AI Coding Tool Trends: Lessons from Kiro

RevenueCat's Vibe Coding Catfe slide.

Japanese Ver: https://speakerdeck.com/s4yuba/kirokarakao-eru-aikodeinguturunochao-liu

Event link：https://revcat-catfe.tokyo/

X(Twitter)：https://x.com/gaishi_narou

Transcript

1. AI Coding Tool Trends: Lessons from Kiro (2025/07/30) Vibe Coding Catfe Speaker: Oikon
2. Self Introduction Oikon Software Engineer R&D at an IT Company Hobbies: Personal Development & AI Tool Research Today’s Topic: Kiro 𝕏 (@gaishi_narou): Zenn:
3. Today’s Message ❌ Not a "How-To" Guide for Kiro ⭕️ But a Look at What Kiro Teaches Us
4. Recent AI Coding Tool Trends GitHub Copilot (2021~) The pioneer of natural language code generation in VSCode (15+ million users) Cursor, Windsurf (2023~) VSCode forks with an AI-first IDE philosophy and various AI features Claude Code, Gemini CLI (2025~) Autonomous coding AI with high autonomy for implementation, testing, and deployment From "AI that assists with coding" to "AI that codes autonomously"
5. Challenges of Autonomous Coding AI Even Claude Code and Gemini CLI aren’t perfect. Unnecessary rework: Rushes to implement without a plan Lack of project context: Limited by the context window Irrelevant code generation: Context can be polluted by unnecessary information While effective for initial Vibe Coding, personal development, and small projects, challenges remain as projects scale.
6. Introducing Kiro An AI-driven IDE from AWS (July 14, 2025) VSCode fork Preview version available Takes you "from prototype to production" A new approach that brings structure and planning to AI coding -> 3 key features (HP: https://kiro.dev/) (ref) Introducing Kiro
7. Kiro’s 3 Key Features 1. Spec-Driven Development Clearly define "what to build" upfront. 2. Agent Hooks Trigger automated tasks based on events. 3. Steering Manage persistent context for your project. A mechanism to ensure the AI correctly understands the project and stays on track.
8. 1. Spec-Driven Development Clearly define "what to build" upfront. 3 Main Documents 1. requirements.md (Defines requirements) 2. design.md (Technical design) 3. tasks.md (Implementation plan) These documents are automatically generated through Vibe Coding.
10. Agent Hooks Automatically execute predefined actions based on trigger events. Trigger Events File Created File Saved File Deleted Manual Trigger Practical Examples Automatically run tests Automatically update READMEs Run security scans etc.
12. Steering Persistent context for projects with multiple files. Steering Files product.md (Product information) structure.md (Project structure) tech.md (Technology stack) manual-created.md (Custom steering file) Configure conditions to provide persistent context based on file types.
14. Kiro’s Challenges Preview version with a waitlist Only supports Claude Sonnet 4 and Sonnet 3.7 AI agent execution is slow The best approach is to use Kiro for requirements, design, and specs, then implement with autonomous AI agents like Claude Code.
15. Key Takeaway 1: What We Can Learn from Kiro VSCode -> Cursor, Windsurf -> Claude Code, Gemini CLI -> What’s next? Spec-driven development helps the AI accurately understand What to build. Engineers are freed from the implementation details of How to build. A shift from ad-hoc Vibe Coding to a structured development process. Engineers are gradually freed from coding and can focus on upstream design. (ref) From Chat to Specs, Kiro and the future of AI spec-driven software development
16. Key Takeaway 2: What We Can Learn from Kiro Insights into Context Engineering AWS’s vision for the "context AI agents need" (Specs) Solutions to context engineering challenges: Separating persistent context from minimal, necessary usage (Steering) Constraining the AI with rules outside of the context (Agent Hooks) Enabling all engineers to define requirements and create appropriate task granularity (Specs) Most AI tools "make strong engineers stronger." Kiro can "bridge the gap between junior and senior engineers."
17. Summary The development of AI coding tools is advancing rapidly. For now, context engineering requires careful consideration. Kiro offers valuable insights into the future of AI-powered coding. As AI models improve, the way we code will also change. It's important to understand the motivation behind new AI tools and to actively try and choose the right ones for your needs. Let’s start by trying Kiro!
18. References Official Documentation Introducing Kiro (2025-07-14) Kiro Homepage (2025-07-24) Concepts - Docs (2025-07-24) Hooks - Docs (2025-07-24) Steering - Docs (2025-07-24) Technical Articles & Blogs Kiroとコンテキストエンジニアリングの時流 (laiso, 2025-07-19) 【Kiro】Kiro(要件定義) × Claude Code(実装) で役割分担させてみる (ryu-ki, 2025-07-15) Over 15 million developers now use this AI coding tool from Microsoft (Windows Central, 2025)