How Large‑Model AI Is Transforming R&D Workflows

Introduction

Today's topic explores the development and application of large‑model technology in the R&D field. While large models already assist code writing, the discussion asks whether they can fundamentally reshape workflows and key processes.

Agenda

Intelligent R&D tool development

Enterprise implementation experience

Developer practice

Q&A session

1. Evolution of Intelligent R&D Tools

We first review the history and direction of intelligent R&D tools. Early tools such as GitHub Copilot appeared about two and a half years ago, initially offering code‑completion suggestions. Over time Copilot added features like chat‑based language conversion and unit‑test generation.

With the rise of large models, products have rapidly evolved from simple assistance to collaborative AI that can generate code from requirements, even setting up testing environments, moving beyond mere code snippets to services that support entire engineering workflows.

The recent changes can be grouped into code‑generation volume, shifting tool responsibilities, performance optimization, capability upgrades, and innovative user‑experience interactions.

2. “Go Fat”: Scaling Code Completion

The term “Go Fat” describes the growing scale of code‑completion. Traditional completion inserts a gray suggestion line based on model inference, turning a few keystrokes into a Tab press. Current tools aim to expand this capability while preserving user habits.