Hey Reader,

Let me describe a very "productive" afternoon.

You sit down to start building your Python project.

Then you want to check which framework to use.

A Reddit thread says LangChain is the best, a YouTube guy swears by LlamaIndex, someone in the comments insists both are obsolete now.

Oh, and which vector database should you use?

And should you use FastAPI?

Two hours later, you're totally lost, and you've made nothing you could call "progress".

I know this dance well, because I've done it many times myself 🙈.

And after the last newsletter, I could tell some of you were limbering up for it too.

Btw, you flooded my inbox with your one-sentence project ideas and their "definition of done." (thanks so much for sharing, I'm so proud of you!)

I loved all of them 🙌. From AI career predictor, to an underwriting copilot for reviewing applications, to a personal learning tool, a job-hunting assistant, a meal-plan helper, an expense tracker,...

After reading dozens of them, I noticed a quiet worry: but which tools should I actually use?

So if you find yourself in this "tool paralysis" situation, let me offer 3 thoughts.

---

1. Getting started matters more than the perfect tool

Tools are just... tools. They're good for different things, and the perfect one barely matters when you're starting out.

Take learning guitar.

You wouldn't spend months comparing hundreds of brands to find the exact one the pros play. It's a waste of time. You'd just grab a beginner-friendly one and start practicing.

Later, once you know how you play and where the guitar holds you back, you swap it for something that fits.

Same with working on a project.

Any tutorial insisting you MUST use this tool or that agent framework is usually just one person's setup, dressed up as a law.

Use the tool that seems most reasonable/ accessible to you, understand how it works (this is important!), and see what you can achieve with it.

And when the day comes that you do need to switch, you'll know precisely why, because you'll have hit the wall yourself. That's a reference point you never get if you never start.

---

2. Too many options actually limit you

Here's something that may sound backwards: constraints actually free you. They make you more creative, not less.

During my Computer Science degree, a lot of assignments forced us to use exactly one tool.

We had a module called Graphics Programming module. And we were only allowed to use a JavaScript drawing library called p5.js to build sceneries and games.

The tool is far from cutting edge. It actually looks like it came straight from the 80s, and you will likely never hear of it again.

But it did the job. I hit walls and had to use workarounds. And I felt like an actual graphics artist 🤣.

So pick one tool, commit to it, and see how far it carries you. I bet it's usually further than you'd guess.

---

3. Switching costs almost nothing now

Now, if you're still nervous about choosing wrong - maybe an AI model or tool turns out slow or expensive and trapping you in a painful migration later.

Honestly, that fear made a lot of sense a few years ago. It makes much less now.

These days, an AI coding agent will translate your code from one framework to another, swap your model, or move an entire codebase between languages.

... as easily as it translates English to Vietnamese 🙃:

The thing you're agonizing for hours is, increasingly, a 10-minute undo.

So whatever tool you pick today is not a marriage. It's a first date 😉.

If you've got a project idea, choose one tool for each part you need to get your project working, and write the first ugly version.

If this is helpful, hit reply and let me know! I read every single one 🙂.

Chào,
Thu

P.S: Once your little project is running on your machine, you'll want to share it. That's where cloud services come in. I'm publishing a video this week on using AWS Bedrock to build a document QA pipeline (RAG). Stay tuned :)

---