It was a dream, it had to be a dream. I can’t believe I am witnessing the things that I am. It just wrote 50K loc in a matter of an hour or two, all on it’s own from a list of TODOs that I prepared for it, I can’t believe that this is real!

Are we living in a dream? No!
Great, how do we know that this going to take my job? No! What can I use this for? Yes! What can I get it to do? Yes!

Why

If we are not thinking about it, then you should. I have seen a huge lift in my productivity. In the last 2 weeks alone, this thing has produced several thousand lines of high quality code, tailored by prompts and commands that I have worked on.

Yes high quality, Im talking about Builders and Factories for my large complex data sets. Extracting methods, reusable code in how I design it in my prompt examples, and creating abstractions encapsulation of implementations.

For every PR I use the tool, I write an eval breakdown of the cost, and estimate times saving. I have seen time savings of at least 60% - 80% consistently.

Giving more time mentoring, hosing out fires in production, creating more features, and leading my team, best of yet more time researching how to become even more proficient at AI.

Don’t Miss Out, Learn Now

By early next year if your team is not using AI, you are doing it wrong. And no I am not talking about CoPilot. Going forward, I am making a strong urge to my company that we should not hire anyone that is not using AI, i’m pretty confident about this and will press it to my superiors. I have that much strong conviction in AI.

Over the course of the next couple weeks, I will write a series of block post that will said the light of what I’ve learned on AI and what I will continue to use it for.

Here is my Claude Code work flow. It’s based off a concept called SPARC. I have been working on a few projects, mostly just to see what AI can do on it’s own. Testing it’s limits, but when I work at Nomi Health, or on my personal project Options Deep, I use this workflow.

My Workflow

  • Plan out what you are going to do.
    • Prompt the AI to generate a plan.
    • Ask the AI to go and do some research and generate Design Docs for your review
    • Ask the AI to produce some questions that could reveal flaws in your PRD (product requirements doc).
    • Select portions of the research and findings and responses to your answers and ask it to generate a better Design Document.
    • Ask it to generate Prompts.
    • Ask it to generate a TODO list.
  • Plan out what the code looks like
    • Tell it to generate Pseudocode
    • Tell it to generate a mermaid relationship chart of your data models.
    • Tell it to generate high level mocks of APIs
  • Architect your pieces together
    • Tell it to generate a Architecture document.
    • Tell it what needs to be scalable in your architecture?
    • Tell it what are your business needs given the architecture?
    • Tell it what technologies you utilize in your project.
  • Evaluate and refine
    • Iterate with the AI and improve the design of the architecture and the look of the code.
    • Evaluate it’s pseudocode, how you want the code to look like.
    • Evaluate how each of the data models relationships are created.
    • Ask it how you can enhance performance of the system.
    • Begin to write out the code.
  • Production Ready
    • Create tests around your code base. Read the code for the tests.
    • ask it to document the codebase.
    • ask it to create your CI/CD and enhance it to tailor the project it is working on.