The latest TGIF community call opened with a relaxed tone, but Samuel ‘Sam’ Dare quickly made it clear that Covenant Labs has entered a period where focus and honesty takes precedence over casual conversation.
Bittensor is changing fast and every subnet is being forced to rethink survival, positioning, and long term value.
This call centered entirely on Covenant itself, its past choices, the new reality created by TAO flow, and the path forward.
Looking Back at the Early Covenant Vision
Sam recounted the earliest days of Covenant when Templar was first launched on the root network. The goal back then was straightforward. Reach 2% emissions, stay there comfortably, and enjoy the long cycle of rewards.
As Sam put it, “My get rich quick scheme was to get to 2% emissions and just milk it.”
Everything changed the moment dTAO arrived. He admitted that he was “terrified” when he learned that Jacob ‘Const’ Steeves was pushing dTAO through.
With this development, Covenant had to rebuild its thinking overnight. Yet the team survived the transition and learned from it.
Entering the New World of TAO Flow
The core of the discussion centered on the new emissions model. TAO flow introduces a harsher but more honest competitive environment. It rewards top performers and compresses emissions for others.
Sam described it as an intentional move by the network. “Emissions are a means to an end. They are not the goal,” he said. “If we over index on emissions, we fall into local minimal land.”
He explained that Bittensor is essentially one of the largest reinforcement learning systems ever created. dTAO was a new environment, and now TAO Flow has introduced even sharper edges.
While the shift feels painful, Sam emphasized that it is for the “greater good.”
Acknowledging Miscalculations and Choosing Evolution
Sam did not sugarcoat Covenant’s mistakes. “I will admit it. I am an engineer. I was brought up with proof of work,” he said. But running a subnet with a token economy is similar to running a public company. It requires fiscal discipline and long term strategy.
The new direction centers on two principles:
a. sustainability and responsible spending
b. continuous delivery without slipping into inertia
Covenant cannot simply rely on being early or playing the emissions game. It must ship real products, it had to do so.
Templar: From Milestone to Market
Templar (Subnet 03) has already delivered one of the biggest breakthroughs in Bittensor. Covenant, through this subnet, successfully trained a 72 billion parameter model over the public internet. This alone breaks the traditional cost model of artificial intelligence.
Sam explained why this matters. “If you can train over the internet and match centralized training, the market you can create is massive.”
Templar will now expand into real world use through:
a. training as a service.
b. collaborations with researchers and academic institutions.
c. presentations at DAI, London and NeurIPS Conference, San Diego.
d. exploratory work with programs such as Stargate.
Sam described Templar as Covenant’s “Mona Lisa,” a piece of work that needs to leave the lab and meet the world.
Basilica: Moving Past Early Mistakes
Basilica (Subnet 39) is undergoing a major transformation. Sam admitted that the earlier design over indexed on Lium and missed the deeper opportunity. He credited Erfan for being the person who had pushed the correct direction months earlier.
“Erfan told me to do this four months ago,” he said. “I was wrong to resist it.”
The new Basilica will become a cloud native compute network built for any job, anywhere. The strategy begins inside Bittensor itself by serving internal teams.
Sam stressed that internal users are more forgiving and help refine the product before wider adoption.
Grail: Refining Intelligence Through Post Training
Grail (Subnet 81) focuses on post training and acts as the second half of the intelligence pipeline after Templar. Sam offered a simple analogy.
“Pre training is like university. Post training is like graduate school.”
With both capabilities combined, Covenant can build training APIs, create tuned model variants, and offer fine grained control to researchers who want more flexibility than what automated systems provide.
Sam believes this space is wide open for Covenant to dominate.
Building a Balanced and Sustainable Team
Sam reiterated that Covenant must evolve into a research and product organization. The team is currently research heavy, with eight researchers and only two full time engineers. That will begin shifting slowly and responsibly.
At the same time, Covenant wants to expand its business development capacity. Sam admitted this is not his personal strength, saying, “I am a backend person. I am not a business development guy.”
Hiring must be careful, given the market downturn and limited capital flowing into subnet tokens.
Closing Message from the Host
Sam closed the call on a candid note. Many subnet communities are underwater. Narratives alone cannot sustain long term value. Only real milestones can.
He reminded the audience that Covenant Labs is fully committed to Bittensor for the long haul. “We are not here for noise. We are here to build things that last us and last our children,” he said.
The session ended with a simple message. Keep building.
Be the first to comment