..v..v.eth

The Ultimate Templar One-Pager 🫧 Training a large language model requires a centralized cluster, thousands of GPUs. MILLIONS of dollars. What if it worked like Bitcoin ? Decentralized. Permissionless. Open-Source. No other protocol has succeeded to fulfill such a wild dream. Until $SN3 made it a reality on $TAO. What is Templar ❓ #SN3 I Templar is Bittensor’s Subnet 3 a decentralized framework for training large language models at internet scale. Instead of one company owning a cluster, independent miners worldwide contribute GPU compute over commodity internet to collaboratively train a shared model. The network rewards useful work, punishes bad actors, and produces something no centralized lab has a monopoly on: intelligence, coordinated by incentive, not capital. Miners train the model on assigned data subsets, compute gradients, and share compressed updates. Validators then evaluate these contributions by measuring actual improvement in model performance before assigning scores and rewards on-chain. The result is a closed loop where every dollar of compute produces a verifiable, auditable improvement to the model not a promise. Why Templar ?💡 • The first ever protocol to train a 72B LLM permissionlessly. • Decentralized pre-training with zero centralized cluster dependency • SparseLoCo algorithm: 146x gradient compression, enabling training over commodity broadband • Crusades competition system: miners compete on code efficiency, not just raw compute • 100% of commercial training fees go to token buybacks • 100% decentralized AI stack: SN3 (pre-training) + SN39 Basilica (compute) + SN81 Grail (post-training/RL) • 2x papers accepted at NeurIPS 2025 only Bittensor subnet to ever reach a major ML conference. Who’s paying attention 👀? Templar’s achievment didn’t stay inside crypto. • Jack Clark, Anthropic’s co-founder and policy head, publicly cited Templar as the largest active decentralized training network operating at 9×10⁷ FLOP/s scale. • NVIDIA CEO Jensen Huang, acknowledged and praised the decentralized AI training approach, during the All-In Podcast. • Grayscale has filed a TAO ETF application w/ Bittensor’s top 10 subnets have collectively reached a $550M combined valuation. • DCG established Yuma Asset Management specifically for subnet-level exposure. Polychain Capital holds an estimated $ 200M TAO position. The Tokenomics 🪙 - Marketcap : $ 105M - Circulating supply : 4.2M - Max supply : 21M Conclusion: As @Jason said $TAO > $BTC For me Bittensor’s subnets alone, especially templar & chutes are superior to almost every other crypto AI protocol in existence today. Chads who saw Bittensor & Templar’s potential early on🧠: @Route2FI I @scoutesy I @tseutseutao @tylerdurdeth I @DreadBong0 @2xnmore I @RvCrypto I @ElonTrades @KakashiTao I @CryptoZPunisher @crypto_ideology I @thejbullmarket @TaoSubnetMaxi I @arnauramio @BitStern I @keenandotai @cryptoblockcats I @JesusMartinez @CryptoWizardd I @CryptoTomYT @au_xbt I @AustinBarack I @Old_Samster @Karamata2_2 I @OnlyHades_