AI Lens Series Research: Unconventional Report on AI Coding

The programming paradigm revolution beyond software

For a long time, programming has been defined as a rigorous, logic-driven activity, a process of translating human intentions into deterministic instructions executable by machines. However, AI is overturning this core definition, elevating programming from the act of "Coding" to a higher dimension of "expressing intentions" and "realizing visions."

In 1995, Bill Gates predicted the arrival of the "information superhighway" in The Road Ahead. At that time, it was hard to imagine that everyone could connect to the whole world anytime and anywhere today. Now, we are standing at a new turning point. This is not just a technological upgrade or a tool improvement. Software development is undergoing a paradigm-level transformation, which is also a redefinition of three fundamental questions:

What is programming?
Who can program?
How will the digital world be built and consumed?

Different from the information superhighway, this paradigm shift is moving from imagination to reality at an amazing speed and with a high level of consensus. At AI Ascent 2025, Sequoia Capital placed the present and future of AI Coding from a more forward-looking perspective, predicting that we are entering an "age of abundance." Code is the first market to be disrupted, and this disruption will be a preview of the "age of abundance." It also believes that the continuous evolution of Coding Agents is worth looking forward to because this will not only reshape the entire software industry but also become an important precursor to the AI transformation process of other industries in the future.

This means that when we focus our multi-layered cognitive lens on product, model, value, investment and financing, commercialization, and competitive strategies on the rapidly evolving AI Coding scene, from the internal research data and CEO interviews of leading companies such as Microsoft, Google, Meta, Amazon, and Salesforce, to recruitment data and extensive surveys of developers and creators, and then to the systematic analysis of core innovative enterprises in the AI Coding ecosystem, as well as in-depth insights from nearly 150 in-depth interviews with founders or core builders, we are also looking at the evolution of the software industry and the AI transformation process of future industries through a unique "AI lens."

Under the data lens, AI Coding is rapidly crossing boundaries in terms of penetration rate, adoption rate, and influence from the consumer side to the enterprise side. This overwhelming leading advantage is most directly reflected in the financing market. What's more impressive is the miracle this industry has created in terms of revenue growth.

From the perspective of AI innovation, AI Coding companies are achieving annual recurring revenues (ARR) of tens of millions, hundreds of millions, or even $500 million with a very small team size and at an extremely fast development speed, challenging the growth and business models of large companies. From the perspective of AI transformation, programming is the application area where AI transformation occurs fastest and has the most significant effect within enterprises. Since Microsoft's Github Copilot, AI Coding has become one of the AI application areas with the second-largest financing scale after large foundational models and the fastest-growing. There has even emerged a company worth tens of billions of dollars that was founded only three years ago. Behind this explosive growth lies a systematic reconstruction of the industrial ecosystem.

However, when we focus the lens on product, value, and commercial competition, we can find that there are huge non-consensuses in seven aspects, from data to cognition and then to action, under the general consensus of the AI Coding direction.

Non-consensus 01: What is the best product form for AI Coding? - Local vs. Cloud
Non-consensus 02: Which model should an AI Coding product choose? - Self-developed vs. Third-party
Non-consensus 03: How much value does AI Coding bring to users? - Efficiency improvement vs. Efficiency reduction
Non-consensus 04: What is the ideal payment model for AI Coding products? - Fixed vs. On-demand
Non-consensus 05: What is the attitude of large enterprises towards promoting AI Coding applications? - Radical vs. Gradual
Non-consensus 06: What is the impact of AI Coding on organizational development? - Layoffs vs. Expansion
Non-consensus 07: What will the future market pattern of AI Coding be like? - Professional vs. Inclusive

If market data depicts "what is happening," then these non-consensuses will point to "why it is happening" and "where it is going," as well as what consensuses on products and technologies remain to be reached under the general consensus of the "age of abundance."

The following is the full text of the Report on Non-consensuses in AI Coding: