How I use AI without handing over the work
I use AI the same way I use documentation, debugging tools, code review, and search: as another tool in the process. It helps me test blueprints, learn unfamiliar syntax, explore ideas in more detail, and move faster, but I still decide what gets built, review the output, test the result, and take responsibility for the finished work.
Where it helps
The parts of my workflow where I use it
I use it most when I am exploring a new concept, learning a pattern, testing a rough blueprint, or trying to explain an idea more clearly before I commit to the final implementation.
Concepts and technical blueprints
When I have an idea but have not decided how to structure it yet, I use AI to sketch possible flows, components, data structures, and implementation plans. I treat those as blueprints to test and revise, not finished answers.
Learning new syntax
I use AI to break down unfamiliar syntax, explain why a pattern works, and compare it with languages or tools I already know. I still verify the details with documentation and test the code myself.
Prototyping and testing ideas
I use it to turn rough concepts into small prototypes so I can see whether an idea actually works. Once I understand the direction, I change the structure, improve the interface, and build the version that fits the project.
Debugging
When something breaks, I use AI to help me reason through the error, identify likely failure points, and compare possible fixes. I still inspect the code, logs, requests, and production behavior myself.
Review and cleanup
It helps me find duplicated logic, unclear writing, missing edge cases, and parts of an interface that could be easier to understand or use.
Explaining ideas more clearly
Sometimes I understand what I want to build but need help putting the idea into clearer technical language. I use AI to expand the concept, question my assumptions, and help me explain the final approach to another person.
What stays my responsibility
I am still responsible for the final result
I decide what the project should do and what belongs in the final version.
I review and edit generated code before keeping it.
I run builds, test the interface, and check the result in the real environment.
I verify calculations, data interpretations, security decisions, and technical claims.
I remove suggestions that do not fit the project or that I cannot confidently explain.
I use official documentation when I need to confirm syntax, APIs, package behavior, or security guidance.
Learning and exploration
Sometimes I use it to understand the idea before I build the idea
When I am learning new syntax or working with an unfamiliar API, I use AI to break the problem into smaller pieces, compare the new pattern with something I already understand, and show me a few possible approaches. I do not stop at the explanation. I rewrite examples, test them in the project, check the official documentation, and make sure I can explain why the final code works.
Where I slow down
I do not trust every answer just because it sounds confident
Authentication and security
I do not assume generated authentication or security code is safe. I review the flow, test failure cases, and check the official documentation.
Databases and destructive changes
I slow down around migrations, deletes, permissions, Row Level Security, and anything that could damage or expose data.
Analytics and conclusions
I verify formulas and interpretations instead of trusting a confident-looking answer. Correlation, scoring models, and business conclusions still need judgment.
Privacy
Private data does not belong in a public AI prompt
I avoid sharing passwords, API secrets, private customer or player information, employee data, proprietary files, internal campaign details, or other confidential business material. When I need examples, I anonymize the information or use synthetic data instead.
This portfolio
AI helped with parts of the process. I still built, tested, and maintained the site.
I used AI while brainstorming interfaces, testing technical blueprints, learning new syntax, restructuring copy, reviewing code for a second opinion, and troubleshooting development issues. I still made the final design decisions, wrote the codeconnected the pages, reviewed the code, ran production builds, tested the features, handled deployment, and decided what was accurate enough to publish. We do not vibe-code here but we do use all the tools available.