Pearl Research Labs
@prlnet • 11,919 subscribers
¶ Pearl is a mathematical breakthrough that redefines the unit economics of AI.
Videos
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.
Pearl Research Labs54,136 views • 2 months ago
Today, the Pearl Protocol relies on INT-based matrix multiplication, which requires model quantization. To eliminate this overhead, we are developing our next-generation Proof of Useful Work to natively support floating-point computations. Launching soon, the initial release will support FP8, with FP4 to follow.
Pearl Research Labs24,335 views • 2 months ago
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.
Pearl Research Labs19,036 views • 2 months ago
Proof-of-Useful-Work forced us to invent a quantization format that did not exist. Pearl Protocol adds noise to both the weights and activations before matrix multiplication. That creates a constraint ordinary inference stacks do not have - the model values and the noise must fit together inside INT8. So we quantize the model to an effective W7A7 range, reserving one bit of headroom for the noise. The perturbed operands still fit in INT8, multiplication uses INT8 inputs with INT32 accumulation, and the low-rank noise can then be peeled off exactly, recovering the exact result of the quantized computation. Existing quantization for INT7 were unsatisfactory, so we built our own - a non-trivial fusion of GPTQ-style weight reconstruction, SmoothQuant-style activation smoothing and custom inference kernels. The result is an end-to-end W7A7 pipeline that, in our internal benchmarks, even outperforms many FP8 baselines (in quality and in efficiency). We are writing a paper detailing this technique so stay tuned. Looking forward, W7A7 is only an intermediate step. Our upcoming upgrade is designed to support FP8 workloads directly, eliminating the need for the extra integer-quantization step.
Pearl Research Labs15,604 views • 2 months ago
Proof of work has always demanded a brutal trade-off - to secure a monetary network, you burn energy on the scale of entire countries. Meanwhile, hundreds of billions poured into AI - all for one operation at massive scale - matrix multiplication. For years, these energy markets were disjoint. It was commonly believed that proof of work could never be implemented on top of real-world instances. They were wrong. The same matrix multiplication that powers AI can secure a monetary network as a by-product - every GPU cycle doing double duty. A currency backed by the most valuable computation on earth. A deep dive by Omri Weinstein ⤵️
Pearl Research Labs15,990 views • 2 months ago
Congrats to our partners Together AI 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 Vipul Ved Prakash and the team on the next frontier of inference: lower costs, better performance, and open models at real scale.
Pearl Research Labs14,131 views • 2 months ago
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? Rafael Pass breaks it down and shows why this critique doesn’t hold up. Worth watching 👇
Pearl Research Labs13,540 views • 2 months ago
No more content to load