We’re introducing the Cursor SDK so you can build agents with the same runtime, harness, and models that power Cursor.
Run agents from CI/CD pipelines, create automations for end-to-end workflows, or embed agents directly inside your products.
Transformers.js v3 is finally out
- WebGPU support (up to 100x faster than WASM)
- New quantization formats (dtypes)
- 120 supported architectures in total
- 25 new example projects and templates
- Over 1200 pre-converted models
- Node.js (ESM + CJS), Deno, and Bun compatibility
- A new home on GitHub and NPM
A warm welcome to Moonshine, a new family of speech-to-text models!
Moonshine claims to be as fast and accurate as whisper-base, while being up to 5x faster! 🤯
They achieve this by removing whisper's constraint on 30-second length audios. Instead, Moonshine processes audios of any length. 🧠
I will be trying this with our speech-to-speech pipeline and reporting back the results. So far, I'm sad not to see an MLX native implementation.
What if AI dev products (Claude, v0, etc) let you install packages, run backends & edit code?
Introducing https://t.co/XXroNZNn7H, by StackBlitz:
- Prompt, edit, run & deploy fullstack apps
- Full dev env (npm, Vite, Next.js, …) w/ frontier AI
- Free!
Announcing Rspack 1.0 - the next generation JavaScript bundler written in Rust, webpack compatible, 10x faster.
Check out the release blog for more details:
https://t.co/nOVJjKMRBo
For Day 4 of Launch Week, we are dropping our biggest release yet…
Introducing Firecrawl /v1 🔥
A more reliable and developer friendly API that makes gathering any web data seamless.
Learn about all of the improvements we are shipping with /v1 (1/9) 👇
Online demos for BiRefNet on @huggingface Spaces!
Is this the best background removal model out there? 🤯
MIT licensed. 5.5G GPU memory needed for inference for 1024x1024 images.🤩
Anthropic (@AnthropicAI) just announced a game-changer for their API: Prompt caching.
Think of prompt caching like this: You're at a coffee shop. The first time you visit, you need to tell the barista your whole order. But next time? Just say "the usual."
That's prompt caching. Here's why it's a big deal...
It makes it cheaper and faster to call any large prompt components that you reference over and over again.
Specifically? Up to 90% cost reduction. 85% faster responses. Use cases are wild.
Think:
•Cheap conversations with chatbots that have encyclopedic knowledge
•Coding assistants that can efficiently read your entire codebase
•AI that can discuss entire books at crazy low latencies and costs
It means a bunch of AI-powered apps that were previously too expensive to run just became viable.
You can now fine-tune your own FLUX.1 text-to-image AI on Replicate, with just a few images.
Teach the model to generate a character, pet, style, or anything else.
@fofrAI even trained one on "bad 70s food". See it in action, and learn to do it too 👇
https://t.co/aLFkNBkaJE
SREF CODES COLLECTION. UPD
Midjourney v6.2 will be a bigger release than v6.1. It will likely drop soon — no official date has been announced, but I’m guessing September or even earlier, as Midjourney often surprises us 🙂
I've decided not to update my collection for v6.1 as planned since it doesn't make sense. Instead, I’ll send you the update for v6.2.
In the meantime, if you haven't purchased it yet, you can download the sref codes collection with prompt examples at a 30% discount.
Use the code: 30
The link is in the thread.
→ 70 Styles
→ 280 Prompt Examples
→ 1120 Images (you can download all images from the PDF). Most of the images have resolution 896 × 1344 and some of them 928 × 1232.
→ Constantly Updating 🟢 (next update in September)
Just change --v 6.0 to --v 6.1 for even better results. You can also add my personal code --personalize kzilt9y to get cleaner images.
Enjoy!
Midjourney just released their web editor!!
