Pearl started with a simple question: Why should AI and blockchain mining compete for the same GPUs when the same computation could do both? In this talk, Pearl co-founder Ilan Komargodski explains how that idea became a live proof-of-useful-work network turning AI inference into blockchain security with minimal overhead.

@komargodski The total network computing power has reached 36E again—can your team actually deliver?

@komargodski The idea of turning AI inference into useful blockchain security is pretty compelling — same GPUs, useful computation, and network security instead of competing for resources. That’s a fascinating shift in how proof-of-work could work. 🔥

@komargodski The interesting part is turning the same GPU into an asset serving two markets. Mining secures the network. Inference puts the compute to work. That’s the shift pearl-2:native is testing.

@komargodski We cant mine anymore

Impressive! We should talk. I see big potential.

@komargodski I called it PoUI Proof of Useful Inference, @ganzoai. Ganzo Agentic Blockchain with PoUI/PoA and quantum resistant encryption...

@komargodski Been wondering when someone would build this. GPUs doing double duty instead of competing for the same hardware... question is whether overhead stays minimal at scale.

@komargodski Exploited again

@komargodski Same thesis, different verification path. TensorCash makes proof-of-work BE LLM inference, and blocks are checked by replayable proof-of-inference: statistical replay of sampled tokens confirms which model produced an output. Not TEE, not zkML. Verification stack is Apache-2.0.

@komargodski Look at the computing power of the k pool. You need to continue to fix the loopholes

@komargodski efficiency is finally winning

@komargodski Exploiter is stealing our coin, do something !!!

@komargodski Please pay attention! Prevent the cracking of computing power, and the stand-alone machine can be 80p.