AI Reshapes Programming: New ‘TiaoXiaShi’ Role and Domestic Tools Surge Ahead

AI Programming Sparks Industry Debate

Recent discussions in the tech community have centered on AI‑assisted coding, with headlines ranging from high‑profile endorsements to rapid tool iterations, prompting programmers to wonder whether their jobs are safe and what future paths they should take.

Programmer Transformation and the Emergence of “TiaoXiaShi”

Experts assure that programmers will not disappear, but their roles will evolve. Zhou Hongyi popularized the term “TiaoXiaShi,” describing professionals who train intelligent agents such as OpenClaw. The CEO of Cursor echoed this view, suggesting developers will become “software‑factory architects.”

AI Collaboration Is Already Widespread

Anthropic’s latest report indicates that 60% of development work now involves collaboration with AI agents, and the proportion of AI‑generated code continues to climb. Nevertheless, the technology is far from perfect; accuracy gaps and security concerns still require human oversight.

Shift from Coding to Orchestrating AI

The core competency for developers is moving away from sheer lines of code toward skills such as requirement decomposition, reviewing AI‑produced code, and orchestrating multiple agents. In essence, the role changes from “writing code” to “managing AI that writes code.”

The AI Coding Tool Race Heats Up, Domestic Players Rise

While overseas tools like GitHub Copilot, Cursor, and Claude Code dominate early market share, Chinese products are catching up. ByteDance’s Trae stands out as a domestic champion, offering full‑Chinese support, a free‑forever model, multimodal interaction, and SOLO‑agent collaboration. Open‑source‑friendly tools such as Tabnine and CodeLlama, which support private deployment, are also gaining traction for their privacy advantages.

Tools Now Cover the Entire Development Lifecycle

Modern AI assistants have evolved beyond simple code‑completion plugins; they now assist with requirement analysis, coding, testing, and deployment. Selecting the right tool can boost developer efficiency by several orders of magnitude.

Industry Focus Shifts to Verification Infrastructure

Although AI‑generated code now achieves near‑perfect syntactic correctness (≈99% accuracy), logical errors and mismatches with business contexts remain common. Consequently, the industry is moving from a “generation‑only” mindset to building automated verification pipelines—automated review, testing, and validation systems combined with multi‑agent collaboration to create a generate‑verify‑correct loop.

Terminal Interfaces Supplant IDE Plugins

Terminal‑style interfaces are increasingly replacing traditional IDE plugins to better accommodate the global operation needs of agents like OpenClaw, signaling a shift in how developers interact with AI assistants.

Self‑Play SWE‑RL Breakthroughs Push the Ceiling

Meta recently introduced Self‑play SWE‑RL, a technique that mirrors AlphaZero’s self‑play approach for software engineering. The system pits an AI “attacker” against an AI “repairer” within a code sandbox, enabling training without human‑labeled code data. This self‑play paradigm could allow AI to surpass human coding performance, representing a core direction for next‑generation AI programming models.

Conclusion: Embrace AI Collaboration

The overarching message is that AI is not here to replace programmers but to empower them. Rather than fearing obsolescence, developers should learn to collaborate with AI tools, experiment with emerging platforms, and position themselves to capture the next wave of industry opportunities.