Recently developed projects: 1. Rewrite the order placement engine of the exchange with RUST, allowing for faster and fewer resources to place orders based on the imbalance between anchored quotes and order books. The 7 * 24 price does not jump short continuously, and an 8c16g server can simultaneously place orders for around 100 cryptocurrency pairs, reducing the arbitrage space of the exchange; 2. Refactored the financial system of market makers, with complete reconciliation logic, daily profit and loss attribution, and detailed transaction hedging data that can be audited; 3. In order to develop a prediction market making strategy for shop orders, a prediction market exchange was set up to automatically list the same markets from Kalshi and POLYMARK. Daily shop orders were placed based on POLYMARK quotes, and Kalshi's quotes were used as a fallback to avoid the failure of obtaining a single market situation during shop order quoting, which could lead to quote failure and user arbitrage 4. Neutral factor strategy backtesting and live trading platform, regularly updating factors from research papers, tracking factor performance, manually adjusting strategy factor weights, currently entering the small fund live trading stage. Among these projects, the first and third ones have been debugged for the longest time. AI has obvious flaws in the design of high-frequency, high concurrency, and low latency systems, mainly manifested in its inability to implement nanosecond level interactions through fine memory management, lock free queues, and pre allocation of memory pools. It likes frequent database/file read and write operations, ultimately resulting in a "I/O poop mountain" that seriously slows down system throughput and an explosion of hard disk space. In the short term, it is still difficult to replace backend system architecture engineers. In terms of front-end design, I am accustomed to using HTML to draw an interactive interface, then rewriting it with React+Typerscript after tuning, and connecting it to the back-end interface. UI、 The front-end and product managers are highly likely to merge into a unified 'product engineer' in the future. AI has completely simplified the construction of front-end page components and conventional interaction logic, allowing for one click replication. At this level, the true core competitiveness has become "taste" and "business insights". UI designers with excellent aesthetics and product thinking can directly lead product delivery with the help of AI tools, and the scarcity of aesthetics becomes the absolute premium that determines the visual and interactive limits of the product.