Augment Code AI

Alright, let’s talk more about Augment Code AI Coding Agent — the coding assistant that’s slowly turning me into a lazy (but efficient) developer. I did already analyzed it a bit in my previous post, but here I would like to go more into the details.

I’ve been using it for a month now, putting it through its paces on a real-world project. The version I tested ran on Claude Sonnet 3.7, though they’ve just upgraded to Claude Sonnet 4 (I haven’t had enough time to fully test it yet, but early signs are very promising — more on that later).

After weeks of daily use, I think I’ve reached a point where I trust this thing. Not blindly — I still review its work — but enough that I no longer brace for disaster every time I ask it to refactor something. That’s a big deal.

The Good Stuff (Where Augment Code Shines)

1. It Fixes Problems I Didn’t Know How To Solve

More than once, I’ve described a problem vaguely (sometimes even naming the completely wrong component), but Agent still figured out the right fix. It’s like having a senior dev on call who ignores your rambling and just… solves it.

2. Database model refactoring? Easy.

When I needed a new feature which required some DB model changes, Agent automatically generated a Liquibase script for the migration. No manual SQL, no forgotten columns and stuff— just a clean, working update.

3. Debugging with Proof

One time I really had a special problem to solve, at the end it wrote a test script, ran it, and exported logs to prove the behavior matched what I wanted. That’s next-level helpful.

4. Need to replace current chart lib with other AND keep all features? No problem.

I threw a massive chart redesign at it (new chart library had to be used, reusing as much as possible of code, keeping all features like onClick events, custom chart labels, support for different data types, etc… ). First try, 10 minutes later, it was done — just a couple of minor tweaks needed.

5. Reads The Docs So You Don’t Have To

My favorite moment? When I needed to upgrade vue3-toastify, since current toastify version did not work well with my Vue version. And the new, compatible one, was not backward compatible, i.e. most code for toastify was not usable anymore. Agent went to the online docs, figured out the changes, and fixed it. Chef’s kiss.

6. Self Test Driven Coding Mode

A few times, it realized a test was failing because the original implementation was wrong — so it fixed the code, not the test. That’s awareness you don’t get from most AI tools.

7. Remembers your preferences

Like a good pair programmer, it picks up on your coding style and sticks to it. Less time fixing formatting, more time shipping.

8. A picture speaks more than a thousand words

If you have a problem with design or CSS, which is a little harder to explain in words — make a screenshot of that part and send it to agent. It will probably figure out what the problem is and propose a solution. Very cool.