Diego Iastrubni
Diego Iastrubni
@diegoiast.bsky.social
I know that Qt was written before STL was a thing.. but they did not revert their decision, instead buried themselves in it.

Anyway, see some of the API ideas I was thinking. See the chainability of calls, see the function callbacks, see usage of `auto`:

github.com/diegoiast/sv...
github.com
November 29, 2025 at 10:28 PM
That was my idea when I wanted to create a new GUI toolkit. I thought, what if Qt did use the standard containers? What if instead of MOC it used standard `std::function`? What if instead of re-inventing everything in house (see next comment), I use 3rd parties?

github.com/diegoiast/sv...
GitHub - diegoiast/svision2: A new cross platform C++ GUI toolkit
A new cross platform C++ GUI toolkit. Contribute to diegoiast/svision2 development by creating an account on GitHub.
github.com
November 29, 2025 at 10:25 PM
If you use cmake+Conan, you dont have to deal with GCC directly.

What's preventing you from using it, or meson?
September 5, 2025 at 5:52 PM
...how did you took this image?
August 2, 2025 at 4:54 PM
You should use it on a function level. Then edit the output of the LLM to match your style, remove the extra comments. Then start reducing the diff to your main branch.

It can be used.
July 30, 2025 at 6:09 PM
Recording videos is really hard. In C++, this is extremely boring... as 20% of the time, is spent compiling.

I spent 30 minutes, (14 edited, as I speed it up) writing this patch: github.com/diegoiast/qt...
CTagsPlugin: load tags in a background thread (109) · diegoiast/qtedit4@bc5391a
Loading the ctags files should not be done in the main thread. We now load it in a background thread. This reduced on my setup 1.5 seconds on start up. In theory, this is bad, since others might c...
github.com
July 21, 2025 at 1:20 PM
so std::optional<T&> and std::weak_ptr<T> are basically the same?
June 14, 2025 at 7:41 PM
That is a pointer, with reference counting?

whats the difference between std::optional<T&> and T* ? (besides the auto memory cleanup)
June 14, 2025 at 7:15 PM
Most of the time I was idling, waiting for the LLM ... the output of the LLM is too verbose, and the code it produces is many times wrong, and always (almost...) just bad. You will spend the same amount of time fixing the code, as "coding" it.

We need to manage expectations. #vibecoding
June 8, 2025 at 7:45 AM
@cursor.com.web.brid.gy no, this is not ready for production. While the product did produce something that works, and it was not something I knew how do to, I will probably need to kill this branch and start over.

Even in this small scale, the experience is bad.

#vibecoding
June 8, 2025 at 7:43 AM
4. Useless debug everywhere. I just removed a for loop, which a debug statement inside.
5. Just changing random code:

- if (text.endsWith("\u2029")) {
+ if (text.endsWith("")) {

NO, no no...
6. On some places it removed documentation, in some added. What gives?
June 8, 2025 at 7:39 AM
I did not look at the actual changes... but OMG, what a mess.

1. Functions changed location, making the diff larger
2. `textToCopy = lines.join('\n');` - this was changed to ` textToCopy = lines.join("\n");`
3. Useless function renamed
June 8, 2025 at 7:24 AM
for the last 2 hours, I found that a new feature, removed another (or was badly implemented). Now, I am dancing with the AI, in order to convince it to re-add that functionality.

If this is not fixed, I will start a new conversation and try again. Seems like its out of tokens, and just mumbling.
June 7, 2025 at 12:37 PM
The files being modified should be locked for writing while the model plains on modifying them. At least a visual marker is a must.

A intermediate progress bar on the side panel, and a red top marking on the edited files should be enough. Basic UI/UX.
June 7, 2025 at 9:15 AM
@cursor.com.web.brid.gy understanding when the model finished "thinking" is not trivial. I find myself "hey, its finished lets run!" and everything works, but the model is still thinking.

Now, I code no longer compiles. I will need to `git reset` and start this part again.
June 7, 2025 at 9:08 AM
The UI is annoying. The extra red (old) code just prevents me from seeing the real code. The floating panel, sometimes, is above my code.

As I always need to fix what the machine gives, this just does not help. I don't understand the accept/reject concept. `git checkout` is the reject I expect.
June 6, 2025 at 8:41 PM
On the other hand, @cursor.com.web.brid.gy seems to work. It gets stuff done, but again, in some cases "look, I tried, I cannot do it" and code is messed up. It cannot restore to a previous working state.

Conclusion: git commit after each working iteration (git commit --amend is also OK).
June 6, 2025 at 8:37 PM