Builder / Ai
England, United Kingdom
Joined June 2025
- Tweets433
- Following637
- Followers220
- Likes921
Bro are you claiming you can code better then opus 5.5? If you are then you must be better then the 99.9 % of coders in the world
It’s a shame most people delete their wrong predictions but I’m going make it so we can track it even if they do
loo retweeted
Don't use this
Build your own software using their tokens
So when their tokens become too expensive
You can switch to your own GPUs
Your compute. That you control
So you yourself can't be controlled
You need to control your stack -- or you become enslaved
I feel like a lot of peoples usage resets today anyway . Seems like a cheap reset from Tibo. Even the last one he said was suppose to be a banked one and it wasn’t
Everyone’s clowning on space x rn but they don’t realise . Space x wins by far on compute! Who will be ahead in 1-2 years is almost certainly who has the most compute
loo retweeted
lmao 2 years of building in stealth and your moat lasts 2 days
介绍比Jev快50倍，在你设备上跑的laya-mlx!
只在你的设备上占用最高1G内存
Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统
我将其移植到MLX,并且做了一些性能优化!
视频中就是这个模型在我的本地M3Max上玩贪吃蛇
这个模型能够以每秒决策60次的速度玩贪吃蛇！
github.com/mizorewww/laya-ml…
I imagine it’s such a small amount of people left in the world that can code better then Astra / Fabel
loo retweeted
So I tested DeepSeek V4.1 Flash on real tasks. 2 repos, 105 hidden bugs, find and fix what you can.
Opus 5 (max): 27
Grok 4.6 (max): 27
DeepSeek V4.1 Flash (max): 24
GPT-5.6 Luna (xhigh): 23
Opus 5 (high): 21
A really strong model for everyday tasks. And look at the cost:
Opus 5 (max): $51.33
Grok 4.6 (max): $16.96
DeepSeek V4.1 Flash (max): $1.80
GPT-5.6 Luna (xhigh): $2.50
Opus 5 (high): $38.77
It's at the Pareto frontier.
Other effort levels (high) and extra tests for DeepSeek V4 Pro dropping every hour in this thread 🧵
loo retweeted
NVIDIA is bringing CUDA to Windows on Arm ahead of RTX Spark laptops arriving in October.
CUDA 13.4 lets developers start preparing their applications for the new machines.
NVIDIA and Microsoft are positioning RTX Spark PCs for local AI models and personal agents. Getting existing CUDA software ready is a practical step toward that goal.
GPU acceleration still requires compatible NVIDIA hardware. Existing Arm PCs can help developers begin porting.
loo retweeted
OpenAI might actually kill Anthropic.
GPT 6 Astra uses 21K tokens per task. Fable 5.1 uses 64K. Same price per token, three times the output for the same job.
That is why a two hour Fable task takes Astra 20 minutes. And it is why your Claude Code limits die in 30 minutes.
You are not using Claude too much. Fable 5.1 just burns three times the tokens to do the same thing.
OpenAI built the model that does not need them.
The end of Anthropic starts with a chart like this.
AGI Astra takes control of my paint app for a masterpiece! See token costs at the end of the 1 minute clip
