Vitalik proposed an "ultra-compact chain" solution: validators submit STARK proofs daily, and state storage is compressed to 6 bytes.
Related
Vitalik Buterin released a new proposal suggesting that the consensus layer state be drastically reduced to only 6 bytes.
According to Foresight News , Vitalik Buterin has released a new proposal, "The Extremely Lean Chain." The core of this proposal is to significantly simplify the state stored by each validator in the current Beacon chain (public key, withdrawal certificate, activation/exit epoch, etc.), retaining only 6 bytes (1 byte for effective balance + 5 bytes for deposit tree index). Validators will need to track their deposit Merkle proofs themselves and asynchronously update their effective balance and participation records using daily generated STARK proofs, instead of having these updates processed in real-time on-chain every epoch. The proposal is planned to proceed in multiple phases: removing public key storage and replacing it with a deposit tree index, with daily STARK proofs used to update the balance; subsequently, a daily public key refresh mechanism will be introduced to achieve strong anonymity of validator identities, while naturally supporting Single Secret Leader Election (SSLE). This design is considered feasible in the context of Lean upgrades such as single-slot finality (SSF) and recursive STARK, aiming to significantly reduce consensus layer state overhead and computational burden, support the number of validators to the millions, and provide strong privacy protection in advanced versions.
SemiAnalysis: NVIDIA Delays or Adjustments to Multiple AI Rack-Scale Architectures, Limiting Rubin Ultra's Expansion Path
According to Mars Finance, on July 6th, SemiAnalysis reported that NVIDIA's latest rack-mount interconnect architecture, Kyber NVL144, has undergone a major adjustment just three months after its release, being delayed by more than 12 months to 2028. The main reason is the ongoing challenges in PCB planar design regarding manufacturing feasibility. Simultaneously, the NVL72x2 back-to-back rack architecture has been cancelled. This solution was originally intended to enhance the scalability of pure copper NVLink by deploying two Oberon racks back-to-back, but its complex structure and high operational burden on hyperscale cloud providers (CSPs) led to strong market skepticism and its eventual abandonment. Furthermore, due to the immaturity of CPO (Co-packaged Optics) technology, NVIDIA's larger-scale scaling solutions based on CPO NVSwitch (such as NVL576) may also continue to be delayed, or limited to small-batch trial production. This means that NVIDIA lacks a stable large-scale scale-up solution until CPO matures. Furthermore, the Rubin Ultra product roadmap has also changed: the originally planned "four-computing-chip" version has been cancelled, with only a "dual-computing-chip" version remaining, and the overall system-level performance is expected to drop to about half of the original solution. These adjustments mean that Nvidia's scale-up capabilities in the Rubin Ultra generation are limited. With the CPO NVSwitch unable to be deployed before the Feynman architecture, competitors such as the AMD MI500X or Google TPU v8i may have a relative window of opportunity in terms of scaling capabilities for large-scale training clusters. Meanwhile, Nvidia is expected to fill market demand during the product transition period and maintain overall supply chain pacing by shipping large quantities of Oberon Rubin racks and their "Ultra" versions.
22 companies filed for Hong Kong Stock Exchange listings in one week: AI semiconductor supply chain dominates the headlines, with star companies like Avita among them.
Mars Finance reported on July 5th that, according to disclosures by the Hong Kong Stock Exchange (HKEX), from June 29th to July 4th, a total of 22 companies submitted IPO prospectuses to the HKEX, including Shengwei Times, Liante Technology, Shengjing Network, Yangtuo Holdings Inc., Longxun Semiconductor, Haote Energy Saving, Tongao Testing, Daotong Technology, Keli Limited, Avita, Wubo Technology, Jiaxuan Intelligent, Oulin Biotechnology, Yuanhai International, Silicon-based Flow, Huaneng Design Institute, Senyi Intelligent, Joint Stock Company National Company Kazakhstan Temir Zholy, Liding Semiconductor, Jingze Biotechnology, Hongxinyu, and Jingwang Electronics. Among the 22 companies that submitted their prospectuses, 5 A-share listed companies simultaneously applied to the HKEX, triggering a new wave of "A+H" dual listings. Among them, Jingwang Electronics leads with a market capitalization of 70.949 billion yuan. This leading PCB manufacturer's revenue exceeded 10 billion yuan each year from 2023 to 2025, with a compound annual growth rate of 46% in profits over three years. The company has significant related-party transactions with Luxshare Precision, and the demand for PCBs driven by AI computing power is its core growth logic. Liante Technology is a leading optical module manufacturer from Wuhan and one of the few global suppliers with vertical integration capabilities for 800G/1.6T optical modules, with a latest market capitalization of 38.764 billion yuan. Daotong Technology, Orin Biotech, and Longxun Semiconductor are from the automotive intelligent diagnostics, human vaccines, and high-speed video chips sub-sectors, respectively, and are all star companies on the Science and Technology Innovation Board, with latest market capitalizations of 17.418 billion yuan, 17.319 billion yuan, and 11.562 billion yuan, respectively. All three companies appeared as "spin-off subsidiaries." Wubo Technology is a smart logistics supply chain service provider for