From Pair Programming to Empowering Everyone: How Gate Embraces AI Development Models

Today, many people are unaware that I come from a technical background. Twenty years ago, in investment banks across the Asia-Pacific region, only about fifty people understood algorithmic trading, and I was one of them; at that time, building systems was not about "making them work," but about ensuring long-term reliability under constraints such as latency, stability, monitoring, and risk management.

For this reason, I was initially very cautious about "Vibe Coding," because it could easily imply that programming could be achieved by "issuing commands based on intuition." But after personally using Claude Code, Codex, and Copilot to create some small programs and internal tools, I realized that the issue is not AI writing code, but the deception that comes with the word: when AI is integrated into engineering processes, it resembles more of an extension of pair programming(Pair Programming), where "production" and "review" are combined into the same loop, allowing people to shift their focus from repetitive work to making key decisions and quality control.

At the same time, the reality is clear: it is very difficult for people who do not understand programming to achieve satisfactory results relying solely on Vibe; but in terms of capabilities, AI will significantly enhance efficiency and delivery quality. This explains why one of the most sought-after "advantages" in companies has become more LLM Tokens.

Gate’s stance on this matter is very clear: we do not see AI as a decorative toy, but as the foundation of next-generation productivity. Based on these irreversible trends, and in line with the prevailing internal consensus, we fully embrace AI and various Vibe Coding tools, encouraging colleagues to push boundaries, foster innovation, and truly integrate AI into core development and business innovation processes.

Most importantly, we are willing to invest real money to facilitate this path: the company invests millions of dollars monthly to improve accessible AI capabilities, controllability, and auditability, including more stable model access and guidance, a more comprehensive authorization and compliance framework, internal toolchains closer to business needs, as well as ongoing training and best practice accumulation.

The purpose of these investments is not just to "make everyone code faster" in a simple sense, but to make delivery more controllable, quality more consistent, and repetition more repeatable. In other words, what Gate wants is not just incremental individual improvements, but a systematic upgrade of organizational capabilities: integrating excellent engineering practices with AI acceleration capabilities, enabling more people to achieve higher standards at lower costs, and re-integrating risk management, code quality, and maintainability into the development process.

For me, the conclusion remains practical: don’t be fooled by "Vibe," and don’t demonize AI excessively. The best approach is to reframe AI within engineering methodology, making it a partner, a drafting tool, and an accelerator, while keeping responsibility and standards within reach. The reason behind Gate’s steady investment is our belief that long-term value does not come from "apparent speed," but from "sustainable and reliable speed." After all, in the worlds of finance and technology, working quickly is easy; but the professional difference lies in the ability to operate stably over the long term and to be trusted.