In 2025, the AI programming field experienced explosive growth. From simple code completion to multi-agent collaboration systems, AI has evolved from a supportive tool into an "intelligent partner" capable of independently handling complex software engineering tasks. A report by the Human-Centered Artificial Intelligence Institute (HAI) at Stanford University, titled "2025 AI Index," shows that AI systems' performance on the software engineering benchmark SWE-bench has significantly improved, with scores rising by 67.3 percentage points within one year, moving from low scores in 2023 to 71.7% in 2024. This marks the transition of AI programming from an experimental phase to a mature application stage.
Significant Improvement in Technical Performance
Over the past year, breakthroughs in AI on high-difficulty benchmarks have been remarkable. The report indicates that performance on emerging benchmarks such as SWE-bench has jumped from single digits to over 70%, with AI now able to independently read problem descriptions, locate bugs, modify code across files, and run tests. This "proactivity" and "executability" have become key focuses for major model providers. AI is no longer limited to generating simple loops but now possesses comprehensive capabilities such as context understanding and multi-task collaboration, significantly improving software development efficiency.
Funding Boom and Market Recognition
In 2025, funding for AI programming tools was extremely hot, with independent startups seeing rapid valuation increases. Among them, the AI code editor Cursor stood out particularly. Its parent company Anysphere completed a $2.3 billion Series D funding round in November, with a post-money valuation reaching $29.3 billion, growing more than 11 times in just 11 months. At the same time, the annual recurring revenue (ARR) of the tool surged from millions to over $1 billion in less than 24 months, setting a new record for B2B SaaS growth. This phenomenon reflects strong confidence from the capital market in AI programming tools, with this field being regarded as a high-growth emerging industry.
Thriving Development of Domestic Tools
In China, AI programming tools are also showing vigorous development. Local products such as Trae, an AI-native IDE, have performed well in local scenarios through deep optimization of Chinese semantic understanding and multimodal interaction (such as directly converting design drafts into code). Several tools have achieved full-process automation, supporting agent mode and project-level context processing, significantly improving development efficiency. With the iteration of domestic large models, the domestic AI programming ecosystem is accelerating its catch-up and forming advantages in certain areas, helping developers handle complex project requirements.
AIbase Insight: 2025 is a critical year for AI programming to shift from "assistant" to "main body." Technological breakthroughs, capital investment, and application implementation are driving the industry into a new stage. In the future, as agent systems become more mature, AI will reshape the software engineering paradigm, but it is also necessary to pay attention to the balance between security, compliance, and human supervision. AIbase will continue to track the latest developments in this field and provide professional insights for developers.
