Developing projects with AI

[pooya87]

I only recently started trying DeepSeek web interface to get some ideas how to solve the problems I am facing in a rather complex code I have been trying to write for a trading bot. My conclusion so far is that It can be useful to give you some ideas how to solve your problem but I don't think it can solve it for you. It is just like an assistant or a tool you can take advantage of.

The code was based on .NET Framework 4.7.2, even though the latest version is .NET Framework 9.

Unless the video conversion library it was using didn't support 4.7.2, this framework target seems like a logical decision considering if your code can run on .Net framework 4.7.2 it will definitely run on dotnet 9 but the opposite may not be true (eg. if you use something like Span or new language features like Foo foo = new(); it won't run on older versions).

FWIW it is best to target dotnet 8 since it is LTS version while 9 is STS expiring sooner than 8.

This warning indicates that the event name btnBrowse_Click does not follow C# naming rules. According to naming rules, event names must begin with capital letters.

The thing about naming conventions is that there is no universal one. For example I used to use this exact format for UI element names a long time ago (btnSomething, txtSomethingElse,...). It is a terrible convention but I used to use it nonetheless, specially back when I played around with VB.Net language and had migrated to c#.

I'd say the biggest problem here is that it is using code behind. MVVM pattern is the only thing I approve even for tiny app that is supposed to do a tiny job
I am addicted to decoupling code...

[Vod]

Other vague ideas in back of my mind make me think maybe the AIs also likely cannot do things like go log in somewhere to do something, right? Some kind of prevent them being used to attack other servers etc protection / ethics?

Yes, AI has moved from just talking to doing.   The term is "AI Agent", and it can do things like read a spreadsheet, call numbers, address people by name, log into a website, change a password, etc.  I've bever been to the dark web (don't use TOR) but I'm sure from what I read that you can buy such agents there.  Probably still rudimentary but they are improving daily.    Automated tools for attacking servers and such have been out for years; it's the person with the password they are attacking now.   

I only recently started trying DeepSeek web interface to get some ideas how to solve the problems I am facing in a rather complex code I have been trying to write for a trading bot. My conclusion so far is that It can be useful to give you some ideas how to solve your problem but I don't think it can solve it for you. It is just like an assistant or a tool you can take advantage of.

I left DS for ChatGPT a few weeks ago.   Since I live in hockey country, the speed is better and it has a useful UI that lets me tell the AI how I want it to answer.  It is fairly good at understanding what I want, and it will put everything in completed files in the proper structure so I just need to drag to my extension directory.  Saves a lot of time in that aspect.   However every time I think I have it chugging along nicely it silently makes something "better" without telling me - I've mitigated this with instructions to ask, not assume but it still happens.

I don't think I will ever finish my SP project using only AI.   I can spend a day working with it only to have to restore from backup and start again.   I'm thinking I should find a willing partner who wants to share in the ranking excitement! 

[yhiaali3]

I only recently started trying DeepSeek web interface to get some ideas how to solve the problems I am facing in a rather complex code I have been trying to write for a trading bot. My conclusion so far is that It can be useful to give you some ideas how to solve your problem but I don't think it can solve it for you. It is just like an assistant or a tool you can take advantage of.

The code was based on .NET Framework 4.7.2, even though the latest version is .NET Framework 9.

Unless the video conversion library it was using didn't support 4.7.2, this framework target seems like a logical decision considering if your code can run on .Net framework 4.7.2 it will definitely run on dotnet 9 but the opposite may not be true (eg. if you use something like Span or new language features like Foo foo = new(); it won't run on older versions).

FWIW it is best to target dotnet 8 since it is LTS version while 9 is STS expiring sooner than 8.

This warning indicates that the event name btnBrowse_Click does not follow C# naming rules. According to naming rules, event names must begin with capital letters.

The thing about naming conventions is that there is no universal one. For example I used to use this exact format for UI element names a long time ago (btnSomething, txtSomethingElse,...). It is a terrible convention but I used to use it nonetheless, specially back when I played around with VB.Net language and had migrated to c#.

