I build AI-native products, apps & automations. Writing about tech, AI and things that industry refuses to admit | Feature your brand: →brandmysetup.lol
0-100
- Tweets 9,892
- Following 838
- Followers 20,931
- Likes 27,922
one 3d racing game. 125M tokens.
Cost: $0
built by space bunny, the stealth model nobody has claimed yet
coding: four seasons, traffic, crashes, gears, mobile controls. first playable build in 10 minutes
multimodal: don't explain a feature, sketch it. send the photo and it builds from that
fast: 90 tok/s early on
1M context: load a whole repo in one go
same tokens on claude opus 5.5: ~ $512 (~$47 with caching)
here's the game:
9
2
15
1,627
How Frontier companies make their models look smarter than they are.
- Benchmark overfitting
- Test-set contamination
- Prompt optimization
- Best-of-N sampling
- Test-time compute
- Reward hacking
- LLM-judge gaming
- Cherry-picked evals
2
9
643
Been using grok 4.7 for 2 days.
Here’s my verdict:
- makes simple tasks complicated
- wastes a ridiculous amount of time
- feels so slower than opus
- burns way too many tokens
- overthinks on every shits
- tries so hard to prove its’s smart
1
10
672
What has vibecoding done to this time?
Mfs who don’t have any idea of basic principles of programming now creating their own programming languages.
8
684
Got to know Google doesn’t give a fuck about vanity AI leaderboards.
They are moving towards what actually matters: Profitable AI Infrastructure
2
674
Give me:
Super fast internet
Unlimited tokens
Bunch of monsters
A quite room
And i’m done for the life.
6
20
730
Made this in under 30 minutes using Space Bunny.
Started with a random Pinterest screenshot.
Ended with this.
1M context.
Multimodal.
Zero data retention.
And it’s free for the next week.
At this point, AI isn’t just helping designers.
It might be coming for the whole workflow.
Designers might actually be cooked.
link - spacebunnytest.vercel.app/
12
1
35
1,745
Used makes more sense in terms of bar, potentially after a certain point having %left instead of %used could make more sense, especially if it was like 9% left, but keeping the same bar. Having a red hash over the left and %left in single digits could also add to UX in some cases
1
4
1,962
Xiaomi just dropped MiMo-V2.6-Pro and casually jumped past GPT-6 Luna, DeepSeek V4.1 and GLM-5.3.
What the hell is happening?💀
5
689
