I work at Google Cloud. I use Claude Code for most of my development work. Let me explain why.

It’s not about which tool is better. It’s about which pricing model fits my use case.

The Technical Reality

Let me be precise about what I’m using and why.

Gemini Code Assist is the enterprise product with enterprise pricing. That’s one path.

Gemini CLI is an open source project. It requires a Vertex AI API key or Google Cloud project. The per-token cost might be better, but I’m capped by the quota for Gemini Pro tool calls per day.

For the volume of complex planning and research I’m doing with this 1-person bootstrap experiment, that quota isn’t viable.

Claude Code Max gives me the most cost-effective access to:

  • Opus (the most complex planning model)
  • Sonnet 4.5 (newer than gemini-flash-2.5)

That matters when you’re running multi-agent orchestration workflows daily.

BrandCast as a Comparison Project

The entire BrandCast project started as a real-world comparison between Gemini CLI and Claude Code. I wanted to build something substantial enough to evaluate both tools properly.

Not just toy examples. Not just “hello world” demos. A real SaaS product with:

  • Marketing automation
  • Customer discovery workflows
  • Content generation pipelines
  • Analytics integration
  • Multi-agent orchestration

The kind of work where AI coding tools actually matter.

Why I Still Use Both

Here’s the thing: I use Claude Code for building BrandCast, but I validate everything with Gemini CLI.

Why? Because I work with Google Cloud customers who are evaluating these tools. I need to understand both:

  • What works well in Gemini CLI
  • Where the workflows differ
  • How to communicate patterns that work across both platforms
  • What the experience is actually like for enterprise vs. individual users

Using both tools makes me better at my job. It keeps me honest about capabilities and limitations.

What This Looks Like in Practice

My typical workflow:

  1. Primary development: Claude Code

    • Agent orchestration for marketing
    • Content publishing automation
    • Customer discovery workflows
    • Daily operational tasks

  2. Validation: Gemini CLI

    • Re-implement key workflows
    • Compare agent behavior
    • Test pattern transferability
    • Document differences for customers

  3. Customer conversations: Both

    • “Here’s what works in Gemini CLI”
    • “Here’s the equivalent pattern in other tools”
    • “This pricing model fits your team size”

The Honest Evaluation

Right now, Claude Code Max gives me:

  • Access to Opus for complex planning
  • Sonnet 4.5 for execution (newer model)
  • Higer quota limits for my workflow

Gemini CLI gives me:

  • Open source extensibility
  • Deep Google Cloud integration
  • Lower per-token cost
  • But quota-limited for the volume I need

This will change. I’m watching for:

  • Gemini 3.0 launch
  • Gemini CLI subagents
  • Quota increases for Pro tier

The goal isn’t to pick sides. The goal is to build a solution that’s eventually model-agnostic.

Why This Matters

If you work at a company that builds developer tools, you should use competing tools. Not to bash them. To understand what good looks like from different angles.

It makes you:

  • More honest about your own tool’s strengths
  • More aware of what customers actually want
  • Better at articulating why different tools fit different needs
  • Less likely to fall into feature parity traps

The best product people use the competition. The best engineers evaluate honestly. The best advocates understand the whole landscape.

Building for Model Agnosticism

Here’s the real point: I’m not building a Claude Code application or a Gemini CLI application.

I’m building agent orchestration patterns that will work across models.

The context markdown, the agent configurations, the feedback loops - none of that is model-specific. The patterns transfer.

When Gemini 3.0 launches with better planning capabilities, I’ll re-evaluate. When Gemini CLI adds subagents, I’ll test whether the quota constraints loosen up. When Claude releases the next model, I’ll compare again.

This is what you do when you’re building something real: you use the best available tools for your constraints, and you design for portability.

Right now, for the volume and complexity of work I’m doing, Claude Code Max is the right choice. Six months from now? We’ll see.

What You Should Do

If you’re choosing between AI coding tools:

  1. Understand the quota constraints - Daily limits matter for high-volume work
  2. Evaluate model access - Which planning models can you actually use?
  3. Try both in real projects - Not toy examples
  4. Design for portability - Build patterns that transfer across models
  5. Re-evaluate regularly - New models and features change the calculus

Choose based on your actual constraints and workflow needs.

That’s what I did. That’s why I use Claude Code for BrandCast while working at Google Cloud.

And it’s why I can have honest conversations with customers about when Gemini CLI is the right choice for their teams.

The Bottom Line

BrandCast began as a comparison project. Every workflow I build in Claude Code, I validate in Gemini CLI. Every pattern I discover, I document for both platforms.

This makes the BrandCast codebase useful beyond just building a product. It’s a reference implementation for AI-native development patterns that work across tools.

And it keeps me grounded in what actually matters: solving real problems with the right tools, not just using the tools my employer builds because that’s what’s expected.

Good engineering requires honest evaluation. Good advocacy requires understanding the whole market. Good products emerge from real constraints and real trade-offs.

That’s why I use Claude Code. And that’s why I still validate everything with Gemini CLI.

Both matter. Both teach me something. Both make me better at my job.