Khaled retweeted
Introducing EmbeddingGemma 2, a new open multimodal model that sets the standard for on-device efficiency.
- our first open, natively multimodal embedding model
- handles text, code, image, video, and audio tasks within a lightweight, modular 740M parameter form factor
- ideal for offline, privacy-first RAG when paired with Gemma 4
- outperforms some specialist models more than twice its size
Weights available now on Hugging Face.
Khaled retweeted
Introducing Gemini 4 Argon – our new frontier model.
It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
Khaled retweeted
Life satisfaction is highly consistent with personality traits.
The top trait was "Have a lot of fun".
Khaled retweeted
aq.dev. supports any model. switch models mid-conversation without losing context. bring your own subscriptions. it's in the cloud so works anywhere you go. sharing your work with anyone is dead simple. the list goes on.. everything else feels so outdated.
Khaled retweeted
CUA-S1-FORMS is a tiny open-source specialist (706k params) for form decisions only: it scores FILL/CHECK/CLICK/SKIP from structured elements + document values, runs locally in 7-9 ms. Inspired by Jev's System One idea but trained independently on synthetic data.
Jev is TypeSafe's general hosted decision model (Choice/Score/Noul with calibrated probs). On Cua's form task the specialist hit 99.7% vs Jev's 83.6% (expected—Jev wasn't tuned for it). CUA is narrow/local/open; Jev is broader/API-based.
Khaled retweeted
OPPORTUNITY‼️: @Microsoft will give your startup up to $150,000 in Azure credits. No VC backing required. No equity. No accelerator needed.
This is @msft4startups. The Founders Hub branding is gone, but the offer is bigger than ever.
What you get:
- $200 in starter credits instantly
- up to $150,000 in Azure credits
- access to AI tools for development
- tech resources and dedicated support
- GTM help & access to Microsoft's customer network
Eligibility: pre-seed through Series C, B2B software, AI, or tech, working toward or at MVP, privately held.
🔗Apply → microsoft.com/en-us/startups
P.S. We build the best data rooms and are proud to be part of the Microsoft for Startups program ourselves. → @ThePageform
Khaled retweeted
Making Startups Powerful: paulgraham.com/powerful.html
Khaled retweeted
got my iphone duo. can’t wait to start writing top tier content for my wife and my mom
(h/t @dsog)
🤖 Made with AI
Khaled retweeted
Images 2.5 is here.
I don't think it can solve super difficult math problems, but it is really good and we hope you enjoy it.
openai.com/index/introducing…
Khaled retweeted
Astra is very good at 3D modeling, and I can't wait for all of you to experience it, for now here is a little walkthrough on how I built the demo house for our launch blog post. From a Blender scene to a Unreal Engine 5 walkable experience. 🧵
Khaled retweeted
When I ran payments at Airbnb, we could've easily built our own payment processor, but we chose to partner instead. It was much more economical (and reliable) to do so.
Khaled retweeted
this is a critically important moment for cyber defense with AI; there is not much time to act.
we are happy if you want to work with us or any of our competitors or partners, but please take this moment seriously.
only an urgent and intense collective response will work.
Khaled retweeted
Grok 4.6 ranks #1 on CursorBench for real-world coding
Grok 4.6 just ranked #1 on CursorBench 3.2
Outperforming Claude Fable 5, Opus 5 and GPT-5.6 Sol on real-world coding performance
And what makes this even crazier is the efficiency....the chart gives CursorBench performance against average cost per task, and Grok 4.6 is sitting right at the top of the frontier
Grok 4.6 is insanely capable at coding....delivering frontier coding performance at ridiculous efficiency
Khaled retweeted
I'll probably have a more thoughtful answer another time but the quick answer is that things are similar but not the same.
What I mean by that is that we needed IDEs (and editors in general) to manipulate code. Now, we need harnesses to manipulate the models.
If models end up "eating the harness", we'll need multiplayer ADEs to manipulate that new thing, which is btw what AQ.dev is really building.
Things are similar but not the same.
Thansk @JonLowTO for tagging me!
Khaled retweeted
Introducing Muse Glimmer, an open-weight 30B-parameter model optimized for local, always-on agent workflows.
Muse Glimmer delivers strong performance on key agentic use cases and benchmarks compared with leading models in its size category, and is designed to run entirely on consumer hardware like a Mac or PCs with performant GPUs.
In keeping with our long tradition of sharing fundamental AI research, we’re releasing model weights under a permissive Apache 2.0 license.
🧵👇