ferrous bulk commodities under Yankuang Energy. Based on 2025 Gross Transaction Value (GTV), it is the largest smart logistics platform in China's steel and metallurgical industry. Tongao Testing, spun off from the Hong Kong-listed Anton Oilfield Services, ranks second among professional testing institutions in China's oil and gas industry's integrated TIC solutions market and is a leading independent oil and gas testing TIC service provider. Liding Semiconductor, backed by global PCB leader Zhen Ding-KY, mainly produces high-end IC substrates such as FCBGA and FCCSP. Based on 2025 revenue, it ranks third among IC substrate manufacturers in mainland China. Of the 22 companies, AI and semiconductor-related companies account for nearly half. From upstream chip design (Longxun Semiconductor), memory (Hongxinyu), and IC substrates (Liding Semiconductor), to midstream optical modules (Liantech) and PCBs (Jingwang Electronics), and then to AI infrastructure (Silicon-based Flow), the entire AI industry chain is represented. Among them, Silicon-based Flow has applied for listing under the Hong Kong Stock Exchange's Chapter 18C Special Technology Companies Rules. Established only 34 months ago, its valuation has increased more than 33 times in three years, reaching a latest valuation of 7.74 billion yuan. Alibaba, Huawei, Meituan, and Innovation Works are all shareholders. This time, it aims to become the first "AI Token Factory" listed in Hong Kong. Hongxinyu's revenue exceeded US$1 billion and net profit exceeded US$500 million in the first four months of 2026, representing a more than 30-fold increase in net profit. Based on 2025 revenue, it will be the fifth largest independent memory manufacturer globally and the second largest in mainland China. Multiple high-profile targets from various sectors are simultaneously emerging. In the new energy sector, Avita is a rare Hong Kong IPO target in the domestic high-end new energy vehicle field. The company is jointly established by Changan Automobile, CATL, and Huawei. It achieved operating revenue of 5.645 billion yuan, 15.2 billion yuan, and 25.6 billion yuan in 2023, 2024, and 2025 respectively, achieving a leapfrog growth in revenue over three years; however, its losses are also increasing, accumulating to approximately 11.2 billion yuan over three years. In the AI healthcare field, Senyi Intelligent, as China's largest provider of intelligent hospital healthcare solutions, has served over 800 hospitals. Its shareholders include Tencent, Sequoia Capital, and IDG. Meanwhile, Kazakhstan Railways, controlled by Kazakhstan's sovereign wealth fund, is a state-owned transportation and logistics giant operating a vital land transport corridor connecting Central Asia with China and Europe. Its IPO in Hong Kong, as a key infrastructure company in a Belt and Road Initiative country, has attracted widespread market attention. (Science and Technology Treasure News)
Venus proposed shutting down 87 deprecated markets across multiple chains, requiring users to withdraw by June 29th.
According to Foresight News , decentralized lending protocol Venus Protocol has released a proposal to integrate 87 abandoned markets across 8 chains, including BNB Chain, Ethereum, and Arbitrum. The proposal will be implemented in two steps: first, the reserve ratio will be increased to 100%, and borrowing rates will be significantly raised to encourage users to exit; second, the collateral factor and liquidation threshold for the relevant markets will be reduced to zero during the week of June 29th, at which point related positions will face liquidation risk. Venus strongly advises borrowers or depositors in the aforementioned markets to close their positions or withdraw their assets before this date.
Solana has implemented on-chain governance, setting the entry threshold at staking 100,000 SOL tokens.
According to Foresight News , CoinDesk reports that Solana has enabled on-chain governance. Its GitHub repository shows that the system, for the first time, grants its validators and token holders the right to directly and documentedly vote on the network's future. Submitting a proposal requires staking 100,000 SOL (approximately $7.7-7.8 million based on an SOL price of $77-78). A proposal needs 15% active staking support to enter the voting stage, and two-thirds of the staking votes are required for passage. The results are recorded on-chain. This system empowers delegators, ordinary users who stake their SOL with validators instead of running nodes and collecting staking rewards themselves. Delegators can now overturn validator votes or cast their own vote if a validator abstains, with all votes weighted by the delegator's own stake. The Solana Foundation calls this "stakeholder sovereignty," aiming to retain true voting power in the hands of the actual token holders, rather than entirely relinquishing it to the validators they delegate to.
Solana has launched an on-chain governance mechanism, requiring proposals to receive 15% staking support before they can be voted on.
ChainCatcher reports that the Solana Foundation has officially launched its on-chain governance mechanism, Solana Governance Proposals (SGP). Validators can now submit, support, and decide on core protocol decisions through SGP. All proposals are conducted on-chain, using a staking-weighted voting mechanism and verified via Merkle proofs. Any validator with at least 100,000 SOL delegates can initiate an SGP. Unlike Solana Improvement Documents (SIMD), which focus on technical and protocol changes, SGP primarily serves to express opinions on ecosystem governance. Proposals must receive at least 15% network staking support before entering the formal voting phase.