more sync engine shenanigans. using CF's new oauth stuff... to build a reactive kv editor with the cloudflare api. here it's syncing across chrome and safari, even though the kv api isn't reactive at all. instant reloads and everything.show more

Today, we're announcing four new capabilities on the Anthropic... API to help developers build more powerful AI agents. A code execution tool, MCP connector, Files API, and extended prompt caching:show more

🤔 Imagine if your backend worked like `useState`. No... API calls. No fetching. Just reactive, real-time state that syncs across devices. That’s what InstantDB is doing. With Instant your frontend gets a real-time database. You write queries that stay in sync and work offline. Every save updates immediately, and it comes with built-in auth and permissions. See how InstantDB looks in action in this deeply technical blog post from Stopa ↓show more

Day 11/90 of Inference Engineering How does vLLM work... and how is it used in production? Before we discuss how vLLM works internally, it helps to understand what vLLM is. At a high level, vLLM is an inference engine that is designed to serve LLMs to thousands of concurrent users efficiently while managing scarce compute and memory. The goal for vLLM is to maximize throughput and minimize latency; optimizing for the best inference economics and experience for end users. With every request from the end user, it eventually ends up in the engine core, gets scheduled alongside other requests from other concurrent users, executes on the GPU, and updates the KV cache with the new key and value vectors, and streams the tokens back to the user. The Scheduler decides what requests should execute next while continuously batching requests together to maximize GPU utilization. Continuous batching is an inference optimization that allows new requests to join a running batch as other requests finish generating tokens. This helps with keeping the GPU utilization high instead of letting it sit idle waiting for an entire batch to complete generating. After the scheduler dispatches the selected batch to the Model Executor, the Model Executor prepares the tensors and metadata required for inference, retrieves each request’s block table from KV Cache Manager, launches the optimized transformer forward pass on the GPU, computes the logits, updates the KV cache with the new key and value vectors, and finally returns the results for sampling and streaming. The KV Cache Manager uses the PagedAttention memory layout to allocate fixed-size cache blocks on demand and maintains a Free Block Queue on the CPU that tracks which blocks in the GPU’s Paged KV Cache are currently free. When a request needs additional KV cache space, the KV Cache manager takes a free block from the queue and assigns it to that request, thus avoiding an expensive search through GPU memory for available cache blocks. All of these components form the core of vLLM’s inference engine. The Scheduler determines what requests are executed, the Model Executor determines how those requests are executed, the KV Cache Manager determines where each request’s KV cache lives using the PagedAttention Memory Layout. This architecture enables vLLM to serve thousands of concurrent requests with high throughput, low latency, and efficient GPU memory utilization. Heres a little animation that visualizes everything! - I've also completed the forward pass for my mnist.c project. I had a nice chat with shrey birmiwal, such a knowledgeable guy. Excited to learn more about vLLM and implement a tiny-vLLM one day.show more

I made a thing! 🐱 Clipcat (link in the... next tweet) is an API for rendering videos in the cloud with smooth animations, multiple scenes, at up to 60fps. It's like Bannerbear, but with complex animations / tweening, and a keyframe-based editor.show more

AN ENTIRE PLAYABLE CYBERPUNK CITY FROM ONE SINGLE PROMPT... → give Kimi K3 the idea and it builds you the whole city → move through the streets in 3D → neon lights, traffic, reactive buildings → all running in the browser $3 in API with K3. the same would run about $9 with Fable 5 this needed a full game studio a month agoshow more

Nice! We no longer need a depth sensor, a... Kinect, or even Windows to build amazing interactive installations. Just an Apple device, ARKit, and AI. Rendered with Three.js in Electron, using its new SharedTexture API for direct GPU-to-GPU texture sharing. 🚀show more

Have you been missing the Fudge Pack that contained... pretty much all the main POSTAL stuff? Well, we a new one coming soon and it's packed with even more games to complete your collection!show more

BREAKING: SpaceXAI just added a powerful new 'Browser Use'... plugin to Grok Build. It gives Grok access to a real browser, either the user’s own Chrome with existing logins or an isolated Browser Use cloud browser. Grok Build can now browse websites, scrape and extract data, fill out forms, test web apps, take screenshots and automate complete web workflows. It can also run locally through uvx, with no API key required when using local Chrome. Install command: grok plugin install browser-use --trustshow more

HeroUI Native alpha 15 is live 🚀 What’s new:... 🧩 Select component — simple, flexible, and powerful 🎛️ Refactored RadioGroup, Button, and Chip for a cleaner and more consistent API ⚒️ Small fixes and improvements across the library We even re-created the Raycast AI model selector to show what the new Select can do ⚡️show more

This dev built a prototype soccer game using the... new Meta Quest camera API to track a real ball with simple color detection and it actually works surprisingly well! Source:show more

Added a Tool Directory to Centaur, showing all the... tool integrations that it ships with out of the box, across many categories: news, infra, research, crypto, company productivity and much more. To use them, you just need to configure their API keys.show more

1-bit Kimi K3 performs at Opus 5 level on... 3D physics! We ran our Atomic Chat quant of Kimi K3 locally on 4x B200 against three cloud models and gave them all the same task, to build a giant anvil drop test as a single HTML file with real physics Outputs: K3 1bit (local): 15.8K tokens, $0 API cost Kimi K3 (API): 15.3K tokens, $0.30 API cost Opus 5: 22.8K tokens, $0.77 API cost GPT 5.6: 14.5K tokens, $0.72 API cost All four got the physics right. But only Kimi made a working winch. The drum turns and the chain drags the flat car off the pad. Opus 5 drew the most detail, road markings and sparks on the hit. And you can run a model at this level on your own box now. That still feels insane to usshow more

Make instant apis. Tell your agent in English, it... figures out the code and puts it on a dynamic worker. Get a shareable link to your new live API endpoint. The possibilities with this infra make me giddy!show more

Big news for all DeFi traders, devs, and AGENTS!... We just launched the Compass API. No solidity stress, no authorization, just simple high-level calls execute. