Claude 4.1 Opus: My Co-Developer Since Claude 3.7 - Now Even Better

AI10 min read

August 5, 2025 – 10 min read


Anthropic released Claude 4.1 Opus today. I've been developing with Claude since April 2025, after previously using OpenAI o3, GPT-4, and GPT-4.1. Claude 4 Opus was the game-changer for me, and 4.1 is another step toward fully autonomous development teams.



My Claude Journey: From 3.7 to 4.1


When I switched from OpenAI to Claude in April 2025, I was skeptical. After months with o3 and GPT-4, I was used to OpenAI's ecosystem. But Claude was different. It didn't just understand code, it understood context, architecture, and most importantly: It felt like a real senior co-developer.


My Claude Evolution:

  • **Claude 3.7** (April 2025): The entry - immediately impressed
  • **Claude 4** (May 2025): The game-changer - noticeably better than everything before
  • **Claude 4.1** (today): The perfect co-dev

    • As a solo developer for our AIpuna App project, Claude isn't just a tool - it's my virtual team.



    My Setup: Claude Code in IntelliJ


    I use Claude Code integrated in IntelliJ IDEA. Why IntelliJ? Not because it's better than the others - I've tested Cursor, Windsurf, Devin, and others. The IDE hardly makes a difference anymore today. I just know IntelliJ best, and that's all that matters.


    Why Claude Code?


    For $200/month I get:

  • Claude 4.1 Opus access
  • Limits that reset every 5 hours
  • 24-40 hours of Opus 4 per week (according to Anthropic)

    • $200 sounds like a lot, but if you code all day and need large context, all other offers will make you poor. Anthropic will reduce limits by end of August, but it's still more fair than all others. Especially when you compare the quality, where Claude shines the most.


    Claude Code is, as of now, the best and cheapest for my tasks.



    Why Claude Is Unbeatable for Software Development


    After four months of intensive use, I can say: Claude understands software development like no other model.


    Claude understands software development:

  • **Architectural thinking**: Claude thinks in systems, not individual functions
  • **Consistency across sessions**: My coding style is understood and maintained
  • **Senior-level debugging**: Finds bugs I would miss
  • **Sensible refactoring**: Suggests meaningful improvements, no overengineering

    • The AIpuna Project as Proof:


    Our AIpuna App - a multi-AI-model app for individuals and businesses - was developed 95% with Claude. That's:

  • Hundreds of thousands of lines of production code
  • Custom FastAPI backend (engineered with Claude Code)
  • Flutter mobile app
  • Complete CI/CD pipeline
  • Stripe payment integration
  • Cloud infrastructure (managed with Claude Code support)

    • Time to MVP: 3 months instead of estimated 12 months. As a solo dev!



    Claude 4.1: The Concrete Improvements


    The new features aren't marketing gimmicks, but real productivity boosters.


    Claude 4.1 is according to Anthropic their most intelligent model and leading for coding and agents. I notice the improvements daily:


    1. Better Code Accuracy


    Claude 4 would have needed 2-3 iterations here. 4.1 gets it right the first time.


    2. Multi-File Context Understanding


    This is the killer feature for real projects:

  • I can refactor multiple files simultaneously
  • Claude understands dependencies across module boundaries
  • Automatic impact analysis for changes

    • 3. Agent Capabilities


    Claude 4.1 can really work autonomously:

  • Independent debugging
  • Automatic test creation
  • Performance optimization without guidance


    • The Tool Landscape: What I've Tested


    Before landing on Claude Code, I tried many things:


    Cursor: Good IDE integration, but expensive with intensive use

    Windsurf: Interesting approach, not yet mature

    Devin: Positions itself as "autonomous AI software engineer", but in my experience doesn't come close to Claude Code (my experience may be outdated)

    GitHub Copilot: Good for autocomplete, not for complex tasks

    Various API wrappers: Too cumbersome for daily workflow


    In the end, Claude Code with IntelliJ is my choice - not because it's perfect, but because it works best for my needs.



    The Hard Reality: Limits and Costs


    Anthropic's new limits (starting August 28, 2025) combine two systems:


    Max 20x Plan ($200/month):

  • 240-480 hours Sonnet 4 per week
  • 24-40 hours Opus 4 per week
  • Additionally: Session-based limits that reset every 5 hours
  • Weekly limits reset every 7 days

    • But the comparison makes it clear: For intensive use with large context, Claude Code is still the best value for money. API usage would quickly reach four figures with my volume.



    Why I Stay with Claude


    Every time a new model comes out, I test it. GPT-5 last week? Tested. o3? Tried extensively. The result is always the same:


    Other models with code:

  • Often only understand superficially
  • Generate "tutorial code" instead of production-ready
  • Lose context in complex projects
  • Don't feel like a colleague

    • Claude understands me:

  • Knows my coding style after a few examples
  • Makes sensible suggestions, no wild experiments
  • Stays consistent over long sessions
  • Admits when it's uncertain

    • It's like the difference between a random freelancer and a long-time team colleague.



    My Daily Workflow


    After months of optimization:


    Development Environment:

  • IDE: IntelliJ IDEA with Claude Code integration
  • Additionally: Direct Claude.ai use for planning
  • Version Control: Git with meaningful commit messages
  • Testing: Claude writes tests parallel to code

    • Daily Workflow:


    Morning:

  • Daily planning with Claude
  • Architecture decisions for new features
  • Code review of night commits

    • During the day (Pair Programming):

  • Me: Business logic & requirements
  • Claude: Implementation & optimization
  • Together: Real-time problem-solving

    • Evening:

  • Refactoring session
  • Documentation updates
  • Test coverage improvement


    • Tips for Other Developers


    The Right Expectations:

    Claude is a senior developer, not a magical oracle. Treat it accordingly.


    Context Management:

  • Give Claude all relevant context
  • Use the 5-hour resets strategically
  • Save important conversations for later

    • Tool Selection:

  • The IDE hardly matters - take what you know
  • Claude Code for intensive development
  • API for automation and CI/CD

    • Budget Planning:

  • $200/month sounds like a lot, but is cheaper than alternatives with intensive use
  • Compare with your productivity increase
  • Think about the time saved


    • The Future with Claude


    With Claude 4.1 and the agent capabilities, I see a future where solo devs like me can build complete enterprise solutions. Not in years, but in months.


    For AIpuna this means:

  • Faster feature development
  • Higher code quality
  • More time for business development
  • Competitive against large teams


    • Conclusion: A Real Co-Developer


    After four intensive months with Claude, I can say: It's not hype. For software development, Claude 4.1 Opus is the best partner you can have.


    Is it perfect? No.

    Are the limits sometimes frustrating? Yes.

    Is it worth every cent? Absolutely.


    If you're a solo dev or work in a small team: Claude 4.1 with Claude Code is your chance to compete on equal footing with large developer teams. The $200/month is an investment that pays off multiple times.


    The future doesn't belong to large teams - it belongs to smart developers with the right AI tools. And Claude 4.1 Opus is definitely the right tool.


    Patrik Germann

    Solo Developer of AIpuna App

    Claude user since April 2025

    Comments

    Ready for AI Transformation?

    Let's explore the possibilities of AI for your business together.

    Schedule Consultation
    Book consultation
    TOBG - DLT, Crypto, Mindset, Community