No grass is placed on terrain unity 3d, but the object follows the player. What to do? - Related to what, is, achieve, developer, no

My AI Python Coding Test: Surprising Results

You’ve read the memos and listened to the talks.

Trust me, I get the concern. I’m also a novelist and I’ve read accounts of other writers using AI to bang out books at a rate for which the human being cannot keep up. The silver lining there is that creative efforts undertaken with AI tend to be pretty bad.

But what about the coding side of things?

I decided to put Ollama to the test and have it write some Python programs to see how it fared.

To begin with, I decided to use a locally installed instance of Ollama, with the Msty frontend. I decided to add the frontend into the mix because I wanted it to be as efficient as possible. Although the terminal usage of Ollama is fairly simple, Msty makes some of the attributes more accessible (such as adding new models and Knowledge Stacks and using a prompts library).

Initially, I decided to use the [website] model for the first round of testing. I fed Ollama the following prompt:

Write a Python application that asks the user how many dice to roll, how many sides are on each dice, and then roll the dice the user has entered 1 Write a Python application that asks the user how many dice to roll , how many sides are on each dice , and then roll the dice the user has entered.

Guess what? It didn’t work. It looked as though it was going to function perfectly, but then it wound up stuck in a loop asking How many dice would you like to roll?

There were a few obvious errors in the code. Take a look at line 49, which is this:

first_half = ', '.join([str(result)[:half_points] for result in [website]',')[0:-1]]) 1 first_half = ', ' . join ([ str ( result )[: half_points ] for result in results . split ( ',' )[ 0 :- 1 ]]).

first_half = ', '.join([str(result)[:half_points] for result in [website]',')[0:-1]]) 1 first_half = ', ' . join ([ str ( result )[: half_points ] for result in result . split ( ',' )[ 0 :- 1 ]]).

Ollama’s output had [website], when it should be [website] That’s a pretty goofy error, but it’s easily fixed.

There’s another similar error in the line below that, which is:

second_half= [result[half_points:]for result in [website]',')] 1 second_half = [ result [ half_points :] for result in results . split ( ',' )].

second_half= [results[half_points:]for result in [website]',')] 1 second_half = [ results [ half_points :] for result in results . split ( ',' )].

After making those changes, the program finally runs.

Even then, if you enter a larger number when asked how many dice to roll, the error pops back up, only this time telling you that [website] should be [website] Guess what… that won’t run either!

I then tried the same prompt with the gemma2:2b model. As you probably expected, the code generated wouldn’t work. Again, it wound up caught in a loop, asking how many dice to roll.

If I pare the program down to simply create an app to roll random dice numbers, gemma2:2b gets it right.

I went back to each model and ran different queries to have it create various Python apps (of varying degrees of difficulty) and found it to be hit-and-miss. For instance, I wrote this query for gemma2:2b:

write a python program that accepts input for a people clothing choices and then reports what they should wear 1 write a python program that accepts input for a people clothing choices and then reports what they should wear.

The output of that query worked fine. I then ran the same query with the Llama [website] model, and the code it produced was vastly different, but it ran as well.

I added the DeepSeek R1 model to Msty, and every time I queried, the response seemed more like a long, drawn-out discussion on how to write code. What llama and gemma took roughly 30 seconds to spit out, DeepSeek ran for 10 minutes and gave me nothing I could use other than a long-winded back and forth that felt as random as it was guided.

In the end, here’s what I discovered about using AI to write code:

Start with a simple query, such as Write a program to roll a die. Test the output. Then ask the AI to upgrade the original with a query such as taking that same program and allowing it to ask individuals how many dice to roll. Test the output. Further, refine the application with another query. Test the output. Keep refining until you’re done.

Whenever I used Ollama and Msty to write Python programs with the above tactic, the results were much advanced than diving into something more complex. The other takeaway is that different models are advanced suited for this purpose. For example, skip right past DeepSeek and use one of the Qwen models (such as [website] Coder). When I attempted the same experiment using the [website] Coder LLM, things were a bit more predictable. Almost every time I used this model, the results worked. Even advanced, the code it produced was far less complicated, so it was easier to read and debug (when needed).

