practical response to market feedback, available capital, and the team’s interests. Maintaining trust remained central to that process. His advice to “be a cockroach” captured the importance of surviving and adapting while acting in good faith.</p> <p class="wp-block-paragraph">Today, <a href="https://taodaily.io/tpn-sn65s-genesis-will-put-4gb-ai-models-on-a-2gb-diet/">TPN (SN65)</a> operates with the goal of making language models perform effectively within specific hardware constraints. The process starts with a model, a target device, and the capabilities the user wants to preserve. Compression then seeks the best balance between memory requirements, retained intelligence, and inference speed.</p> <p class="wp-block-paragraph"><a href="https://taodaily.io/shrinking-ai-models-just-got-easier-with-tpn-sn65s-new-toolkit/">Quantization</a> is one of the main techniques involved. By reducing the numerical precision used to represent model weights, it lowers the memory needed to run a model. Mentor explained that moderate reductions, such as moving from 16-bit to 8-bit precision, can preserve much of its capability, although the outcome depends on the model and task.</p> <p class="wp-block-paragraph">More advanced approaches include mixed-precision schemes that protect parts of a model that are particularly sensitive to compression. <a href="https://taodaily.io/tpn-bench-expanded-its-toolkit-to-preserve-ai-performance-during-compression/">Quantization-aware distillation</a> adds another method, using the original model as a teacher to help a compressed student reproduce its behaviour.</p> <p class="wp-block-paragraph">Evaluating the result is a central challenge. A compressed model can perform well on familiar public benchmarks without retaining the broader capabilities users expect. Repeated optimization against known questions can also make benchmark scores less informative about performance on unfamiliar tasks.</p> <p class="wp-block-paragraph">TPN (SN65)’s response combines task-based evaluation with measures of behavioural fidelity. Mentor described <a href="https://taodaily.io/tpns-private-benchmark-tests-models-without-prior-access-to-questions/">Fluid Bench</a> as a generative meta-benchmark that samples knowledge points to produce questions whose exact form is unknown in advance. The approach is intended to make it harder to optimize solely for a fixed question set.</p> <p class="wp-block-paragraph">The team also uses KL divergence to compare the original and compressed models’ full next-token probability distributions. This assesses how closely their predictions align, complementing benchmark results rather than replacing them. According to Mentor, miners can choose different trade-offs between task scores and similarity to the original model.</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">New: Validate a quant tool.<br><br>With TPN Bench quant verification tool, users can deterministically or probabilistically prove that a quantization is indeed a quant of the base model.<br><br>The tool's report shows how often the quant picks the same next word (top-1 match), how far its… <a href="https://t.co/HXQOfG6xxA">pic.twitter.com/HXQOfG6xxA</a></p>— True Performance Network (@TPN_Labs) <a href="https://x.com/TPN_Labs/status/2116117021951750225?ref_src=twsrc%5Etfw">October 30, 2026</a></blockquote><script async src="https://platform.x.com/widgets.js" charset="utf-8"></script> </div></figure> <p class="wp-block-paragraph">TPN’s public workbench at <a href="https://taodaily.io/how-far-tpn-bench-pushes-independent-model-validation/">TPN Bench</a> makes these choices accessible through a configurable workflow. Users submit a Hugging Face model, specify hardware constraints and acceptable loss percentages, and receive a matrix of quantized or distilled variants. They can rank the results by their preferred metrics, with Mentor highlighting the “Optimize” feature as particularly useful.</p> <figure class="wp-block-image size-large is-resized"><img decoding="async" width="1024" height="431" src="https://taodaily.io/wp-content/uploads/2026/10/image-302-1024x431.png" alt="" class="wp-image-26001" style="aspect-ratio:2.3726235741444865;width:624px;height:auto" srcset="https://taodaily.io/wp-content/uploads/2026/10/image-302-1024x431.png 1024w, https://taodaily.io/wp-content/uploads/2026/10/image-302-300x126.png 300w, https://taodaily.io/wp-content/uploads/2026/10/image-302-768x323.png 768w, https://taodaily.io/wp-content/uploads/2026/10/image-302.png 1895w" sizes="(max-width: 1024px) 100vw, 1024px" /