[FREE TOOL]Cointikka: Tools for Bitcoin Holders & Bitcointalk Users

[Stalker22]

~
Is this happening only to me, or is there a bug with this feature?

Its definitely not just you.  I am seeing the exact same thing on my end.  Pretty sure its a bug on their end, likely an issue with their third-party API calls.

@Cointikka, you should probably look into this.

Also, the following link does not seem to be working...

https://cointikka.net/bitcoin-to-fiat-converter

It works fine for me.  Have you checked again, maybe it was a temporary glitch?

[Cointikka]

~

I have been working on those files and by EOD I will upload the update for you guys to check. Thanks to everyone active on this thread for guiding me and helping me find issues in the tool. After I upload the update, I will update this reply.
EDIT:
I have passed an update. The new onboarding or home page now looks like this:

I hope everyone likes it now, with the increase in size and utilizing all the space when viewed on a desktop or laptop. I did find bugs on fiat conversion and halving cycle pages. I have now fixed them. I have added a status and tool status link at the bottom of the home page and in the top right-hand corner of the home page. I will be adding it to all pages in the future. You will be able to see it within the RankUp tracker.
ROADMAP:

• Adding a blog/Insight
• Creating an ANN
• Starting a Review Campaign
• Starting a Signature Campaing

Please visit: 👉 Cointikka to review the changes.

[Dareo]

Please visit: 👉 Cointikka to review the changes.

     

• First of all, I want to say that the Fiat converter is not working. That is, it is not able to fetch the data.And other sections related to it are also not working. That is, Halving cycle, Utxo etc..
• Secondly, when you copy the link to the investment calculation results, it is copying the link to that section, but not the result link.
• Then, when I select the post count of the week in the Signature Campaign Tracker inside the Rank and Merit Tracker, it doesn't give any results.
  
• Another issue is that you have given your Bitcointalk link incorrectly in the contact section. This link is given there "https://bitcointalk.org/index.php?action=profile" , but this "https://bitcointalk.org/index.php?action=profile;u=2568422 "link should actually be given

[joniboini]

I wonder if there's any plan to add customized region for the tax filling feature? Maybe the user can define their own taxing method per their own country if you can't fetch from a reliable database on the internet. At least tool like that is rare in my area.

I also checked the DCA calculator and it seems like the max limit is a year ago or so, which is quite limiting given other tools can offer more than that. I think changing the base currency can be an interesting option too.

[Cointikka]

I wonder if there's any plan to add customized region for the tax filling feature? Maybe the user can define their own taxing method per their own country if you can't fetch from a reliable database on the internet. At least tool like that is rare in my area.

That is a good idea. I will surely look into it. I was stuck with the data fetching issue from Coingecko. I have forwarded a patch again today, and hopefully it should work now.

I also checked the DCA calculator and it seems like the max limit is a year ago or so, which is quite limiting given other tools can offer more than that. I think changing the base currency can be an interesting option too.

By the end of this week, both the investment calculator and the DCA calculator will allow the max limit to last five years. I am currently working on it.

[joniboini]

By the end of this week, both the investment calculator and the DCA calculator will allow the max limit to last five years. I am currently working on it.

That's good. I think you can also add the option of buying a specific amount of sats per month instead of limiting it to fiat only. For example, a user can target buying 0.0001 BTC per month to see how much the cost & profit they make after some time has passed. I remember someone said it's called value averaging before. I'm not sure that's the right term, but at least this feature feels quite unique compared to regular DCA calculator imo.

[Cointikka]

As asked, I have made the updates to these three tools: Investment Calculator | Tax Calculator | Fiat Converter
What Updates did I do?

• I can now confirm that the fetching of the real-time Bitcoin price has been solved.
• I have now built a database that has the last 5 years' Bitcoin prices from 15th August 2021 till today. It daily fetches the current price of the previous day
• Users can now input either USD or BTC to check their investments using the DCA or Investment calculator.
• I have added regions where the tax calculator will function, with the accepted currency and corresponding calculations made accordingly.

BTW.. @Dareo, in the Signature campaign tracker, after selecting the dates, please click the Sync button at the top. It will retrieve your data. I will improve it later, as I am currently focused on getting the blog started.
I thank @joniboini for suggesting me ideas to improve my project and @Dareo in helping me find bugs.

[ABCbits]

I read your post about AI checker feature on other thread and i thought i better share my thought there.

