Agentic AI Engineering in Practice: How AI Engineers and Forward-Deployed Engineers Build with Claude Code, Codex, and Gemini

                     Agentic AI Engineering in Practice: How AI Engineers and Forward-Deployed Engineers Build with Claude Code, Codex, and Gemini
               



Tópico:
                     Agentic AI Engineering in Practice: How AI Engineers and Forward-Deployed Engineers Build with Claude Code, Codex, and Gemini
               
Categoria: Tutoriais | FreeCodeCamp Premium
Idioma Principal: Português (Conteúdo de Tecnologia)

Conteúdo do Tutorial / Guia Passo a Passo:
-------------------------------------------------------------------------
A practical, three-tool guide to the AI-native software development life cycle (SDLC): Plan, Design, Build, Test, Deploy, Maintain, reimagined for agentic coding.

In March 2025, a small nonprofit research group called METR published a chart that made many engineering leaders sit up straighter than usual.

Working backward through six years of model releases, METR measured the length of the software task an AI agent could complete on its own, which they defined as the amount of time a skilled human professional would need for the same task, and found that number has been doubling roughly every seven months since 2019 (METR).

That curve isn't about autocomplete getting a little better: it's about agents crossing from "finishes a function" to "finishes a feature" to, on the current trajectory, "finishes a sprint."

The adoption numbers already reflect that shift. Google Cloud and DORA's 2025 State of AI-assisted Software Development Report found that 90 percent of developers now use AI at work and more than 80 percent say it has increased their productivity, even though roughly three in ten still report low trust in the code the models produce (DORA).

Stack Overflow's 2025 Developer Survey puts a similar number on habitual use: 84 percent of developers now use or plan to use AI tools, up from 76 percent the year before, and about half of professional developers reach for one daily (Stack Overflow).

AI-assisted coding skipped the novelty phase and arrived as the default way software gets written. At the same time, most teams still haven't updated the software development lifecycle they built for the previous default.

That mismatch is the subject of this piece. Anthropic's Applied AI team published a framework in 2026 called the AI-native SDLC, built around a single observation: once an agent can write and revise code faster than a human can review a pull request, the bottleneck in software delivery doesn't disappear so much as relocate (Anthropic).

This guide walks through that framework stage by stage (Plan, Design, Build, Test, Deploy, Maintain), and shows you how to implement each stage with whichever agentic coding tool you have access to: Claude Code, OpenAI Codex, or Gemini CLI. You'll see configuration files, markdown artifact templates, and CI workflows for all three tools, plus a worked example of how a single forward-deployed engineer uses this pattern to cover work that used to require a five-person team.

This guide is one framework, implemented three ways: a practitioner's playbook grounded in config files and command output.

Table of Contents

• Prerequisites

• What is an AI-Native SDLC?

• Where the Bottleneck Moves Once Code Gets Cheap

• Claude Code, Codex, and Gemini CLI: One Framework, Three Vocabularies

• How to Run the Plan Stage

• How to Run the Design Stage

• How to Run the Build Stage

• How to Run the Test Stage

• How to Run the Deploy Stage

• How to Run the Maintain Stage

• How One Engineer Covers a Five-Person Team

• The Plan Stage

• The Design Stage

• The Build Stage

• The Test Stage

• The Deploy Stage

• The Maintain Stage

• Pre-flight Checklist Before You Go All-in on an Agentic AI-Native SDLC

• Conclusion

• What to Explore Next

Prerequisites

Before you start, make sure you have the following:

• One agentic coding CLI installed: Claude Code, OpenAI Codex, or Gemini CLI. You only need one to follow along. The sections below give you the e

... [O tutorial continua no link abaixo] ...