AI is making almost every news headline and is attracting more and more people. Everyone has their own use cases — from writing and research to coding and application development. This might sound great on paper, but is the AI output actually usable without the right human input and knowledge?
To take an example: if AI can build a car, would you have the guts to drive it? Is everything in place? Are the brakes working correctly? Beyond even considering getting behind the wheel, there’s a long checklist of safety things to verify before that car goes on the road.
How I use AI in development
Returning to current, real-world usage — I’m already using tools like Claude for coding when developing RunningLog and Planaro. But in order to extract real value from the AI output, I need to bring my own knowledge, my software development and architecture skills, to give AI the right input in the first place.
And I don’t stop there. The AI output doesn’t just get thrown into production for users to test directly. I review everything — doing code review, checking if what I requested is actually there, verifying how the new code integrates with the existing codebase, and testing the new functionalities within the app.
Why the extra effort matters
This entire process might look like too much, but it’s there to ensure that both RunningLog and Planaro remain stable and that users can rely on them without bumping into issues. Keeping the applications reliable and offering a good user experience is always the goal when doing development.
At this moment, AI is hit or miss. For some features, its code is fine from the first attempt, reducing development time to just a few minutes. However, when it makes mistakes — mixing API versions, breaking layouts — those errors can make the app almost unusable. That’s why catching these mistakes early is critical, before AI keeps iterating on top of them and making things worse.
The bottom line
Using AI in software development is powerful and can compress development time from days to minutes. But it can also fool inexperienced people into thinking they just need to write some input and they’ll have a working application. The reality is that AI makes good developers faster — it doesn’t replace the skills needed to build reliable software.
