I’ve been building a Retrieval-Augmented Generation system (bear with me)
… and I took a week off. Just one week.
In that time, so much changed (looking at you, Make-com - adding agents to your workflow) that I now have to restart the project. My week-old system is already outdated.
What pulled me down this rabbit hole in the first place was something my team lead said:
“I think we’ll look back on this as the time before and after AI—just like smartphones.”
That line stuck, so I figured I’d better get on the bus.
But now I’m wondering if this is still a season to learn more than launch.
Every AI subreddit and YouTube channel is full of people hawking their apps and “quick wins.” But if you can’t trust platforms not to update their LLMs, pricing models, or feature sets every other week… how do you build a product that depends on them?
This isn’t some AI-generated chin stroke. This is a legit question:
What areas of AI actually feel stable enough to build real products on?
I feel good using automation to scale my own services—but selling standalone, turnkey tools? Are we there yet?
