Development PracticesConference50min
It's Dangerous to Code Alone! Take This: Developer's AI Survival Guide
The session warns against “wow demo” AI coding and its hidden costs: technical debt and verification bottlenecks. It teaches production prompting, context management, MCP tools, agents/subagents, and spec-driven development to build reliable, maintainable AI-assisted software.
talk.summaryAiDisclaimer
Arnaud JeanAWS
Muhammed Salih GülerAWS
Right now, your feed is flooded with developers claiming they casually "vibe coded" an entire application before breakfast. It is incredibly easy to feel like you are missing out or falling behind the curve as the rest of the industry races ahead. But here is the truth nobody is sharing: the industry is caught in a massive "wow demo" trap.
While it looks like magic when an AI generates an app in minutes, that raw speed is creating a hidden crisis. We are seeing a massive surge in unmaintainable technical debt and a crippling "verification bottleneck" where developers are spending more time untangling AI hallucinations than writing actual logic.
In this session, you will learn:
- Prompting for Production: You will learn structural techniques like few-shot and chain-of-thought prompting to minimize logical errors and guide the model effectively.
- Context Management: You will learn how to observe the dynamic context your AI sees, manage token limits, and avoid "context rot" so the AI doesn't lose track of your codebase.
- Developer Tools: Learn to build and use MCP servers to safely ground your agents in real data, documentation, skills, and external APIs.
- Agents and Subagents: You will understand why monolithic agents fail at complex tasks and how to securely delegate specific workflows to specialized subagents running in parallel.
- Spec-Driven Development: You will use SDD to make your formal specifications the definitive source of truth.
While it looks like magic when an AI generates an app in minutes, that raw speed is creating a hidden crisis. We are seeing a massive surge in unmaintainable technical debt and a crippling "verification bottleneck" where developers are spending more time untangling AI hallucinations than writing actual logic.
In this session, you will learn:
- Prompting for Production: You will learn structural techniques like few-shot and chain-of-thought prompting to minimize logical errors and guide the model effectively.
- Context Management: You will learn how to observe the dynamic context your AI sees, manage token limits, and avoid "context rot" so the AI doesn't lose track of your codebase.
- Developer Tools: Learn to build and use MCP servers to safely ground your agents in real data, documentation, skills, and external APIs.
- Agents and Subagents: You will understand why monolithic agents fail at complex tasks and how to securely delegate specific workflows to specialized subagents running in parallel.
- Spec-Driven Development: You will use SDD to make your formal specifications the definitive source of truth.
Muhammed Salih Güler
Salih is a Senior Developer Advocate at AWS and Google Developer Expert with over 10 years of experience spanning fullstack, mobile, and cloud development. He specializes in bridging complex engineering challenges with developer communities through hands-on code, content, and technical strategy.