Watching Subnet 2 push Verifiable AI forward is exciting. Over 2.2 billion verified proofs on-chain proves zkML can scale. Builders, agents, and inference apps now have verifiable oracles powered by zero-knowledge circuits. #zkML#Bittensor#VerifiableAI#Subnet2
2.2 Billion verified proofs on-chain. That’s not just a milestone it’s proof that #VerifiableAI is scaling.
Subnet 2 is powering the next generation of Inference, Agents, and Builders with verifiable oracles through zero-knowledge circuits.
#zkML#Bittensor#VerifiableAI#Subnet2
This weeks Livestream report will be on Friday, July 17, 2026 here on X, LinkedIn, Facebook & YouTube!
https://t.co/kil40IAqQR
We will also be hosting our weekly Game Day live with our Grand Championship Round! This will be a don't miss live event!
Calendar it today and LFG!
I can see applications in industries like civil aviation and industrial manufacturing, where accuracy, accountability, and traceability are essential. Looking forward to seeing how this technology evolves. @SertnAI https://t.co/RiAzNtBhcD
One challenge with AI in regulated industries is trust. A prediction alone isn't always enough when multiple organizations, auditors, or regulators need to verify how a decision was made. Transparent, verifiable AI will become increasingly important as adoption continues to grow.
I like the idea of cryptographic proofs acting as mathematical receipts. They securely connect the model version, input, and output into a tamper resistant record, while staying compact at under 500 KB, making verification practical rather than burdensome
Enterprise AI is moving past pilot theater. The harder phase is production, where outputs affect workflows, audits, contracts, and operational decisions.
Sertn is built for that layer: not just running AI, but preserving verifiable records of what happened.
1/ AI-driven science is not just about making research faster.
The more interesting shift is that models are starting to become part of the reasoning environment itself, helping decide what gets tested, what gets ignored, and what becomes “promising” enough to pursue.
Verifiable AI moves closer to adoption when verification becomes efficient.
DSperse shows that targeted proofs can deliver trust without the burden of full model verification.
Huge step forward by @inference_labs 🚀
Trust is becoming one of the most important topics in AI. That’s why I’m paying attention to @inference_labs and their work on making AI outputs verifiable without excessive costs. 🚀