Pratham: Coding Tasks Outsourced to AI, Development Remains a Human Endeavor
Pratham stated that coding has become a task that can be outsourced to AI, while development is fundamentally a way of thinking that still requires human responsibility.
It defines coding as the execution of specific tasks, now efficiently completed by AI; development involves architectural decisions, product judgments, and systems thinking, which AI cannot yet replace.
In market mechanisms, developer resources are shifting from repetitive coding labor to higher-level architecture and product design. AI coding tool providers benefit from subscription and integration revenue, while traditional lower-level programmer positions are under pressure. Overall, software development funding is accelerating towards AI-enhanced teams and companies driven by product thinking.
Source: Public Information
ABAB AI Insight
Pratham, as an observer of developers in the AI era, has previously discussed the disassembly of engineering roles by AI, continuing the perspective emphasized by Jack Dorsey and others on "rebuilding rather than layering". This highlights the essential distinction between coding and development, a viewpoint that will gain broader resonance after the widespread adoption of AI proxy coding tools in 2026.
In terms of capital pathways, engineering teams are reallocating human budgets from pure coding positions to senior architects and product managers. Resources are leveraged through AI coding assistants like Cursor and Goose, with the strategic goal of retaining human core judgment while using AI to accelerate iteration cycles, avoiding the complete commodification of the entire development process by cutting-edge lab models.
Similar cases include the shift in graphic design from manual PS operations to Midjourney generation plus human aesthetic control, and early automation testing replacing manual QA, leading engineers to focus on testing strategy design. Current software development is in a transitional phase from code-intensive to thinking and systems design-intensive.
Essentially, this is a technological substitution: repetitive coding tasks are fully taken over by AI, with the mechanism being that after breakthroughs in large model contextual understanding and tool invocation capabilities, the marginal cost of execution efficiency approaches zero. This leads to a concentration of pricing power from coding skills to product architecture thinking and system-level decision-making capabilities, while forcing developers to transition from "code writers" to "problem definers."