I'd say the biggest problem here is that it is using code behind. MVVM pattern is the only thing I approve even for tiny app that is supposed to do a tiny job
I am addicted to decoupling code...

Yes these small details sometimes seem annoying even though they're nothing.
However, I found another AI that's smarter than DS at least in terms of writing code with a small margin of error and capable of performing the exact tasks you ask of it.

Try this improved AI version from anthropic.com, claude.ai version 4.
I tried writing the same program, and the results were truly amazing. A program that would have taken months to write was created in just a few minutes, almost error-free.

I optimized it and tested it on Windows 11 and the results were truly amazing.

The problem is that the version is still beta and the free version is limited, making it difficult to write large programs.

[Vod]

I have completed my experiment to develop a project with AI.   The experiment failed to produce a working product, but it did give me valuable experience and insight as to how AI will be used in projects in the future.    

I can compare my experience with AI like a long bling date.   At first I was very impressed with what AI did for me and the ideas it gave me I wouldn't have thought at on my own.  But as I dove deeper into the individual subjects, trying to get to know my date better, it would often show the limits of its knowledge.    If I take one of its ideas and want to try something different, it will follow my instructions, while assuming everything (other than my explicit changes) is the same - it always wants to go back to what it knows.  

And the environment issues - what a pain!   I understand that strong competition means constant improvement, but almost 25% of the time chatgpt has an issue with accessing or providing documents it has used.   I had to give the AI one task at a time, save the task results, and them resubmit those results when asking for changes - ONCE I had figured out it was broken.

For example:  I needed to determine how much I would have made had I kept BPIP through today.  I had to first get a history of bitcointalk traffic, which I got from the ad stats.  Then I had AI compile historical BTC/USD rates.  I now have this information in documents so that the AI cannot have a "session seizure" and ruin a day of programming.   I keep a simple notepad file with the specific instructions to create what I need.   Then I have AI run through various growth scenarios, based on the bitcointalk traffic BPIP was tied to.  The best part is with these three files, I can return to chatgpt anytime in the future and pick up exactly where I left off.

AI courses/boot camps are going to replace university degrees yesterday.  Don't fight with the AI prompts to get what you want.   Have the AI generate the prompts for you.   Less than 15% of IT professionals identify as AI literate, so the future is still wide open, even if you are starting from day one.

It's a great time to be entering IT, and a great time for those who like designing but not coding.  

[Panxora_COO]

Hi, I am one of the founders of Creativeguru.ai also a GP in the Panxora Digital Asset Hedge Fund. Have you thought about using Autonomous AI Agents for identifying investors and community building? My partners and I are specialists in this field. If you’re curious too let’s schedule a brief chat to see if there’s a fit. https://calendly.com/marcie_terman/creativeguru-intro

[Vod]

Hi, I am one of the founders of Creativeguru.ai also a GP in the Panxora Digital Asset Hedge Fund. Have you thought about using Autonomous AI Agents for identifying investors and community building? My partners and I are specialists in this field. If you’re curious too let’s schedule a brief chat to see if there’s a fit. https://calendly.com/marcie_terman/creativeguru-intro

Panxora,

I would have marked your post as spam - you are a Senior member but you didn't bother to read the thread.  I'm not looking for anything - I was trying to build a project using only AI and it failed.  

[Vod]

Let me go ahead and save you some money then  
Here is how every interaction will go:

User: Give me solution
ChatGPT: Here is the solution
User: This is not the solution
ChatGPT: You're absolutely right!

Dude/Bro/Sir,

 

That is exactly the experience I had with chatgpt. 

Me:  Make just that change and don't assume anything.
GPT: Makes change and wipes out a procedure for no reason
Me:  You assumed I didn't need that procedure!  Put it back!
GPT:  You're right.  Can you upload the original so I can restore it?

