Twitxy Saved

Zain Shah

Verified account This account is verified. Learn more

@zan2434

teaching machines @figma previously: @cmmnknwledge (acq) @opendoor @openai @ycombinator s13
San Francisco, CA zainshah.me Joined June 2009
3.2K Following 21K Followers 2.1K Posts

Latest posts 5

Zain Shah Verified account @zan2434
Interesting result! Astra can directly interpret and manipulate another LLM's weights to make it 6x better at math! And it's not just a benchmax - other capabilities also improve 3x.
Another scifi thing people said superhuman AI would be able to do, and frontier LLMs already can.
Rikiya Takehi @rikiyatakehi Nov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.

We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weight-sp…
View quoted post on X
2 replies 0 reposts 24 likes 2.2K views View on X
Zain Shah Verified account @zan2434
I think @typesafeai struck a very interesting chord with jev. It is both fundamentally the same technology as before, and fundamentally new in terms of how it actually gets used due to the new economics/ergonomics.
There has long been talk of cheaper, faster models being transformative for software, but the truth is that historically the options compromised capability to get there AND still weren't quite cheap/fast enough to be integrated in new places, so they still competed with frontier LLMs.
Instead, Jev is an OOM cheaper & faster, while also being just as capable, and it's still infinitely flexible and general! Just for a subset of the infinite tasks LLMs were suited for. Pretty excited to see where this goes, and more examples of this kind of disruption!
Sebastian Raschka Verified account @rasbt Sep 20
It’s easy to dismiss Jev it as “just a classifier”.

But people (me included) who have been training encoder-style models for classification for many years know they were usually special-purpose and limited in some way.

The breakthrough of Jev is that it generalizes well.

And I’d say the secret sauce is probably more in the data than in the training algorithm. (Plus a nice API design on top of it.)
View quoted post on X
5 replies 1 reposts 37 likes 5.5K views View on X
Zain Shah Verified account @zan2434
Big news everyone - we’re excited to share that @cmmnknwledge, the team behind Flipbook, are joining @figma to help build the future of creating with AI. It is so rare to find a team and product at this scale that believes as much as we do in both the unlimited capability of AI and human creativity. The future is bright and we’re excited to share what we build together soon!
Common Knowledge Verified account @cmmnknwledge Sep 15
Excited to share that @cmmnknwledge, the team behind Flipbook, is joining @figma. In our conversations with the Figma team, we found a deep alignment in a shared focus on amplifying human creativity while centering human agency. Looking forward to sharing more soon.
View quoted post on X