Magdy Saleh

🇪🇬 🇪🇬

[ali @waterloo_intern]

https://t.co/qe2HpEQZPd

[Baseten @baseten]

GLM 5.2 is live on Baseten.

5.2 is built for agentic engineering: stronger coding, sharper agentic reasoning, and a long context window built to run hours-long tasks.

Test it today: https://t.co/4sOMV0WF3k

[Magdy Saleh @Magdyks]

@finn_hulse hyperloglog is amazing. max number of leading zeros observed in the binary representation of incoming number n -> answer is ~2^n.

[Magdy Saleh @Magdyks]

Opus 4.5 ASIC will be so much fun

[Magdy Saleh @Magdyks]

@dimiwonders This is so sick! Congrats

[Dimi Nikolaou @dimiwonders]

Today AI Video stops being slop.

Introducing Wondercraft Video, an AI video studio built for real work.

Create explainer videos, trainings, product launches, ads, and more by describing what you want.

RT and comment “WONDA” and I’ll DM you 1,000 free credits.

[Adam McKola @AdamMcKola]

They spent all week trying to tell us they'd beat 2008 Manchester United and couldn't even beat 2026 Manchester United. 😭

[Magdy Saleh @Magdyks]

@dimiwonders @vic_weller Go to @Sightglass and arsicault!! @mbadran22 fave pastry shop

[Magdy Saleh @Magdyks]

@WillManidis The coptic monks are fascinating to study

[Magdy Saleh @Magdyks]

@AshesAmbler @IterIntellectus Yes!

[Magdy Saleh @Magdyks]

@dimireadsthings Oooo coaching from the master himself. Im down

[Magdy Saleh @Magdyks]

@gabriel1 Yoga did wonders for my posture

[Dimi Nikolaou @dimiwonders]

Today we launch Branded Pages - the first way to create and share content in a branded webpage.

Create your content on @wondercraft_ai and in 1-click, share it on a custom fully branded web page.

Comment by tagging your company and I’ll share a custom podcast in a branded page

[Magdy Saleh @Magdyks]

@dimireadsthings @AcquiredFM This was such a good episode

[Geoffrey Angus @GeoffreyAngus]

Struggling with context management? Wish you could just stick it all in your model?

We’ve integrated Cartridges, a new method of leveraging sleep-time compute for learning long contexts, into Tokasaurus, an inference engine optimized for high-throughput 🧵

[Predibase by Rubrik @predibase]

🚨 Qwen 3 is here—and it’s a beast.

From a featherweight 0.6B model to a 235B Mixture-of-Experts (#MoE) powerhouse, Alibaba’s latest #opensource LLM is turning heads.

But here’s what most people don’t know: you can deploy and serve any #Qwen3 model securely inside your own private cloud (#VPC) in minutes—no need to share data, no painful infra buildouts.

In our latest blog, we show you exactly how to get Qwen 3 running on private infrastructure with just a few lines of code using Predibase.

Here’s what you'll learn:

🧠 Deep dive on Qwen 3's key features: hybrid thinking, massive language support and benchmarks

🚀 How to serve any Qwen 3 model—from 0.6B to 235B—on your own terms

🔐 What it takes to keep your data private and models secure with VPC-native deployments

📈 Going from prototype to production on Predibase with sub-second latency, private deployments, multi-LoRA serving and robust post-training capabilities

Check it out: https://t.co/xgVuUmVfhN

[Magdy Saleh @Magdyks]

@shaoruu @cursor_ai Tab triggered autocomplete. I want to finish my thought before the AI does its thing. I had it setup with LSP in vim and it made staying in flow so much easier

[Magdy Saleh @Magdyks]

@amasad At the internet cafe

[Saam Motamedi @saammotamedi]

Huge release from @Predibase today -- the first end-to-end platform for Reinforcement Fine-Tuning

Bringing the techniques that power DeepSeekR1 to any open source model and data