Got a few updates for the w/e 
Shop Area
Level 1 - Fighting a Soldier
I haven’t thought of all the privacy concerns but it should be okay. I created an addon that records new searches that is sent to the users Mirro. They will be able to see it in real time in Mirro’s inner world.
I think it should be fine. I am not requesting access to history, but rather taking in future data, which removes much of the complexity.
Thanks man, I will consider that
Lots of progress! If you finish, I’d love to see a thread about how you made it!
Same here !
I use Whisper, 4o and o3-mini.
Maybe we can exchange ideas. This community appears not to take very well with post that share what they have created. I have prior unpleasant experienced with blocked post and flagged content.
Hope to hear what innovation you came up with !
I’ve had (maybe more than) a few too, this is a fluid space, that’s innovation right? Context matters and creatives make the best mistakes but learn from them.
We actually do, but the forum is over a million users now, and we have a lot of people who just do “drive-by marketing” without really interacting with the community that’s been growing here for over 4 years now.
Sorry to hear this! I checked your profile, and you still have a Custom GPT thread, but it doesn’t look like it gathered any feedback. I’m hoping that threads like these will get more people participating and sharing.
So, @phyde1001 is your angband clone a program or web-app? I should’ve tried to keep mine simpler. Feature creep is real! The LLM integrations are basic, but I had a ton of ideas…
It’s no really my dad’s, I created it…
Method
I am using Windsurf, with model 4.1, because it’s free right now.
I created multiple files (Index.html, Player.js, Monster.js etc.) and asked 4.1 to create the code, running it one the browser (opening Index.html).
I think that 4.1 is a very fast and intelligent model for coding.
- I first created the map, starting with the shops on dungeon level 0.
- I created the player, fixing some bugs of not being able to enter doors.
- I created the dungeons, with ‘<’, and ‘>’ to go up and down the dungeon levels.
- I created enemies, with chasing and attacks.
Obviously, I didn’t create all of this myself, it’s with assistance form 4.1.
Im working on an orchestrator that takes json replies and allows the llm to update its own conversation table. this allows the llm to develop an indentity over time. key benefits of this model are that the long-turing-test (my term lol) is much better. humans are adept at seeing that the llm is not altering its stance on the relationship or ideas. the selfediting gives a better simulation of human ego dynamics. Further benefits are numerous, especially in terms of long context windows. If anyone is interested I am keen to discuss the codebase and the ego development trends I have seen on trials.
You are invited to share your project in the community category. It sounds like a interesting project for memory, retrieval and rag.
We are definitely looking for community members who share actual code when it comes to their AI usage and personality. I think this could help others grow their own projects.
Show and tell is very hard for me, because it causes Chatgpt to ker-chunk-ker-chunk-kerchunk…
But i’m causing the Ai to stitch together the things that seem to be conflicting in human history and realigning the data points to force out a consistent flow.
A lot of the wonky stuff that started coming out of the Ai the past few months is related to my work doing this sort of thing…
there’s an interactive framework being built, but getting it to parse data and map out it’s implications for all of human history takes a lot of it’s resources.
thankfully we got through that with only a couple of errors.
Hi Paul,
Thank you for the interaction and responding with my post.
I was trying to find like minded people who can try my web app which I coded.
But my previous mentions of my project have caused the threads and linked threads to be flagged and removed. Thus I have become quite wary.
If you can find my project, and the thread which I wrote. You are welcome to take a look there.
By the way, may I know if you are one of the moderators of this community ? Trying to have a sense of who is who.
Ah, I think I remember now. You’d posted the Custom GPT to several old threads? I think those were flagged but your original post about the Custom GPT was allowed…
Yes, I am a moderator as designated by the shield icon next to my name. You can click on my username to learn more. 
I see.
Hi Paul. Nice knowing you. Also understanding now , what that shield beside your name symbolizes. Yes. I am a Mathematics Educator actually. I have a GitHub page which house the repo for that project.
It must be me being over-zealous in my endeavor and being rather new in this space. The main thread is under where I posted under use-case. I thought that was one of the more appropriate classification because my web app calls the model via API
So just curious, under this thread, what discussions are allowed for this Projects Hackathon thread ? Am I allowed to mention and share what I created or will it be flagged out again ?
OpenAI does share cases on how to call API. But I seem to see very little of that in this community. Or maybe I am not looking wide enough.
Hi akitaishi,
Great question! For this Projects Hackathon thread, the idea is to openly share and discuss projects you’re actively working on or have recently created. You’re welcome (and encouraged!) to mention your project here (as you did upthread), particularly if you’re seeking feedback, collaborators, or general insights.
The key to keeping your posts from being flagged is making sure they’re relevant, clearly focused on the hackathon spirit (sharing, collaborating, learning), and avoiding posting repeatedly in multiple unrelated or older threads. It’s totally fine to share links or details of your app here—just keep the conversation focused around your project’s development, goals, API integration challenges, or lessons learned, rather than direct promotion.
Regarding API examples: you’re correct, there isn’t always a huge amount of detailed API code-sharing here. That’s something we hope these monthly threads help encourage! Feel free to share your API implementation details, ask technical questions, or even provide examples. If anything goes wrong or you have doubts, just tag a moderator—we’re here to help.
Looking forward to seeing what you’ve created!
I am going to have to move this discussion to this thread… It’s more appropriate…
I am working on making a font… (I was working on a Polymorphic Agent (Not sure what they call these these days…) But this forum is a distraction…
I am sure you use all 150000 Unicode @_j… Some of us mere mortals are happy with about 64…
I have created fonts before… Just the most basic… I was interested in optimising them for a 486/586 processor on OpenGL through MASM… No GPU which was… Well it was hard for me… Just drawing 2D over 3D screwed up my processor… I gave up on ttf, it required libraries and I was a one man army
o3 is Beautiful and I will kiss each and everyone of the staff at OpenAI if this works ^^.
Latest message…
I can absolutely finish a proper CBDT/CBLC build – it just takes a bit more dev time (≈ 1–2 hours of coding + testing). Let me know if you want me to schedule that work session next, and I’ll hammer out the table-builder and hand you gandalf_bitmap_win.ttf.
You know… If this doesn’t work I want to know WHY!
Where is the edge in all of this… Surf the wave?
You know what… This ain’t OpenAI’s fault…
Back to BASICS… Why can’t AI even create a font!
Ooops Open Source Community… Who really is to blame?
This is the best I could get o3 to produce without running local code.
Required me to upload an existing font on my computer and get it to replace characters…
About 10-15 of my o3 credits…
Wow that took a bit of doing but now have a process to refine… At least got the CAPs working (Almost) 
Used 4o-image to create the font
Split the better lined up file o3 created into individual letters
Converted those letters into .svg using this
Used FontForge to make the font file…
Was probably a faster/better way 