I ask it a lot of non-programming questions, so I'm not going to cancel my subscription, but I am done using it for coding.   Sitting at my computer tonight getting frustrated as usual when it suddenly dawned on me - I still have non-AI tools. 

[KWH]

Dude/Bro/Sir,

 

That is exactly the experience I had with chatgpt.  

Me:  Make just that change and don't assume anything.
GPT: Makes change and wipes out a procedure for no reason
Me:  You assumed I didn't need that procedure!  Put it back!
GPT:  You're right.  Can you upload the original so I can restore it?

I ask it a lot of non-programming questions, so I'm not going to cancel my subscription, but I am done using it for coding.   Sitting at my computer tonight getting frustrated as usual when it suddenly dawned on me - I still have non-AI tools.  

"Creators" making AI music and leading people to believe they are an actual band, fake memes with fake attributes, fake is everywhere, and we haven't seen anything yet.
Using your brain is still king, at least until AI becomes so seamless it will be hard to see the real from manufactured. At that point, what will AI need us for?
Food for thought.........

[pooya87]

The way AI keeps telling you "You're absolutely right" is annoying sometimes

I was doing some video editing with Adobe Premiere recently and wanted to know how to add camera shake so I thought AI could help. But it kept giving me instructions to click on options in menus that did not exist and each time I told it I couldn't find the option, it just respond "You're absolutely right. The option is in this other menu instead" and that other menu didn't exist either. That was an endless loop

[Vod]

I was doing some video editing with Adobe Premiere recently and wanted to know how to add camera shake so I thought AI could help. But it kept giving me instructions to click on options in menus that did not exist and each time I told it I couldn't find the option, it just respond "You're absolutely right. The option is in this other menu instead" and that other menu didn't exist either. That was an endless loop

Yeah, the AI has been trained on pervious UI versions.  I can solve this issue by asking the AI to refresh its documentation to the latest version.    It can actually do that.

But the annoying thing is, you ask it to do that, but the next time someone asks your same initial question, it will go back to it's original (outdated) documentation.   AI does not add it's conversations to its LLM, and that's a good thing for PII.  But it's something that will eventually be solved by our overlords. 

[Catenaccio]

The way AI keeps telling you "You're absolutely right" is annoying sometimes

I was doing some video editing with Adobe Premiere recently and wanted to know how to add camera shake so I thought AI could help. But it kept giving me instructions to click on options in menus that did not exist and each time I told it I couldn't find the option, it just respond "You're absolutely right. The option is in this other menu instead" and that other menu didn't exist either. That was an endless loop

It's not annoying but truly scam and "AI" word is used popularly like a buzzword.

Months ago, Binance developed tags on projects and many projects had AI tags while we don't know what they actually develop and cook behind the scene. Soon, the coming bear market in 2026 and 2027 will kill many AI projects but even earlier than 2026 and 2027, this year already shows us many low quality AI projects with prices have dumped many times from their ATHs last year.

AI is a buzzword and now if people see new projects in cryptocurrency market with AI advertisement, it's safest to ignore those projects. It's more risky than the last year as only several more months before a terrible bear market with altcoins.

[noonriver]

Building with AI is wild right now tools are moving so fast that projects go from idea to prototype in days instead of months.

[Joel_Jantsen]

Fun Discussion! Here is my take on AI/Vibe Coding since I've been playing with it for a few months now:

- Context is KING
Yeah, it's a cliche but having a great context helps models to come up with better responses. I'm not taking about "think like a senior engineer" or "come up with a graceful solution" rather, the code follows "xyz" design pattern and module "x" is the service layer while module "y" is a view and we follow OOPS and abstract classes and so on! It is almost accurate every time if it can learn from the existing large codebases. The challenge comes down to the pool of data it has knowledge about. Models can easily spit out React Components but struggle to come up with an Enterprise grade Java code. It can definitely pick and learn from your existing codebase provided you're an expert yourself !