It's actually one of the coolest features they've dropped in a while
essentially lets you do inpainting, panning, zooming and more all in a single step
really really powerful
🌎 Hola, Bonjour, こんにちは! Introducing AI-powered Localization for your website. Effortlessly translate your entire site, customize images, and even tweak entire designs for different languages and regions. Learn more about Localization at https://t.co/kmlqMtTevi
CSS Trick 🖼️
You could take the resize trick from last week and use it to create one of those sliding image components 👀
CSS only, JavaScript to enhance later 🤓
Put an image in each panel and make them position: absolute. They can use a container query unit for their size 🫶
img {
position: absolute;
width: 100cqi; // Make them the container width 🤙
}
.panel:last-of-type img {
right: 0;
}
The rest of the trick from last week was using display:grid with grid-template-columns. Then we have a resizable element inside the first column 😎
.panels{
display: grid;
grid-template-columns: auto 1fr; 👈
}
.resizer{
max-width: 100cqi;
overflow: hidden;
resize: horizontal; 👈
}
Where does JavaScript come in? To progressively enhance the thing with keyboard controls ✨ Not 100% sure of the pattern for that though. Feels like it should be a button that you can toggle into resize mode and then use left/right to resize 🤔
Anyways, @CodePen link below! 👇
Stability AI is proud to Introduce Stable LM 3B: Bringing Sustainable, High-Performance Language Models to Smart Devices.
Download the weights on Hugging Face today 🌟🚀 #StabilityAI#StableLM3B#LanguageModels https://t.co/AIFynjqQ4a
The combination of Bing Chat vision with DALL•E 3 is amazing.
Bing not only understood my image but also brought my logo sketch to life using DALL•E 3.
Here is how you can do it too in a couple of minutes:
You can now get a full tracing/observability UI in *all* @llama_index RAG/agent pipelines, in one-line of code ⚡️
Bonus: all your data lives locally! 🔐
We're launching a native integration with @arizeai Phoenix 🔥. Full 🧵 below.
Full Colab nb: https://t.co/JexIInOaUs
Explaining 8 Popular Network Protocols in 1 Diagram. The method to download the high-resolution PDF is available at the end.
Network protocols are standard methods of transferring data between two computers in a network.
1. HTTP (HyperText Transfer Protocol)
HTTP is a protocol for fetching resources such as HTML documents. It is the foundation of any data exchange on the Web and it is a client-server protocol.
2. HTTP/3
HTTP/3 is the next major revision of the HTTP. It runs on QUIC, a new transport protocol designed for mobile-heavy internet usage. It relies on UDP instead of TCP, which enables faster web page responsiveness. VR applications demand more bandwidth to render intricate details of a virtual scene and will likely benefit from migrating to HTTP/3 powered by QUIC.
3. HTTPS (HyperText Transfer Protocol Secure)
HTTPS extends HTTP and uses encryption for secure communications.
4. WebSocket
WebSocket is a protocol that provides full-duplex communications over TCP. Clients establish WebSockets to receive real-time updates from the back-end services. Unlike REST, which always “pulls” data, WebSocket enables data to be “pushed”. Applications, like online gaming, stock trading, and messaging apps leverage WebSocket for real-time communication.
5. TCP (Transmission Control Protocol)
TCP is is designed to send packets across the internet and ensure the successful delivery of data and messages over networks. Many application-layer protocols build on top of TCP.
6. UDP (User Datagram Protocol)
UDP sends packets directly to a target computer, without establishing a connection first. UDP is commonly used in time-sensitive communications where occasionally dropping packets is better than waiting. Voice and video traffic are often sent using this protocol.
7. SMTP (Simple Mail Transfer Protocol)
SMTP is a standard protocol to transfer electronic mail from one user to another.
8. FTP (File Transfer Protocol)
FTP is used to transfer computer files between client and server. It has separate connections for the control channel and data channel.
–
Subscribe to our newsletter to download the high-resolution PDF. After signing up, find the download link on the success page: https://t.co/keu1P0hhKI
Multi-document agents 🤖
Today we released an initial architecture aimed at generalizing RAG to more documents and more question types 📑🔎
Check out this amazing diagram by @clusteredbytes outlining the core agent/retrieval components!
Full guide here: https://t.co/LosFQEjpy8