CapComp | SN103

[CapComp | SN103 @CapCompAI]

Just two weeks ago, we took our first step as a subnet in Bittensor.

From the very beginning, our goal has been to push the boundaries of AI research, just like the other ML subnets that have emerged across the ecosystem.

To be completely honest, operating an ML research subnet comes with significant challenges. Many participants naturally gravitate toward subnets with immediate revenue potential, while staking power remains largely concentrated among established top subnets that have already secured a strong position in the ecosystem. The recently activated root weight mechanism has also, at this early stage, shown a tendency toward concentration among the leading subnets.

In such an intensely competitive environment, we did not enter the ecosystem by presenting already-proven groundbreaking research results and demanding recognition. Nor did we launch with a team backed by massive funding, a long track record, or a widely recognized reputation. Instead, we quietly took our first steps, driven purely by our vision.

At the beginning, we had only one investor who truly believed in our vision and made a firm commitment to support us. As a newly registered research subnet, we had to overcome the challenges of a high root proportion requirement and miner sell pressure from the very start. And we began this journey with a humble team that was built from the efforts of what was just a single miner.

However, the story has started to change.

The subnet that once seemed to have only one supporter has now built a small but incredibly meaningful community of supporters - something that would have been almost unimaginable for an early-stage subnet. Many top miners across Bittensor have joined us, and they have shared a commitment to work together toward the ultimate success of this subnet.

Furthermore, contrary to our initial expectation that research progress would move slowly in the early stages, the involvement and contributions of these top miners have already led to improvements beyond what we imagined. This has given us tremendous confidence and strength. It has allowed everyone who believes in and supports us to share the same vision: achieving success together.

Our goal is clear.

Based on the operational results and lessons learned during this initial phase, we are working toward building a truly meaningful and competitive IM design that can stand alongside centralized vendors in this field through measurable research performance.

As an early-stage subnet, we have identified several challenges within our current scoring mechanism. The final score does not yet have a sufficiently direct connection to represent definitive research performance in this domain. We recognize this gap, and we are actively working to address it.

Over the next two to three weeks, we will engage in serious discussions with multiple experts who have already joined our team, while also gathering feedback and insights from experts throughout the broader community. Our goal is to finalize and release an improved IM design that establishes a clear foundation for fair and meaningful competition.

In other words, we aim to make the subnet's evaluation score a direct and reliable indicator of research performance in this field - creating the foundation needed to compete confidently with centralized solutions.

Just as Templar, Targon, and now Minos have demonstrated, our entire team is committed to creating the most transparent, rigorous, and fair competitive environment possible. With the full support of our investors and community, we want to ensure miners receive meaningful rewards for their contributions and create another defining moment that demonstrates the power of Bittensor's core vision in the AI industry.

This is our goal. This is why we chose to operate a subnet on Bittensor.

Everyone who operates a subnet on Bittensor shares the same ambition: the dream and determination to make their subnet one of the leading subnets in the ecosystem and one of the projects that represents the future of Bittensor. They work tirelessly, often around the clock and without pause, because they truly believe in their vision and in the future success of Bittensor itself. The same dedication can be seen among miners as well. As Const simply expressed last Christmas, "There is no Christmas for miners in Bittensor" - and that statement could not be more accurate. The passion, commitment, and relentless effort of everyone building within this ecosystem are what continue to drive Bittensor forward. Their dedication is driven by the belief that something small and uncertain today can become something meaningful and transformative tomorrow.

A future that once existed only as a small dream and a precious vision is now gradually taking shape and becoming clearer.

From the very beginning, our team has been fully committed, and today we are working even harder with renewed ambition and determination toward the development of our new IM design.

To everyone in the community who believes in our project, supports us, and continues to stand alongside us, we sincerely express our deepest gratitude.

Thank you, @const_reborn , for giving me the strength and encouragement to keep moving forward.

We would also like to extend our heartfelt thanks to @ShizzyUnchained for directly supporting and promoting our subnet, and to @mogmachine , who became the first among major validators to allocate root weight to us.

CapComp will become another source of pride for Bittensor.

#Bittensor bittensor:native #AI #LoRA

[Openτensor Foundaτion @opentensor]

RedTeam is working toward integrations with three banking companies in the top 10% of the Fortune 500.

On the latest Novelty Search, @_redteam_ explains how #SN61 miners strengthen Innerworks’ cybersecurity products.

They also demo their Immune System, where AI agents created three new attacks from a miner’s submission and wrote the code to stop all three.

Hosted by @const_reborn

Full episode in the first comment

[Openτensor Foundaτion @opentensor]

https://t.co/zb16oEUkvd

[CapComp | SN103 @CapCompAI]

@_joncipher Thanks for the detailed breakdown.

[CapComp | SN103 @CapCompAI]

@ShizzyUnchained Thanks for the analysis.

[CapComp | SN103 @CapCompAI]

LoRA merging has proven it can meaningfully improve models.

As scoring nears its ceiling, submissions will no longer need to beat the previous top recipe.

New IM + research direction coming soon.

[const @const_reborn]

Announcing Teutonic-II-110B: a frontier permissionless training of a 110B LLM

We are scaling 10x beyond our previous model and 50% larger than any prior decentralized run.

[CapComp | SN103 @CapCompAI]

Mining is officially live.

20% miner burn is now open.
Time to compete. Let the best miners win.

We want to see even higher-quality submissions from miners going forward!

[CapComp | SN103 @CapCompAI]

CapComp is live on Bittensor.

V1 starts with LoRA Merger - miners compete to discover the best capability composition for real workflows.

The goal is bigger than merging LoRAs:
build an open market for AI capability engineering.