Recent development projects:

1. Rewrote the order placement engine of the exchange using RUST, achieving faster speeds with fewer resources by placing orders based on anchoring quotes and order book imbalance features. 24/7 continuous price updates without gaps. A single 8-core 16GB server can handle approximately 100 trading pairs simultaneously, reducing arbitrage opportunities on the exchange;
2. Reconstructed the market maker's financial system with complete reconciliation logic, daily profit and loss attribution, and detailed hedge transaction data that is auditable;
3. Developed a market maker order placement strategy for prediction markets, and conveniently set up a prediction market exchange that automatically lists the same markets pulled from Kalshi and POLYMARKET. Daily order placement is based on POLYMARKET quotes, with Kalshi quotes used as a fallback to prevent order quote failures caused by market data retrieval issues, which could be exploited by users;
4. A neutral factor strategy backtesting and live trading platform, regularly updating factors from research papers, tracking factor performance, manually adjusting strategy factor weights, and currently entering a small-capital live trading phase.
Among these projects, the first and third took the longest to debug. AI has clear limitations in designing high-frequency, high-concurrency, low-latency systems, mainly because it lacks fine-grained memory management, lock-free queues, pre-allocated memory pools, and other techniques to achieve nanosecond-level interactions. It prefers frequent database/file read/write operations, which ultimately produce severe system throughput bottlenecks and "I/O garbage," with exploding disk space. In the short term, backend system architecture engineers are still difficult to replace.
In frontend design, I prefer to use HTML to create an interactive interface first, then rewrite it with React + TypeScript, connecting the backend APIs. In the future, UI, frontend, and product managers are very likely to merge into a unified "product engineer." AI has already simplified component building and routine interaction logic for frontend pages, allowing one-click duplication. At this level, the core competitive advantage truly becomes "taste" and "business insight." A UI designer with excellent aesthetics and product thinking can directly lead product delivery using AI tools, making aesthetic scarcity an absolute premium that determines the upper limit of product visuals and interactions.