📉 Scandal in $TAO and a -25% crash
The reasons for the crash, Covenant AI's exit, and how will this change DeAI?
Covenant AI - who built the largest permissionless LLM in history, praised even by Nvidia's CEO and Anthropic's Co-Founder - has officially left Bittensor along with all their projects:
Templar (SN3)
Basilica (SN39)
Grail (SN81)
🟢 The Covenant AI team was pulling up to 25% of emissions and triggered a massive rally. Over the past two years, they achieved the largest decentralized pre-training of a heavy LLM model in history built on Bittensor - Covenant-72B.
They proved in practice that training neural networks across untrusted internet nodes is real. Their departure is the loss of a flagship use case for the entire network.
🟢 Project founder Sam Dare released a scathing statement, accusing Bittensor co-founder Jacob Steeves of creating "decentralization theater."
Covenant AI's accusations:
🔴 Centralized Governance
- A unilateral decision by leadership to halt $TAO emissions for the Covenant subnet.
🔴 Censorship
- Revoking the Covenant team's moderation rights in their own community channels.
🔴 Sabotage
- Unilaterally shutting down the subnet's critical infrastructure.
🔴 Financial Pressure
- Targeted, large-scale token dumps by the founders during conflicts to suppress the price and exert economic pressure on the team.
Covenant claims the network isn't run by decentralized code, but by the personal will of one man. Because of this, they are packing up their models and research, and leaving to build somewhere else.
Bittensor = isn't 1 product, it's a network of subnets:
• every subnet = a market/model/niche
• they earn rewards in TAO
• and their tokens are tied to TAO through staking mechanics
If trust drops → TAO drops → ALL subnets drop
🟢 Amidst this, $TAO crashed by 25% (down to $250). The exact number of subnets that will voluntarily leave the network after Covenant is still unknown, but this precedent has caused genuine panic.
The fear that even a giant project can be manually "canceled" by leadership is forcing builders to look for alternatives.
All of this overlaps with the new, harsh mechanics of automatic Deregistration, which is currently forcing weak subnets into a massive liquidity "death spiral" and wiping them from the network.
⚠️ Short-term cons:
• A massive reputational blow to the narrative of Bittensor as a fair "free market"
• The departure of top-tier brains and a ready 72B LLM model to competitors
• FUD in the community, liquidity outflow, and a drop in market cap
✨ Long-term pros:
• Antifragility and exposing weak spots
• Any crisis reveals vulnerabilities
• The transition to dTAO was supposed to solve the centralization issue, but this scandal highlighted the risks
This will force the network to urgently overhaul its Governance towards full automation to regain trust.
Jacob Steeves responded to the conflict radically: according to him, this exit will cleanse the network and lead to the birth of the "first subnets that operate absolutely autonomously (headless) as true commodities" - without politics, community managers, or hard feelings.
🕯 Bittensor is going through its hardest and most painful growing pains. A perfect consensus doesn't exist yet, and governance is still subject to human flaws.
This stress test will either force the network to evolve into true decentralization, or slowly kill the ecosystem in favor of more transparent competitors.
Personally, I'm confident that everything is moving only for the better in the ecosystem; this is a plus for decentralization. Huge bets have already been placed by Smart Money into the ecosystem, and the necessity to evolve has been triggered.
One thing is certain - the community that trusted the Top Subnet, CovenantAI, experienced a RUG ☠️
Waiting for updates from Bittensor and CovenantAI.
Was Covenant’s exit pre-planned? What do you think, should Covenant have warned the community?
Subscribe ➡️ Crouton Digital Official | Website crouton.digital⬅️