Proof-of-Useful-Work proposes that the same GPU computation used for AI workloads can also secure a blockchain network.
But that immediately triggers an economic objection - if miners are already getting paid for running AI jobs, does the security of the network collapse?
@PassRafael breaks it down and shows why this critique doesn’t hold up. Worth watching 👇
In his last paper, @PassRafael modeled the economics of proof-of-useful-work and found it lands in one of three worlds, set by a single efficiency parameter. Two of them beat plain Proof-of-Work.
His talk breaks it down:
Together AI @togethercompute just raised an $800M Series C at an $8.3B valuation.
It serves ¶Pearl-certified models as a live inference endpoint, and it's betting on cheaper open-model inference at scale, the same compute Pearl's proof-of-work is built to run on.
Pearl's premise, from the @prlnet team:
The scarce inputs behind modern AI are energy and data, and pearl-2:native is minted directly from the compute that fuses them. Running the model and mining the coin are the same operation.
July.
holidays? no, numbers.
@BNBCHAIN doesn’t stop for the summer.
- 65.2M $BNB burned so far, out of the original 200M supply.
at today’s price (~$550), that’s roughly $35.9B…
- 35 quarterly burns already completed.
- 134.79M circulating today.
- 34.79M left before hitting the 100M hard cap
the 36th burn scheduled for this month.
no pause, no slowdown, just steady deflation quarter after quarter.
BuildNBurn.
In a world where most GDP is produced by non-humans, the real currency isn’t money, it’s compute (data + energy).
The question that motivated @prlnet was whether it’s possible to create a currency directly from this scarce resource. A currency pegged to intelligence.
Our partnership with https://t.co/IwfRAocLPA (@togethercompute) showcases how Pearl changes the unit-economics of inference, and how it could help fund the AI buildout.
Couldn't have said it better than @vipulved.
Pearl’s math breakthrough is that Proof-of-Work can now run on top of arbitrary matrix multiplication AxB, with vanishing o(1) overhead.
The key difference: the miner choose (A,B).
This “2-for-1” technology means the hundreds of billions - soon trillions - of $$ in AI compute can secure Pearl’s PoW chain, effectively for free.
The result? the world’s most economical inference.
https://t.co/4urCcXuA4E
Congrats to our partners @togethercompute on their $800M Series C fundraising round.
Few teams have done more to make open models fast, accessible, and production-ready.
We’re excited to keep working with @vipulved and the team on the next frontier of inference: lower costs, better performance, and open models at real scale.
Since Bitcoin, a major open problem in distributed systems was whether proof-of-work consensus can be implemented on top of real-world computation rather than useless random hashing. This challenge of PoUW was repeatedly conjectured impossible by researchers and thought leaders. They were wrong.
Quantum readiness is not a theoretical “someday” problem anymore.
Last week, the White House signed an executive order requiring federal agencies to identify where they rely on quantum-vulnerable cryptography and start moving high-value systems to post-quantum standards.
For blockchains, this matters even more. (1/6)
The real currency is not money. It's energy and data - the two scarce resources whose fusion creates intelligence.
We believe that this dramatic shift in the production of knowledge compels us to rethink the fundamental properties, purpose, and creation process of money.
ÇÖZÜM BULMAK YERİNE CEO BENİ ENGELLEDİ !! @cz_binance e sallamayı biliyorsun ama
SPK AYKIRI OLARAK YAPILAN TUM REKLAMLARI (MAALESEF FENOMENLERDE CEZA ALACAK) HEM SPK YA HEMDE REKABET KURULUNA AVUKATLARIMIZ İLE BİRLİKTE İLETİCEZ görelim bakalım
Activation quantization sits on the critical path of LLM inference, so every microsecond matters. At @prlnet, we built a fused blockwise Hadamard quantization kernel that combines asynchronous memory management, intra-warp primitives, and other low-level GPU optimizations, making the Hadamard transform effectively free in production. A significant portion of the kernel was developed by an agentic tool we built in-house to automate CuTe DSL kernel development.