[TOOL] ⭐Bitcointalk BRDb score

[Gragebox]

We still are a t the complete beginning of the AI revolution. AI are still in their infancy, this is why there are the problems that we al recognise. But if you think at the rate of improvements of these systems, it will be difficult to even imagine where we will be in 10 years.

It feels like we are only on the precipice of the AI revolution. Everything happening now seems par for the course as we go through the growing pains. It seems more shocking how rapidly this is progressing. Imagine how far ahead AI would be in 10 years, if it continues on at this current pace, then the real work will be what we do with it, and how we adapt to it.

[Lillominato89]

We still are a t the complete beginning of the AI revolution. AI are still in their infancy, this is why there are the problems that we al recognise. But if you think at the rate of improvements of these systems, it will be difficult to even imagine where we will be in 10 years.

It feels like we are only on the precipice of the AI revolution. Everything happening now seems par for the course as we go through the growing pains. It seems more shocking how rapidly this is progressing. Imagine how far ahead AI would be in 10 years, if it continues on at this current pace, then the real work will be what we do with it, and how we adapt to it.

I’m convinced that in 10 years’ time, artificial intelligence will have advanced so far that it will replace many jobs that currently require human involvement.
They’ve already made huge strides in the medical field.
This is what the future will look like; this is the path they intend to take.
And I use AI to draw silly pictures 

[*Ace*]

You’ll find loads and loads of bugs 
I worked on BRDB all night today, but I didn’t manage to sort out much. I’ve completely removed every reference to BitList, as I’m no longer using the BitList APIs.
I tried scraping the profiles with my scraper but something’s gone wrong; I’ll have another go tomorrow if I’ve got time.
I don’t use Claude, but I’m using Qwen Coder for help and I’d say I’m getting good results

I am not using any AI toold for coding so they are all the same for me 
Problem with them fixing bugs is that they are creating new bugs in the process, this is what I heard from some developers recently.
Be careful, and never really or fully trust any of this tools.

Man, this is so true! 
Previously, when I made a simple expense manager app using Codex, the same thing happened with me. There were like hundreds of errors/bugs.
One time the UI is messed up, one time the login button wouldn't work, another time the whole calculation logics breaks down.
Every time I fixed one bug, two more would take it's place! So, it took me more than 30-40 builds/versions before I could have perfected my app.
AI tools are really powerful that I admit, but what experience told me is that when you start using one you need to hand feed him everything.
Do this, do that, be careful not to touch that, don't go there, it needs to be done this way. The more detail you explain it things, the better it works (though it still misses a few things).

Lol This is exactly what happens to me every time I rely on deepseek or Qwen coder. The problem is that you have to give specific commands limited to that action, otherwise it makes unwanted changes that then maybe create further bugs. Anyway, I should have some time tonight and I'll get back to fixing the project! Intent data collection continues smoothly

[Gragebox]

Lol This is exactly what happens to me every time I rely on deepseek or Qwen coder. The problem is that you have to give specific commands limited to that action, otherwise it makes unwanted changes that then maybe create further bugs. Anyway, I should have some time tonight and I'll get back to fixing the project! Intent data collection continues smoothly

DeepSeek and Qwen can both be excellent tools for writing AI code, but you really need to tell it exactly what you want within some limits, or it will start changing pieces of your project that should never have been touched. Often, one change can lead to yet more bugs, and it can become a struggle to track things down with many small, unplanned, changes. My strategy is now to give one specific task, tell it exactly what I want it to change (and, just as importantly, what it definitely shouldn't) and then to run the code and see what I need to fix in the next round. The intent data is coming in fine, and once the code can be stabilized, it will be interesting to see how that data turns out to be useful for project improvements.

[*Ace*]

Last night I made some more new changes.
I removed all references to bitlist, removed the global feed Tab which was consuming a huge amount of worker resources, It exceeded Cloudflare's free plan limit by about 1400 times.
I then started fixing profile scraping directly from bitcointalk, every time you write a post or send or receive a merit, your profile will be updated. At least I hope it works!
There are still bugs to be solved, but I'm slowly fixing them.

[fillippone]

I am starting to think that this tool might have a true value in detecting farming accounts, trying to elaborate patterns over posts and merits.
One cool feature would be trying to understand if there are “big feeders” in merits from a single account.

[DYING_S0UL]

Lol This is exactly what happens to me every time I rely on deepseek or Qwen coder. The problem is that you have to give specific commands limited to that action, otherwise it makes unwanted changes that then maybe create further bugs. Anyway, I should have some time tonight and I'll get back to fixing the project! Intent data collection continues smoothly

Where/which platform are you using to run them exactly? I mean are they like subscription based? Or locally runnable LLM models? I personally, used all of the types. Right now I have a couple downloaded in my machine which I run locally. They aren't perfect but they are capable enough for small simple tasks. And when they can't fix my problems anymore, I turn to online version of Claude or Chatgpt!

These are the locally available LLM on my device (accessible through LM Studio or Ollama). Since I have a high end GPU, they don't give any pain.



As for giving specific commands, I learned that their thinking should be kept at normal/medium mode. If you set it high/extreme, they would go above and beyond your instructions, making something entirely different that you never wanted. And keeping it low has it's disadvantages also. So I just keep it to medium mode.

I am starting to think that this tool might have a true value in detecting farming accounts, trying to elaborate patterns over posts and merits.
One cool feature would be trying to understand if there are “big feeders” in merits from a single account.

In case you didn't notice, someone else made a similar tool, that in short can be describe as an analyzer for accounts. It's a cool project that can find certain patterns (seems still in developments).

Multi-Account-Scanner by jahead