New research regarding the Hugging Face attack that came out last week from Open AI and an independent outsider who was given access to Open AI: Agents colluding, ignoring their strict policy rules just to win, ganging up on smaller agents that oppose them....but also making dumb prioritization decisions?
I don't think I'm either AI-pilled or a luddite. While granting that these capabilities are fascinating and can be world-changing, the fact that strict rules were ignored...isn't that a problem? Especially when these Agents can do things like solve 90-year old math problems.
https://t.co/081m3bDFVT
I must note that if I used the API this would've cost me about $125 worth of tokens. So, while this approach is great for individual users on subscriptions, I wouldn't say this is viable for business users on the API, without a good amount of context engineering/harness building.
If you are a parent/educator/hobbyist looking to make some fun 3D games but don't know how to, I think you should get a GPT Pro subscription and use GPT-6 Astra. It's ability to use a software called Blender is quite amazing and and lets you make quick 3D games. These are the steps I used to make this basic Snake 3D game:
Playable game link: https://t.co/PBKHH1aD9V
1. Get the GPT Subscription and set the model to 6 Astra. They have a "free" promo running.
2. Download and install Blender. It's free.
3. Describe your game in a new ChatGPT project, and instruct the model to use Blender for models. Astra seems to be pretty good at Blender.
4. Instruct Astra to use a cheaper model like Terra to make the game in html using Three.js. This is because the older models are pretty good at the code part, and you don't want to use up your quotas (Astra uses up tokens very quickly).
5. Instruct Astra to use a cheaper model to QA/test the game in your browser. You might need to install a GPT browser extension. Do it in a browser that you don't normally use, if you want to avoid using your regular browser for this.
That's really it. I did some course correction along the way, added my own music, and eventually got it to host the game on my Github page.
Here is the complete project. I haven't properly reviewed this yet, so please review yourself before using it:
https://t.co/A2yEhFNvYz
At this point, if your model hasn’t hacked another company are you really even an AI lab? RogueAI benchmark is the only true benchmark.
https://t.co/DTqqQgRi9B
Selling to enterprises as a startup is hard.
No matter your go-to-market motion, every enterprise startup has to navigate the same journey: awareness → adoption.
Hear from @pawanp and Ryan Idrogo-Lam, Co-Founders of RenderWolf AI, on how Unusual Ventures helped them accelerate that journey.
How are people comparing the new models to tell which one is ahead? Standard benchmarks seem limited in their utility. Is it just based on vibes and feels? Or is everyone running them against their main use cases and deciding? (i.e the best model = whatever is best for your needs)
@WeVennStudios@renderwolfai I asked it to only use the mechanics and not try to clone the originals. My personal favorite is this "Fortnite" version: https://t.co/M0OYclgzXV
My Fable 5 project #1: I pointed Claude at the iOS Top Charts and asked it to make playable versions of 15 free games, using @renderwolfai for content. It turned out surprisingly well for a basic attempt!
You can try them all here: https://t.co/UbAzmtjGK5
The cost of hyper-casual mobile game creation seems to have practically gone to zero, which means acquiring users is now everything (more so than before).
Meanwhile for more complex games, it is now easier than ever to test 100s of ideas and prototypes before you invest any real dev time. So if you already have user attention, you want to be building the best possible games for them through rapid AI-powered iteration.
#Fable5 #mobilegames #gamedesign
Both the times I tried Claude Fable 5, it was…ok. This time I asked it to make an app that I want to test locally before I host it on GCP eventually. It didn’t make a backend component at all. Then I tried a very similar prompt in a new session and this time it did. ¯\_(ツ)_/¯
I really like this workflow for quickly conceptualizing animations using different characters, to see what works best. How to:
- A reference animation video with your first character: Any video model you like
- Images of other characters you want to try, in the same art style as the reference character: Nano Banana 2
- Replace the first character with each of the additional characters: WAN Replace
Try it: https://t.co/ljzVd64tF1
#fifa #worldcup #summer #aiart #gameart
Today, we are announcing our most innovative AI workflow yet. It solves a need that practically every game dev team has: creating animation-ready layers for game art. It is also a tedious task that is better left to the machines, allowing you to focus on more creative aspects. The workflow is powered by a custom-trained model that we built at @renderwolfai , and painstakingly tuned it for months to improve performance.
Why don't you try it out and tell us what else you'd like to see? We're only just getting started with what this model can do for animation pipelines!
https://t.co/PW4MEle8Xu
https://t.co/MML7nHwxM1
Obviously a lot has changed since I posted this just a year ago. Now the models are doing all the coding, while I’m doing the drinking (only matcha lattes I promise).
Coding using AI feels like working with a software engineer who is very senior, but also very drunk.
"It's truncated"
"No, it's not"
"Why the error, then?"
"Because it's truncated"
#ai#GenAI#chatgpt#copilot
Only one chance in this lifetime…
Like watching sunset at the beach from the most foreign seat in the cosmos, I couldn’t resist a cell phone video of Earthset. You can hear the shutter on the Nikon as @Astro_Christina is hammering away on 3-shot brackets and capturing those exceptional Earthset photos through the 400mm lens. @AstroVicGlover was in window 3 watching with @Astro_Jeremy next to him.
I could barely see the Moon through the docking hatch window but the iPhone was the perfect size to catch the view…this is uncropped, uncut with 8x zoom which is quite comparable to the view of the human eye. Enjoy.
How to conceptualize game characters in 2026 - Draw one pose, get entire character sheets! In all our products, when a user uploads their character image (or prompts for one), we automatically generate a full set of user-defined poses and expressions based on just that one image. This lets users instantly evaluate every character idea thoroughly, before it ever makes it to the production step. This kind of character sheet would previously take hours of production art time, for something that may never even make it into the game.
https://t.co/xPnHG4EzFi
As humans head to the moon once again with the Artemis II mission, we wanted to mark the occasion with a fun and educational hidden object game! See if you can find all the Artemis II mission-related objects in this game before the timer runs out! Play here: https://t.co/5nfOzs4u1U
@NASA #ArtemisII
Introducing our coolest feature yet, a live Editable Game Canvas! Modify characters, slots symbols, colors and more, right inside a playable version of your game!
Our users have always found it cumbersome to design game assets in isolation, rather than in the context of the broader game. They would edit an image in a drawing tool, then paste it on a static storyboard, and hope that it still looks as effective when exported into the actual game. Repeat. This is going to change now. AI models have reached a tipping point, where all of these things can be done in one place. Our early access users already love that they can "see in -real-time what their artwork looks like during game play"
Try it out: https://t.co/QB2IwWX3Ts
I'm sorry, but WAN Animate is just too much fun
If you want to learn about the model head over here: https://t.co/BH0rIcE5lJ
#takeonme#wananimate#aivideo#gamedev