Multimodal, hardware, and AI Coding form the hottest landscape of the AI track in the second half of 2024. AI Coding, which was still unknown last year, has now become one of the few new stories in the AI application layer.
**TikTok’s algorithm leader plans to leave and start a business in AI Coding**
– **Chen Zhijie’s resume**: Chen Zhijie, the algorithm leader of TikTok, recently plans to leave and start a business, focusing on the direction of AI Coding. Chen Zhijie once served as the chief technical architect at Baidu and was responsible for advertising algorithms. In 2020, he joined ByteDance and successively took charge of the algorithms of domestic products such as Douyin, Toutiao, and Xigua Video, as well as TikTok’s algorithms.
– **AI Coding becomes a new trend for going overseas**: In 2024, AI Coding products such as “AI programmer” Devin, Cursor, and OpenAI o1 became extremely popular globally, and the AI Coding track was rediscovered. Alibaba Cloud and ByteDance respectively launched products such as Tongyi Lingma and MarsCode. Some startup projects have also emerged in this track, such as AIGCode founded by Su Wen, a former investor of China Growth Capital, and Ming Chaoping’s “Xinyi Yima” which started financing at a valuation of 50 million US dollars in the angel round. Although Chen Zhijie’s project is planned to be implemented in China first, the ultimate goal is to go overseas.
**Advantages of AI Coding’s development overseas**
– **Standardization lowers the threshold for scale**: Coding has no cultural or language restrictions and is easy to standardize. An investor believes that AI Coding products lower the threshold for scale.
– **Strong overseas model capabilities and high customer willingness to pay**: Model capabilities and downstream customer willingness to pay are key factors for the development of AI applications. The models with the strongest coding capabilities are all in the United States, such as o1 and Claude. Domestic models are relatively weak in understanding and generating code. After StackBlitz’s AI Coding product Bolt integrated Claude 3.5 Sonnet, its ARR grew from 0 to 4 million US dollars in just 4 weeks, and the number of weekly users exceeded 100,000.
**AI Coding is a good business**
– **Demand drives annual revenue growth**: Algorithm engineers are essential in all industries, and large models can improve programming efficiency. According to a survey report by Menlo Ventures, in 2024, the adoption rate of AI Coding-related products ranked first in major scenarios, reaching 51%. Downstream demand makes AI Coding a good business with an annual revenue of tens of millions of dollars. Anthropic’s software development and code generation business achieved a 10-fold increase in annualized revenue in nearly 3 months. Microsoft GitHub Copilot’s ARR is about 300 million US dollars, contributing 40% growth to Github’s annual performance. Cursor’s ARR reached 65 million US dollars in November 2024, a year-on-year increase of 6400%.
– **Productization ability is a key competitive factor**: Many AI Coding startup companies want to replicate this growth myth. Some AI practitioners believe that what AI Coding projects ultimately compete on is often not model capabilities but productization capabilities. Most entrants benchmark against Cursor and create Copilot products for natural language conversations. This tests the ability of product forms to integrate with customer IDEs and integrate into the programmer’s workflow. Strong productization ability is the core competitiveness of many Chinese AI startups that have a voice overseas, and it is also an important reason why investors are optimistic about the overseas expansion of domestic AI Coding projects.
