Video Game Coding with OpenAI o1

[samuelepesce]

This is the only video that shows how useful the thinking process is.
Still, it's an awesome upgrade

[felixb.]

At this point we will be able to program GTA 6 ourselves before GTA 6 💀

[Arron_Mottram]

I'm actually very excited about the new model because GPT-4o couldn't properly imagine what the game it encodes looks like. When I asked it to fix one mistake, it made several mistakes in another part of the code, and so on indefinitely

[ricobuilds]

🥺 to all junior devs...

Mission Failed. You'll Get 'Em Next Time
(jk, there's still plenty room to make it bro/sis)

[pcode]

Welcome to OpenAI new model cycle!
1. AI Influencers starts to predict end of software devs jobs.
2. Hype fire is big again
3. Model show its real possibilities and is tested by people
4. Model starts to make funny mistakes (It will allways do, its statistical machine, not true inteligence - ask model for it yourself)
5. Everything goes back to normal, till new model.

For me its already boring :D

[philipwilkin]

That is insane.
Even with GPT 3.5 and 4 I am able to write simple code well enough to get it to run (I have barely any coding experience). Looking forward to this update.

[MildlyAutisticApe]

Coding devs everywhere sweating profusely “Don’t worry boys it’s just hype”

[anonymouskerman]

Like 501!😃 I hope that Open AI will focus on the OPEN a bit more in the future.

[IndianTinker]

"Thinking" is an interesting choice of words

[kirillsemenchenko]

I built a full ‘Vampire Survivor’-like game with level-ups, weapons, bosses, etc. in ten minutes. That’s really cool.

[garfield]

Claude did this in 6 seconds. Without thinking. This was an extremely bad example.

[KristianTheDesigner]

This is very exciting! One thing though, I hear the word "think" alot in regards to o1. "It thinks before it gives the result." I would like to get their (Open AI) definition of "think" in this case.

[MilezAwxy]

in case you haven't noticed, current chatgpt version will give you a completely different version of code most of the time if regenerate answer. (you can try it) which isn't supposed to happen. sure maybe v1 and v2 slightly different in minor things like attributes... but the current chatgpt seem to make wild guesses and see what sticks by modifying huge chunks of code. this seems like is going to be the solution.

[Macala]

I’ve been teaching a middle school elective class trying to get students to design their own python games with ChatGPT. We started about a month ago. The struggle is just getting them to be creative! They I’ll show this video in our next class.

[coolandamazingchicken]

I was actually wondering yesterday when they will make coding better

[nicheva]

For non-dev's - that part where you're like 'I'll just jump across to this other program and run it' is a barrier to entry to most people. They don't know where to run the code or what file format or even if you get it all working, what to do with it then etc - You need a 'The thing is working - post to market button' otherwise all these games will sit forever on people's desktop.

[VoodooDg]

sadly thats all it can do. as soon as u need 2, 3 or more classes interacting with one another all the llm's fail miserable. even with cursor.

[lordlucifer]

Sounds good in theory, but with just how limited anyone is with usage on all your good so called openAI models, I don't think anybody is going to be doing any big projects that are going to require a good amount of back and forth input and output. I don't think 5-10 responses are going to do justice.

[PocasterClips-]

Boy o boy, as a cs major sophomore, i am truly not be able to enter a cs field after i graduate. Gotta switch to data science

[realharo]

Tbh I thought GPT-4 could already do this.