Passionate about AI, even more if it's local. Automating my life.
Joined August 2026
- Tweets80
- Following12
- Followers25
- Likes43
Oooh oh. @grok how does it compare to GLM5.3 flash with these numbers?
Introducing SWE-2 Flash, our new fast & efficient coding model.
On leading evals, it outperforms all American open-weight models, even ones 3x larger.
To balance transparency and cyber risks, we are building a program to securely share the weights with partners & researchers.
1
33
Impressive. Would need to test this.
Today we release Ripple-65B Preview, our 65B dense LLM we pretrained from scratch.
Ripple sets a new state-of-the-art for U.S. open-weight models in agentic coding tasks, surpassing significantly larger models including Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B.
Learn more: ricursive.com/ripple
19
Before you give an AI agent your email and calendar, ask 3 things:
Where does my data live ?
Who can read it ?
Can I take everything and leave ?
If the answers aren't written down somewhere, that's an answer too.
1
1
35
Love asking this question when I switch models, to see the way the model “thinks”.
I asked GLM5.3 flash and Deepseek v4.1 Flash 1017.
“How does it feel knowing you live in the 2 ASUS GX10s on the left?"
Deepseek seems more philosophical, GLM more friendly.
With the picture of my tiny homelab.
1
1
89
My passion for local AI brought my passion for coding back to life.
Nowadays, if we feel it, the only limit left is imagination. We can ship things every single day.
I got really inspired from latest weeks from the work of @MiaAI_lab @plotarmordev @ashxhart, thanks for shipping amazing stuff.
Now I'm cooking something I care about. Built for people, and for their families.
Hope it can be available in the next 1 to 2 weeks. Stay tuned.
1
2
91
Well it ain’t much compared to some crazy folks there.
2x DGX Sparks with @MiaAI_lab recipes, 1x Framework Desktop with video / audio models and Hermes.
Bought few months back… now I would need to somehow up my insurance and make sure my home security is 24/7 up and running 😂
1
66
@MiaAI_lab
GLM-5.3-Flash on 2x DGX Spark: quantized the dense side + DFlash2 draft + lm_head to EXL3 on top of @MiaAI_lab's E2 recipe. 65→79 tok/s (+22-26%), 97.6% top-1 fidelity, acceptance intact.
github.com/Alexbob0/glm53-fl… · upstream: issue #124
2
97