When AI companies like OpenAI or Anthropic build models like ChatGPT or Claude, they don’t just train the model once and ship it. They spend months doing something called “post-training”, teaching the model to actually be helpful, follow instructions, and give good answers instead of just predicting text. This post-training is expensive, secretive, and controlled…
For a long time, “Decentralized AI Training” was viewed as a science experiment, cool in theory but impossible in practice. The narrative was simple: “You can’t train massive AI models across the internet. It’s too slow.” But this week, that narrative shifted. Jack Clark, co-founder of Anthropic (the company behind Claude AI) and former policy…
For years, the final stage of AI training has remained quietly centralized. After a model is pre trained, the most important work still lies ahead. This final phase, known as post training, is where models learn to reason better, follow instructions more reliably, and perform well on complex tasks like mathematics, coding, and logic. Reinforcement…
SUMMARY: Covenant Labs delivered the 17th episode of TGIF at the NeurIPS conference in San Diego, highlighting major progress across Templar, Basilica and the broader Covenant 72B initiative. The brief emphasized how Templar unlocks decentralized pre-training at a scale traditionally reserved for hyperscalers, earning strong interest from researchers and industry partners. Basilica’s relaunch also introduces…
SUMMARY: In a TGIF community brief held on October 24, Covenant Labs shared major progress across their full AI pipeline on Bittensor. Templar continues to lead decentralized pretraining, outperforming centralized baselines and scaling the largest live distributed model. Grail is now showing successful post training results on mainnet, while Basilica prepares a shift to real…
