Using AI agents to code?
I promise I won't call you a vibe coder.
For the past 6 months I've been vibe coding a lot (damn!), and although there's a lot of buzz around it, you can't really do too much without studying a bit.
"But that tutorial said I can create an app in 15 minutes with just one prompt."
I'm sorry, that's simply not true.
If you have no idea of what you're doing you're still gonna have bad results, just automated. You still need to tell the agent clearly what to do and validate the results. So, the agent is just a very fast car, but you're still the one driving.
Tell it to build the wrong thing and congratulations, you now have the wrong thing. In half the time. So you basically become a Product Owner.
But here's the good side: once you get it, you're way more productive than before.
The trick is understanding roles. You can set up agents as a code reviewer, a tester, a security checker, someone monitoring your prod logs. If you're smart about it, you basically have a team working for you.
So to get better results, study:
- prompts
- roles and agents
- context management and memory
- observability
And for software development, you still gonna need basics of UX, programing logic, devsec and devops.
Feels overwhelming? It's not. There's a ton of free stuff out there. I'll be sharing the good ones here and making them easier to digest, so stay tuned!
By the way, how are you using agents in your workflow right now?