Pearl

Pearl Research Labs published its whitepaper in April 2025, introducing cuPOW, a proof-of-useful-work consensus replacing hash mining with verifiable matrix multiplication for AI.

On April 27, 2026 Pearl's mainnet went live as a Bitcoin-forked L1 using PoUW, with a fixed 2.1 billion PRL supply and no pre-mine or VC funding.

On May 15, 2026 AI cloud provider Together AI announced an exclusive partnership, launching a discounted Pearl-powered inference endpoint tied to network compute.

In June 2026 PRL began trading on centralized exchanges including CoinEx, BigONE and SafeTrade, broadening market access for the token.

In July 2026 Pearl Research Labs joined NVIDIA's Inception program, building an accelerated inference platform for Hopper and Blackwell GPUs.

The Pearl (PRL) mainnet launched on April 27, 2026. It is a Bitcoin-codebase L1 that replaces hashing with GPU matrix multiplication (called 'proof of useful work'), with a max supply of 2.1 billion coins, no pre-mine and no team or VC allocation; all supply is released through mining. In May, Pearl partnered with Together AI to launch a Gemma inference endpoint priced over 25% below market, with the discount subsidized by PRL mining issuance.

Price and supply: PRL rose from about $0.28 in August to $2.10 on October 14 (up 228% in a month), with a circulating market cap of about $700 million, but it is listed on only two small exchanges. Roughly 1.01 million new PRL are issued daily, and about 305 million tokens will be mined over the next 12 months, nearly doubling circulating supply. Controversy: in June, independent research found that the dominant mining software contained no AI code and that random data could still earn shares, which the co-founder acknowledged permits 'spam mining'; a testnet upgrade aims to cut inference overhead from 10-15% to 0-3%. Pearl was among the most-talked-about projects during Token2049; the project has no buyback or burn mechanism, and its value rests on the future AI-compute settlement narrative.

Pearl launched its Proof-of-Useful-Work (PoUW) Layer-1 mainnet on April 27, 2026, with a block interval of about 194 seconds and a native token PRL capped at 2.1 billion. PoUW lets miners run matrix multiplication (the core math of neural networks) to secure the network while producing AI computation. Progress during the year: on May 15 it partnered with Together AI to launch a Gemma-4-31B-it-pearl inference endpoint priced over 25% below market, subsidized by PRL emissions; on June 12 the MoE hard fork went live, upgrading certificates from v1 to v2 built around Mixture-of-Experts workloads; on September 14 it released an FP8 floating-point PoUW scheme (roughly doubling throughput) and launched the OpenAI-compatible Pearl Inference API, and partnered with Attestable to build an AI trust layer. Future strategy and roadmap: Pearl's roadmap must answer a question academics have raised directly: consensus can verify that matrix multiplications were computed but not whether they served paying customers or random matrices, so usefulness has to be settled by economics rather than the protocol. Its answer is to let inference providers earn both customer revenue and PRL rewards, undercutting pure miners' cost line, then convert the subsidy into real calls through a Together AI Gemma endpoint priced more than 25% below market, with Attestable's privacy proofs making enterprises willing to route inference there. The risk is equally clear: inference is billed in dollar prepaid credits, and no payment or burn loop yet links PRL to real compute demand, so without call volume the network remains a miner wearing AI clothing.

Omri Weinstein is co-founder and CEO of Pearl. He holds a bachelor's from Tel Aviv University and a PhD from Princeton in computer science, was a postdoc at NYU's Courant Institute, has taught at Columbia since 2017 and is now an associate professor at Hebrew University. He was a researcher at NVIDIA and chief scientist at VAST Data, wrote the useful proof-of-work paper in 2025 and founded Pearl Research Labs.

Ilan Komargodski, a co-founder, holds a PhD in cryptography from Weizmann, researched at Cornell Tech and NTT Research and now teaches at Hebrew University; he co-authored the founding paper and leads zero-knowledge proofs and post-quantum signatures.

Idan Sugarman, the third co-founder, comes from business and investing, is a partner at Zero Knowledge Ventures and co-founded the AI healthcare company QuantHealth, handling organisation and commercialisation. Weinstein owns the research agenda, Komargodski runs the cryptography group, Sugarman runs people and operations, drawing researchers from Hebrew and Weizmann doctoral circles.