DaveTheFighter Game

[Mahmoud Fayed]

Hello

Game: DaveTheFighter (20 Levels)

Video: DaveTheFighter Game - Developed using the Ring programming language, RayLib and Claude Code Opus 4.6

Install:

ringpm install davethefighter from ringpackages

Run:

ringpm run davethefighter

Screen Shots:

World (1) - Levels from 1 to 5

World (2) - Levels from 6 to 10

World (3) - Levels from 11 to 15

World (4) - Levels from 16 to 20

Greetings,

Mahmoud

[Bert Mariani]

Hello Mahmoud

Seems every morning when I look ...

You developed a new game overnight !

Faster than a Speeding Bullet.

Regards

Bert Mariani

[Mahmoud Fayed]

Hello Bert

>> "Seems every morning when I look ..."

>> "You developed a new game overnight !"

>> " Faster than a Speeding Bullet" 

Thank you very much :D

All of these games (CodeRoom3D, Tank3D and DaveTheFighter) are developed using Claude Code (Opus 4.6)

The Ring/RayLib source code is 100% generated through prompt driven development

CodeRoom3D game contains 1985 lines of code

Tank3D game contains 2485 lines of code

DaveTheFighter game contains 4409 lines of code 

The numbers without comments

So, to reach this speed of development

(1) Most of these results are because Claude Code (Opus 4.6) is already a great model and can generate good enough code 

(2) Using good prompts, that explain what is required, in a few but clear sentences. 

(3) I tell the AI the next information about Ring (based on need)

(4) Trusting the tool, when you discover an error, don't try to solve it yourself, just copy/paste the error message and let the model solve it. If the model can't try to tell him more abstract ideas.

(5) The model can learn from documentation & samples of Ring code. You can while writing prompts give him links to chapters in Ring documentation or attach complete programs that are written before. In some tasks like developing extensions I attached all of the configuration files related to RingQt. 

(6) RingRayLib is improved based on using the Model, he reported many times that some functions are missing, all of these functions are added to RingRayLib. So the development becomes more faster and faster, these new functions are listed in Ring 1.26 - what is new?

To get this update run the comment: ringpm update ringraylib 

Getting this updated version of RingRayLib is done automatically if you installed (CodeRoom3D, Tank3D or DaveTheFighter using RingPM)

(7) The more we develop games using Ring, RayLib and Claude Code (Opus 4.6), the more we know about the good things the model can do and reuse it.

For example, (Generate sound/music files and add it to the game) is a common prompt that I used in different games to support sound/music. The model generates these files, update the source code to use it. 

All of these are advantages, but we have disadvantages too

(1) I started by using $20 subscription and in a few days, I switched to $100 subscription, and I think this is a problem. When using Claude Code (Opus 4.6) to develop something, programming is no longer free and depends on your time/effort and a single computer where a little kid can program using his father machine. Now using these AI model comes with a cost (of-course 100$ is not a problem for me) but we don't know the future, and how the prices of this model could grow. Also, we don't know about usage restrictions to specific countries that could happens at any time for any unknown reason in the future.

(2) I used to be able to program while being offline, but using AI, the situation is different, I need an internet service all of the time!

(3) Using AI to generate the code, sometimes gives you code/patterns that are already given to other users. This is different from using open-source code or samples from the internet. Because this time we may don't know! The AI will do this without telling you.

(4) You have to be careful about the architecture/performance. In most cases AI will not improve these things until you tell him to do so. 

(5) Some prompts take long time (few minutes) to be executed. Sometime the execution is interrupted for unknown reasons! 

But in general, I love the experience/results, and using AI for Ring code generation becomes one of my favorite tools. Ring as a language is very suitable for this purpose

(1) Ring code is short which enables the AI to use few tokens

(2) Ring comes with a lot of libraries, which enables the AI to use it without installation effort at the developer side.

(3) Ring support creating DSLs with natural commands, this advantage could be used in the future, so we write prompts in English and generate code in English 

Greetings,

Mahmoud

[Mansour Ayouni]

Hello Mahmoud,

Thank you for this feedback — it’s very interesting.

Also, we don't know about usage restrictions to specific countries 

For example, in Niger, the available time and token budget, as well as overall model performance, appear to be noticeably lower than in Tunisia. I haven’t conducted formal tests, but I suspect that developers in Europe may receive more resources, and those in the United States possibly the most.

By contrast, the Qwen and even Grok models seem consistent across locations, maintaining generous deep-thinking capabilities and extended memory budgets in both places.

These aspects are very important to me, as they reflect a certain ethical positioning. If geographic disparities do exist, this could raise important concerns and should be handled with great care.

Anthropic is (currently) the best at coding, but in the old Microsoft way that risks locking programmers — and programming itself, one of the last remaining intellectual disciplines — into its ecosystem and tools.

All the best,

Mansour

--

---
You received this message because you are subscribed to the Google Groups "The Ring Programming Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ring-lang+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ring-lang/6825ee57-64ed-4e4c-bd19-b862307e9cadn%40googlegroups.com.

[Liam OBrien]

Hi Mahmoud,

Do you create a test suite to verify that the code is working as expected or is this something that the model provides automatically? How does it handle the issue of code coverage?

Kind Regards,
Liam

[Mahmoud Fayed]

Hello Mansour

>> "Anthropic is (currently) the best at coding, but in the old Microsoft way that risks locking programmers — and programming itself, one of the last remaining intellectual disciplines — into its ecosystem and tools."

The company still a new company, while it's growing very fast and has a very important role now, we still need to wait to know things will go

I am optimistic that they will keep their success and will provide better and better services along the time

Since using AI tools we have the generated source code, I think if we follow the next steps everything will be fine

(1) Understanding and reding the generated source code, caring about the code quality.

(2) Using AI tools to generate libraries/tools that increase the abstraction level and improves productivity

(3) Using AI tools with restrictions/care about security

The second point is important, because the better libraries/tools we have, we close the gap between using AI or not using it.

Greetings,

Mahmoud

[Mahmoud Fayed]

Hello Liam

>> "Do you create a test suite to verify that the code is working as expected or is this something that the model provides automatically? How does it handle the issue of code coverage?"

When using AI to develop/prototype games, I just use manual testing after each update.

When using it to develop a library, I generate tests too that can be automated.

Greetings,

Mahmoud