1. Downloading the AI model could took few minutes. You may want to show more stats (such as remaining MB or download speed).
2. Is it right to assume 2nd AI model (TMR AI-text-detector) generally or overall better than 1st AI model (ModernBERT AI-detector )?
3. What does "LABEL_0" and "LABEL_1" actually mean?



I also tried your tool with a reply i quoted some time ago on https://bitcointalk.org/index.php?topic=5456516.msg66822985#msg66822985, but the result of each detector is different.

[joniboini]

If I'm reading it correctly, your website use the models from huggingface? It seems like they're quite big, unless I misclicked them. One of them are nearly 2 GB, which is quite significant for some devices. Can I easily remove the models after I use it? It makes no sense to store the model for a long time if I don't want to check texts regularly.

[Cointikka]

1. Downloading the AI model could took few minutes. You may want to show more stats (such as remaining MB or download speed).

That is a fair point and a good suggestion for future updates.

2. Is it right to assume 2nd AI model (TMR AI-text-detector) generally or overall better than 1st AI model (ModernBERT AI-detector )?

Not really; both are shown side by side as independent models. Model 2 has been trained on a newer benchmark, whereas Model 1 is trained on an older benchmark, but you can say Model 2 is more accurate.  

3. What does "LABEL_0" and "LABEL_1" actually mean?

This is something I am working on improving. But, according to the official documentation for Model 1, LABEL_1 = AI; LABEL_0 = Human. For Model 2, it is a best guess based on an example code snippet on its official page.

I also tried your tool with a reply i quoted some time ago on https://bitcointalk.org/index.php?topic=5456516.msg66822985#msg66822985, but the result of each detector is different.

No detector is 100% reliable; that is the reason there are three, and each is trained differently. I also encountered the same problem for a couple of my replies, and if a forum member is using a writing assistant app to write here or is using translation apps, then the model detects them sometimes as AI.

If I'm reading it correctly, your website use the models from huggingface? It seems like they're quite big, unless I misclicked them. One of them are nearly 2 GB, which is quite significant for some devices. Can I easily remove the models after I use it? It makes no sense to store the model for a long time if I don't want to check texts regularly.

Yes, you have misclicked; you might have checked the folder size that contains all models, which comes to be 1.93 GB. The total size of my models is 217 MB. You can verify this by using the AI text checker once and then going to the browser cache page. If you want, I can guide you on where to check if you can tell me which browser you are using.

[joniboini]

You can verify this by using the AI text checker once and then going to the browser cache page. If you want, I can guide you on where to check if you can tell me which browser you are using.

So I assume clearing my browser cache (or specifically cache from your website) will clear the model too? Maybe it's worth adding an option to clear or auto-clear the model after the user finish using it. Some users probably don't know about cache and whatnot and stumble upon your tool just because they're looking for an AI checker tool.

[Cointikka]

You can verify this by using the AI text checker once and then going to the browser cache page. If you want, I can guide you on where to check if you can tell me which browser you are using.

So I assume clearing my browser cache (or specifically cache from your website) will clear the model too? Maybe it's worth adding an option to clear or auto-clear the model after the user finish using it. Some users probably don't know about cache and whatnot and stumble upon your tool just because they're looking for an AI checker tool.

I have worked on your feedback and the feedback given by @ABCbits. I would request you both to check the tool and tell me what I can do to make it an option for the team that does AI text checking in the forum. Thanks again for your support and suggestions.

[joniboini]

I just checked and I can see the option to clear the model from the browser cache. That's great.

I'm not sure if the feature of checking by user ID is new or not, but it seems it's down for now? I copied your UID as a test OP, but it failed to grab any posts from your post history. Shows a generic error, basically saying the service might be down due to a change in Bitcointalk format or server issue.

[Cointikka]

I just checked and I can see the option to clear the model from the browser cache. That's great.

I'm not sure if the feature of checking by user ID is new or not, but it seems it's down for now? I copied your UID as a test OP, but it failed to grab any posts from your post history. Shows a generic error, basically saying the service might be down due to a change in Bitcointalk format or server issue.

I am currently working on my product and want to demonstrate to the AI text-hunting community that it is effective. I plan to use an exisiting scraper that can extract up to 10 pages of a user’s post history. When an AI text hunter enters a user ID, the system will automatically analyze the last 5 posts to determine whether the content was generated by AI. The results will be presented in the same way as when content is submitted manually. If the AI text hunter is not satisfied with the outcome, they can use the same interface to manually copy and paste the content for further verification.

It is still a work in progress, and it will take some time.