Another thing is not to expect perfect results. You will have to tweak things and even try out different models. I even ran across issues with Msty tanking on me, which helped me draw this simple conclusion:

The companies creating AI want you to believe their tools are as capable as you are at writing code and that is not exactly true. When you use AI to write code, it’s imperative that you comb through every line in the output and test it because, more than likely, you’re going to have to spend a good amount of time debugging.

I was actually excited about writing this piece because I’d tested Ollama and Msty with some fairly basic applications, and it performed admirably. When things got more complex, however, AI let me down.

In boardrooms worldwide, top decision-makers are likely asking their IT executives, “What more can we do with AI?” It’s a reasonable question as artif......

The global enterprise AI market is expanding rapidly, and more and more businesses are exploring AI’s potential to drive innovation and efficiency. Th......

Identity security is broken — and it’s costing companies millions.

Over 80% of data breaches last year stemmed from compromised credentials or stolen......

Why I Want to Be a Front-End Developer and How HNG Will Help Me Achieve My Goals

Last year, I made it to Stage 5 before getting kicked out. It was disappointing, but here I am, giving it another try. You might wonder why I’d put myself through this again. Well, that’s exactly what I want to share—my journey, my setbacks, and why I’m more determined than ever to become a frontend developer.

I wish I had an inspiring story to tell. Something about discovering coding at a young age or dreaming of building websites since childhood. But the truth is, I needed a career change, and frontend development happened to come my way.

Sometimes, opportunities don’t appear because we go looking for them. They find us when we’re ready. What started as a simple decision has turned into something I genuinely love.

Frontend development blends creativity with problem-solving, and that’s what draws me in. Modern web applications demand more than just functionality. They need to feel smooth, intuitive, and easy to use. That’s why companies often hire [website] hire [website] developers to build fast, dynamic interfaces.

Because failure taught me something valuable.

Last year, I struggled with time management, asking for help when I needed it, and testing my code properly before submission. Those mistakes cost me. But they also showed me exactly what I need to work on.

HNG is challenging, but that’s what makes it worth it. It pushes you to think fast, work under pressure, and collaborate with others. It’s not just about learning to code. It’s about learning to solve problems, stay consistent, and grow as a developer.

To do superior, I need to be superior. So here’s what I’m focusing on:

Engaging with the community, both learning and helping others.

Documenting my progress to reinforce what I learn.

Going beyond the assigned tasks to build real-world projects.

Managing my time well and staying ahead of deadlines.

This time, I’m not just here to complete stages. I want to grow into a developer who writes clean, scalable code and understands how to build things that last.

The tech world moves fast, and the best developers are the ones who keep learning. They don’t just code; they solve problems, collaborate, and improve every day.

That’s why I’m here again. Not just to try, but to do advanced.

Maybe next year, I’ll be writing about how HNG helped me go from a beginner to a confident frontend developer. Until then, I’ll keep showing up, learning, and taking one step at a time.

In boardrooms worldwide, top decision-makers are likely asking their IT executives, “What more can we do with AI?” It’s a reasonable question as artif......

Slow is officially the new down. That’s a major finding of Catchpoint’s SRE study 2025, with 53% of study respondents agreeing with this expression, ......

The global enterprise AI market is expanding rapidly, and more and more businesses are exploring AI’s potential to drive innovation and efficiency. Th......

No grass is placed on terrain unity 3d, but the object follows the player. What to do?

everything is working well, the objects tab is working, and the *relief tab is working too. BUT the tab with the addition of grass on the terrain does not work. I'm creating an object, but it's just not being added to the terrain. (everything is fine with the model, *the most popular unity asset on the grass). Please help!

Apache Kafka is known for its ability to process a huge quantity of events in real time. However, to handle millions of events, we need to follow cert......

In the world of programming, errors are inevitable. Whether caused by unexpected inputs, system issues, or edge cases, they have the......

He: I'm Jun. I'm the tech lead of the data platform at Netflix, doing management and workflow automation.