One of the most difficult problems to solve in on-chain derivatives exchanges is not about making order matching faster, but rather whether users still have sufficient information to retrieve their assets when the operator stops providing services. On October 23rd, Celestia Labs published the Derive case, which describes how this on-chain options platform continues to publish state difference data to Celestia within its V3 architecture, and coordinates this with Ethereum-based custody and proof of validity. The case study also lists the platform’s scale and throughput metrics; however, since it is a supplier’s introduction to a collaborative project, not all of the commercial forecasts contained within should be taken as independent audit results.

A company engaged in cross-chain payments often doesn't lack assets; rather, the assets are on the wrong chains. For example, Chain A has euro-stablecoins, while Chain B requires settlements. As a result, the operations team has to maintain inventory on multiple networks. On October 23rd, Circle announced that Circle Gateway would start supporting its Bitcoin product cirBTC and the euro-stablecoin EURC in addition to USDC, allowing developers to manage unified balances across supported chains separately according to their assets. This change occurred at the asset integration layer of Gateway. It cannot be said that all assets on the chains have gained the same liquidity, nor can it be claimed that cross-chain transactions no longer involve settlement or counterparty risks.

The U.S. Department of Labor released its weekly unemployment insurance report on October 22, which presented a set of figures with contrasting trends: As of the week ending October 17, the number of first-time unemployment claims, after seasonal adjustment, was 194,000, which is 6,000 less than the revised figure of 200,000 from the previous week; the four-week moving average was 197,750, a decrease of 1,000 from the previous week. However, as of the week ending October 10, the number of people receiving unemployment benefits continuously was 1.728 million, an increase of 25,000 from the revised previous figure. Although the weeks covered by these two indicators are not the same, they serve to remind the market that a smaller increase in new unemployment cases does not mean that those who have already lost their jobs will be able to find new jobs quickly.

After large models are applied to enterprise operations, the most common issue is not that the models cannot answer general questions, but rather that they do not understand the company's own product catalog, claims processing rules, or multi-round service processes. On October 22nd, Amazon Web Services announced that the "training session" in the serverless model customization of Amazon SageMaker AI was now officially available. Developers can write their own post-training loops, define loss functions, reward signals, and multi-round reinforcement learning processes; the platform is responsible for GPU distribution, distributed operation, fault recovery, and checkpoint saving. What this solves is the management burden of training infrastructure, and it does not decide for developers what tasks are worth training or what results are considered correct.

According to the Shanghai Securities News, on October 23, the first batch of 18 active ETF products were launched for sale. On the same day, E Fund, Morgan Fund, and Fullgoal Fund all issued announcements to end the fundraising process ahead of schedule. Reporters learned that on the first day, the total fundraising scale of these active ETF products had already exceeded 10 billion yuan. Market participants believe that such products not only fill a gap in actively managed tools within the market but also subject the public offering fund's active investment research capabilities to daily disclosure of holdings.