[joniboini]

I plan to use an exisiting scraper that can extract up to 10 pages of a user’s post history. When an AI text hunter enters a user ID, the system will automatically analyze the last 5 posts to determine whether the content was generated by AI.

5 post sounds like a small sample. But I guess there's no need to burden the hardware too much if the suspicion towards some accounts aren't big. Then again, if they do use AI generated text, the chance they only use once in a week seems very low. I believe it's nice if there's an option to extend the range automatically, without needing to copy paste manually. Hope it works as intended.

[Cointikka]

5 post sounds like a small sample. But I guess there's no need to burden the hardware too much if the suspicion towards some accounts aren't big. Then again, if they do use AI generated text, the chance they only use once in a week seems very low. I believe it's nice if there's an option to extend the range automatically, without needing to copy paste manually. Hope it works as intended.

After reading your reply I have worked on the AI text checker. It took me some time to get it done but it is now complete and has been updated.
This is how it works:

When you go to AI text detector. You will see two tabs; I am certain that you have seen it before after reading your reply.
Now comes the best part. I am going to list how it works, as it makes understanding the UI simple. You need to click on the tab that says Bitcointalk User(by UID)

• Enter the UID
• Click on Fetch last 20 posts
• It will display the last 20 posts written by a user
• Then click on Check this post
• As soon as you click, the system uses all three AIs to verify the text content
• After that, it displays the result.
• After the update, the result is now evaluated using all three AIs, and a final result is shown. [/b]
  

The three AIs do their job perfectly, but I feel the third AI does its job efficiently. BTW, the first AI, which is a JSON application, cannot be considered an AI now. I am working on teaching it as of now. So if anyone wants to ignore its results, then I am okay with it.

[joniboini]

I've checked the new update, I'm not sure which first or second AI you're referring to though. For example this model[1] result seems way different compared to the other one. Based on that model, your post above seems to be flagged as 82% AI, unless I misread it. Is this correct?

Is there any plan on adding more models btw? Or maybe an option where users can utilize their own LLM to test the result.

[1] Model 2 — TMR AI-text-detector (RAID-benchmarked)

[Cointikka]

I've checked the new update, I'm not sure which first or second AI you're referring to though. For example this model[1] result seems way different compared to the other one. Based on that model, your post above seems to be flagged as 82% AI, unless I misread it. Is this correct?

Is there any plan on adding more models btw? Or maybe an option where users can utilize their own LLM to test the result.

[1] Model 2 — TMR AI-text-detector (RAID-benchmarked)

Give it a try now or whenever you are free. There are a few minor changes I need to work around. Otherwise, the tool is good to go. I have replaced model 1 with a newer version and I am still working on model 2. Thanks again for your feedback.

Edit: I am quoting my reply here for you to understand how good my AI model is now. I posted this on the Reputation board on the thread where they report AI Spam is Reported.

Hi @nutildah

I worked on your feedback, and I am excited to share that my AI detection models are far better than what you tested earlier. I am also eager to see you test it once again. I took the liberty of working on the @LoyceV post with my upgraded AI detection page. I am posting a few screenshots for you guys to try it.

BTW.. Model-2 can now detect anyone using Grammarly or a translator, as it will give a verdict that it is AI-generated content, whereas Model-1 does not. Since I use Grammarly, you will see that most of my written content is flagged as AI by Model-2.

Here is the first screenshot:


Here is the next screenshot with a detailed explanation, read section by section.

I am still working on UI refinement. As of now, I am confident that you will like it. I will post detailed info in my thread on the Project Development board later.
Thanks.

Edit: So as an example, I used your UID @joniboini to do a test wheteher the Model 2 can detect that you are using a translator and I found it works fine like a good wine. I am posting the screenshot. BTW.. After entering the UID, my page does allows me to check all your last 20 posts. I did it and checked it.

Here is the first screenshot. I would request you to give it a try with my UID  

The post where you used a translator which are now equipped with AI

The post where my Model 2 said yes is because it was written in your native language or normal English, which it cannot detect


And this one
I am working on multilingual detection, and if that gets going then you will be able to find those who are using AI on your local boards.

[joniboini]

I'm not sure about how accurate the model 2 is then. Because I don't use any grammar checker or AI to help me refine my post. Maybe there's a parameter that automatically flag my sentence as AI generated or using grammarly? It would be better to add more models to get more judgments I guess, since relying on models that can provide conflicting result would be bad. Then again, I guess the limited models cause this to happen (plus the number of samples we take).