与「OPTIMUM」相关的搜索结果

据官方披露，Optimum 主网 10 月 20 日上线以太坊，1kx 领投 1100 万。 16 家节点商、2.2 万验证者密钥接入，140 万 ETH 关联质押，传播快 2-4 倍。 999 张 TurboCard 除 gas 免铸，官方明说非代币承诺，别上头。 #OPTIMUM#

Following Korea Blockchain Week (KBW), the @get_optimum team is traveling to Singapore for another busy series of engagements. On October 6, they will partner with @ambr_io to co-host an exclusive event focused on agentic trading, bringing together builders and researchers to explore automated, AI-driven strategies and decentralized market infrastructure.

We talk about blockchain TPS all the time. Higher throughput. Faster execution. More transactions per second. But how fast does data actually move across the network? Blockchains can indeed process transactions quickly. However, the resulting data still needs to be transmitted to various nodes scattered throughout the network. Blocks, transactions, attestations, blobs, and other payloads must reach the nodes that require them. Instead of trying to boost TPS directly, @get_optimum focuses on the propagation layer specifically with mump2p, which employs RLNC as part of its approach to moving data between nodes. Because blockchain performance isn't just about how much data can be processed; It’s also about how efficiently that data can be moved.

Today’s internet is actually full of data, but that data is still scattered across many chains, applications, and systems. The problem isn’t just how to store data, but how to enable various environments to access and use it together. @get_optimum introduces DeRAM as a unified memory layer to connect fragmented data. That way, apps, AI agents, chains, and other systems can operate with a more consistent state. From isolated data → shared memory → a more interconnected internet.

Blockchains don't just need faster execution. They need faster ways to move blocks, transactions and blobs between nodes. That's the problem mump2p is built around. Move the data faster. @get_optimum

Ethereum clients don’t need to be rebuilt to move data faster. Run mump2p as a sidecar: fragment block data with RLNC, forward useful fragments immediately, and let the client keep doing what it already does. Standard flow: receive full block → reassemble → forward. mump2p flow: fragment → forward → process on the fly. The goal isn’t just faster blocks. It’s permissionless connectivity, optimized data flow, and validator-grade performance that isn’t reserved for the most powerful machines. Ethereum Client → mump2p → High-speed Transport Network @get_optimum

♾ Optimum Pulse: September 21 - 27, 2026 🔻🔻🔻 1. "Must-watch" Podcast @get_optimum CEO Prof. @MurielMedard explains on A Global Tech Podcast how to optimize latency in DeFi and why a hybrid approach between TraFi and DeFi as most likely. 🔗 Full interview with @AGTPinsights (with correct timestamp): ---- 2. ETHSofia 2026 CPO @sajidazouarhi attended the panel discussion Building Under CROPS: What the Ethereum Foundation's Mandate Means for Developers and Institutions alongside representatives from European Ethereum Institute, Ethereum Foundation and 0xbow. 🔗The full recording will be available soon ---- 3. APAC Tour - Part 2 Save the Optimum team's schedule this later September and early October 👇 🇰🇷 Korea Blockchain Week: 📍 Sept 28 - Ethereum Korea One Genesis 📍 Sept 29 - signature CODED@KBW event 🇸🇬 Token2049: 📍 Oct 6 - co-hosting an event with AMBR 📍 Oct 7 - Digital Asset Summit (DAS) 2026 ---- 4. Weekly Role Upgrade - Refined Batch 29.

♾ Optimum Pulse: September 14 - 20, 2026 🔻🔻🔻 1. Post-Quantum Security CEO Prof. Muriel Medard wrote an article for CoinDesk about the path to quantum secure blockchains. 🔗 ---- 2. New mump2p testing partner @get_optimum × Twinstake 🇬🇧 Let’s see how mump2p’s ultra-fast data propagation can push Twinstake's MEV APR further. 🔗 ---- 3. Meet the Optimum team this month Later this September, in Korea 🇰🇷 📍 Sept 28th - Ethereum Korea Conference 📍 Sept 29th - CODED@KBW, co-hosted by Optimum, Google Cloud, Four Pillars, LINE NEXT, Auros. 🔗 🔗 ---- 4. Careers Optimum is hiring a Crypto Markets & Trading Lead 🔗 ---- 5. Mystery Update The newest number is 894,600. Admin Jeff hinted that everyone will be surprised with those numbers. Anw, I'm surrender 🏳️ 🔗 ---- 6. Community Call Community Call with Jeff is back. The full recap is available 👇 🔗 ---- 7. Weekly Role Upgrade - Refined Batch 28.

Thoughts About Scaling Law Scaling, but not only of parameters. Every model release now ends with the same question: how many parameters? It isn't a question that can be answered on its own. Parameter count is only meaningful alongside three others — how much data you have, where you intend to spend your compute, and who will run the model, under what conditions. The field learned this the hard way. Kaplan et al. (2020) fit an exponent that told everyone to grow parameters faster than data — roughly 2.7:1 — and the industry complied: GPT-3, Gopher, MT-NLG. Hoffmann et al. (2022) redid the experiment across four hundred models and found the compute-optimal split is closer to 20 tokens per parameter, and that with sufficient compute the two should grow at the same rate rather than drifting apart. The error in the earlier fit compounded with every order of magnitude of compute, which is why the largest models of that generation were the most misallocated. The trillion-parameter round was, in retrospect, a detour the whole field took together and then reversed. Chinchilla wasn't the end either. It optimized training compute for models that would be trained once and evaluated. Today a model is called billions of times a day and inference dominates lifetime cost. Put inference into the objective and the optimum moves toward smaller models trained far longer — deliberate over-training, which is what Llama-2-7B and Gemma-2-9B were doing at roughly 290 and 889 tokens per parameter. Sparsity moved the target again. In a MoE model two quantities have to be kept apart: total parameters govern roughly how much the model can hold — knowledge, facts, the long tail — while activated parameters and effective depth govern roughly how far it can think, how many steps of a causal chain it can carry before it comes apart. A dense 20:1 ratio does not transfer. And the ratio isn't a single number at all: Roberts et al. (2025) find the optimal tokens-per-parameter is task-dependent, with memorization favoring more parameters and reasoning favoring more data. Follow-up work on MoE observes that at fixed TPP, pushing total parameters higher actually degrades reasoning, while activating more experts reliably helps it. This matters for what we are building toward. Finding a vulnerability is not a retrieval problem. It doesn't come from having memorized more CVEs; it comes from carrying a twenty-step chain of inference to the end without losing the thread. That capability does not live in total parameter count. Which brings us to this release. Total parameters appear to matter up to a threshold — enough to hold the world — after which additional capability comes from scaling elsewhere: effective depth per forward pass, and above all post-training. GLM-5.3 is our controlled experiment on that claim. Same base, same architecture, same total and activated parameters as GLM-5.2. One month of scaling long-horizon environments and RL. The gains are not marginal. Well, scaling has more than one dial. We turned the post-training one this time because it had the most slack left in it — not because the others are finished. Base model size, pretraining data, compute spent per forward pass: all of them are still on the table, and we will come back to each. What this experiment taught us is that the dials do not have to be turned together, and that the one worth turning next is rarely the one that was worth turning last. We are not done scaling. Next time, maybe mid-training, pre-training, and even more.