I have been using AI for coding a lot, and that part gets most of the attention. Fair enough. It is impressive.
But honestly, some of the most useful ways I have used AI have been much more ordinary than that. Not “build me a company.” More like: help me think through this real-life problem with actual constraints.
That is the part I keep coming back to. It is not just about speed. It is about getting a strong draft when the problem is messy.
So here are four very practical things I have done with AI recently.
1. Protein shakes, but with real constraints
Someone I know was dealing with health issues and had reached the point where eating itself had become difficult. The doctor said: give protein shakes.
Sounds simple enough, right? Not really. Because there were constraints. Real ones.
The person also had diabetes and high cholesterol. So this was not just a case of throwing random fruit, peanut butter, ice cream, and protein powder into a blender and calling it healthy. It had to actually fit the situation. It had to be gentle enough, useful enough, and ideally... taste good enough that they would actually drink it.
That last part matters more than people think.
If it does not taste good enough to actually drink, it is not really helping.
An LLM was surprisingly useful here. Not as a doctor. Not as the final answer. But as a very fast way to work through constraints and get to a good starting point. Higher protein, more thoughtful about sugar, better for cholesterol — and still realistic enough that someone would actually want to drink it.
That kind of shortlist usually takes weeks of trial and error. We had it in minutes.
2. Planning an Alaska trip without losing days to research
We were planning a trip to Alaska.
Before LLMs, this is exactly the kind of thing that would have turned into days of research.
Which places should we visit? How do we make sure the drive time each day stays under two hours? What places can we eat at if we are vegetarian? Which tours are actually worth it?
That is not hard in the sense that any one question is impossible.
It is hard because there are too many moving parts.
And that is where an LLM really helped.
Instead of starting from zero, we had a pretty good draft itinerary in minutes. Not perfect. Not “book everything exactly like this.” But a real starting point.
The value was not that it planned the whole trip for us.
The value was that it compressed the “where do we even begin” part from days to minutes. Here is the itinerary we ended up with.
3. Vegas food planning for a group trip
We had a Vegas trip coming up for a pool tournament, and one of my friends asked whether I already had a list of places for us to eat.
Now this was not just “find good restaurants in Vegas.”
It was a group of five. One vegetarian. We preferred Asian food. It could not be too pricey, could not be on the Strip, had to be highly rated, and needed to include three specific places we already knew we wanted to try. On top of that, lunch had distance restrictions. Dinner did not.
So the real task was not just finding popular places. It was finding good food that actually fit the trip. About 30 minutes later, mostly as a joke, I sent the group this link.
And the funny part is: it was actually useful.
Some of the places on that list were exactly the kind of hole-in-the-wall spots we would never have found otherwise, and they turned out to be really good.
Because again, the problem was never really “find good restaurants in Vegas.” It was “work through a bunch of real constraints and still end up with places worth eating at.” That is where AI is quietly very good.
4. Going from app idea to production much faster
And yes, coding is still a huge one for me.
On personal projects especially, going from idea to production has never been faster.
I was tired of paying a monthly subscription for a dividend tracker. Then I found Tiingo.com — generous free tier, everything I needed. I figured I could just make an app for myself. MVP in a day. FinanceForest: Portfolio released to the App Store in 10 days.
Not because the LLM replaces thinking. For me, it does the opposite. It frees up more of my thinking for the parts that matter.
I can spend more time on architecture, product direction, tradeoffs, and feature ideas instead of getting slowed down by every tiny implementation detail or by the weight of starting from scratch.
That changes the whole energy of building. You can explore more. Try more. Ship faster. Learn faster. And if you use it properly, you can still maintain product quality while moving much quicker than before.
What ties all four together
The common thread across all four is pretty clear to me. I do not find AI most useful when it is pretending to be magical. I find it most useful when it helps with messy problems that have real constraints.
Food restrictions. Travel time. Vegetarian options. Group logistics. Product architecture. Feature planning. Different problems, same pattern.
AI gives me a better starting point, faster.
Not the final answer. Not better decisions. Not a replacement for thinking. Just a much better draft than staring at a blank page.
On a lighter note, I asked ChatGPT to generate an image of me based on everything it knows about me.
Pretty close, honestly :)