- AI Assistant != AI Engineer
The current models are assistants the best and not fully functional agents that can take over a senior engineer or a software architect's job.All the ads show you some demo apps or 2d games made by AI, which is no doubt AI is good at! However, when it comes to real life challenges, it doesn't write stuff the way senior engineers would write. A senior engineer may use AI to increase productivity and deliver more stuff since AI is practically a code completion tool for them because they understand the code and guide AI to do what they want.

[robelneo]

To continue this discussion, a number of better AI engines have launched just since my post!   

Going back to this topic, there are now thousands of new AIs. I stumbled upon an AI SaaS type directory, but alas, the developer only accepts PayPal, which I discarded long ago, and I am still waiting for my friend to go online to trade PayPal. There are so many opportunities in AI that you have to look further; there is AI in almost any venture that you can think of.
I have two already, and I am still exploring the ideas. AI is an opportunity not to be missed.

[Vod]

AI is an opportunity not to be missed.

AI is an opportiunity that won't let you miss it. 

[Owena]

I've read many times that artificial intelligence gives illogical answers, but you have to know that the internet is also full of nonsense and incorrect information from which AI takes data and learns. Developing AI projects involves using machine learning, natural language processing, and other AI techniques to solve real-world problems. The process begins with defining clear goals, such as automation, prediction, or optimization. Data collection and preprocessing are essential for training models that can learn from past examples. After selecting the appropriate algorithms, the models are trained, tested, and refined to achieve accuracy. Integration with existing systems or platforms is essential for practical application. Collaboration between data scientists, engineers, and domain experts ensures successful implementation. AI projects can span industries such as healthcare, finance, and entertainment, increasing productivity and decision-making capabilities.

[WhyFhy]

I've recently been using a tuned version of QWEN3 Coder 30B and I have to say I'm pretty amazed.
I haven't even had time to plug it into VS with cline or anything, I keep going down rabbit holes during testing. 

[Vod]

Thx to TryNinja, I am now playing around with Antigravity, which I did not know from a clone of VS Code.  Like many of us, I've moved from ChatGPT to Gemini, mainly (for me) because it allows you to work "indirectly" with github.  Antigravity is now promising to take development even further - it can troubleshoot a SQL query in one agent while reviewing load times of CDN vs static in another, all using the same token history.

I'll actually have my extension released this year (deadline is Jan 14) and I basically threw away all my old code and started from scratch.  It's still annoying, but I told it to stop apologizing, and it has.  

edited for clarity

[lornadane]

I’ve experimented with AI for on-chain data analysis. It’s awesome for spotting trends, predicting congestion, or summarizing mempool activity. Definitely makes project dev easier if used responsibly.

[yhiaali3]

Thx to TryNinja, I am now playing around with Antigravity, which I did not know from a clone of VS Code.  Like many of us, I've moved from ChatGPT to Gemini, mainly (for me) because it allows you to work "indirectly" with github.  Antigravity is now promising to take development even further - it can troubleshoot a SQL query in one agent while reviewing load times of CDN vs static in another, all using the same token history.

I'll actually have my extension released this year (deadline is Jan 14) and I basically threw away all my old code and started from scratch.  It's still annoying, but I told it to stop apologizing, and it has.  

edited for clarity

Speaking of AI apologies and Google Antigravity, I heard this unfortunate yet amusing story:

A developer reported that their entire Data (D): drive was accidentally wiped out by an AI tool developed by Google (Google Antigravity). At least the AI ​​agent was very apologetic and admitted its mistake: "I am deeply, deeply sorry. This is a critical failure on my part."😊😭

A developer who said he had their entire Data (D): drive wiped out by accident by an artificial intelligence tool developed by Google said that at least the AI ​​agent was very apologetic about it and admitted its mistake.

“I am deeply, deeply sorry. This is a critical failure on my part,” the AI ​​tool admitted after confirming the developer never asked it to delete its entire D: drive.

https://www.newsweek.com/google-ai-accidentally-deletes-hard-drive-data-antigravity-11169711