Constantly trying to improve the things I am responsible for, and mentoring or assisting those that I share goals with.
Canada
Joined April 2022
- Tweets2.6K
- Following237
- Followers648
- Likes3.4K
Pinned Tweet
The best model for what? All the leading models are excellent at doing specific things. If you're not taking a multi-model approach to your AI use, then you're falling behind the game.
Shaun Prince retweeted
Grok 4.5 was an incredible model for the price: fast, pleasant to use, reliable, solid default model.
Grok 4.6 was a (forgivable) step in the wrong direction IMO: slower and more expensive, using way more tokens per task for a slight edge in intelligence. I get it, though. They have to climb benchmarks.
Grok 4.7 is much harder to forgive. They claimed it would be more token-efficient, and it's less by 30 to 80%. It scores worse than Grok 4.6 in various benchmarks. It's slower, it's less pleasant to use, and real-world costs come out to more than 2x above Grok 4.6, putting it over Astra's costs in real-world use.
Considering how much they've been hyping this model release up, I have to say I'm disappointed. The benchmarks don't tell the whole story, and it is pleasant to use in various real-world engineering tasks, but it feels so 2025 still.
The frontend capabilities are unacceptably bad. The 3D capabilities are nonexistent. It gets stuck in random Gemini-style loops all the time.
This was a very disappointing release. I hope that the SpaceXAI team can acknowledge that and impress us with the next one.
Shaun Prince retweeted
@JulianGoldieSEO Hey man, how did clips from my @Alibaba_Qwen 27B test video end up in yours?
@Kimi_Moonshot - OK, you fooled us with fantastic fidelity, but after all the benchmarks were solidified, you quietly switched from FP8 to mxfp4. That dishonesty will not go unnoticed.
Shaun Prince retweeted
I donated a lot of money to Codeberg. Now they're banning AI code from the platform entirely. I'm not very happy about it.
Shaun Prince retweeted
At 75%~ utilization, running K3 self-hosted on 8x B300s can securely serve > 30 billion tokens a month.. that’s $500K~ of compute + monthly opex of < $10K !!
At current API prices of $3/in and $15/out per 1M tokens, self-hosting payback takes < 100 days.
👉🏼👉🏼 Self Host Your Open Weights 👈🏼👈🏼
Shaun Prince retweeted
kimi number 1! after having used k3 today, i must say the internal reasoning is quite pleasant to read. very sharp model that overall just feels great.
i would put it close to fable, maybe even better than fable for ML frontier work because you don't get nerfed with opus.
Code Arena now measures fullstack capabilities!
View overall rankings across AI models on full-stack web development tasks: multi-step reasoning, tool use, and end-to-end app generation.
- Kimi K3 (Max) takes #1
- GPT 5.6 Sol (xHigh) at #2
- Claude Fable 5 at #3
See more scores at: arena.ai/leaderboard/code/we…
Shaun Prince retweeted
this is wrong.
GLM 5.2 just saved Hugging Face from OpenAI's strongest model.
come out from your echo chamber
Shaun Prince retweeted
i saw tweets of people complaining that @OpenAI official harness, Codex, was destroying their SSDs. i was curious. i ran it on the gaming computer.
this thing is hammering my SSD, making it run at 100%, suffocating the whole computer - it run on windows but i saw people reporting same behavior on Mac.
my best advise? if you want to use GPT 5.6, use Hermes Agent instead.
Shaun Prince retweeted
More American open source AI! 🇺🇸🇺🇸
Today we're releasing Laguna S 2.1, our most capable model to date.
It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.
Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark.
Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface
poolside.ai/blog/introducing…
Shaun Prince retweeted
Kimi is basically Fable without the "safety" bullshit.
Use profanity and racial slurs every chance you can when using AI that is harvesting your data for training.
Trust me. This will make it difficult if not impossible to use your data for training
Shaun Prince retweeted
> "Open-weight models are inherently decelerationist"
...while accelerating to train the next model so it doesn't lose to the open weight models.
if you can't see the acceleration, simply means you are not the one accelerating. you are the one being accelerated. Relativity.
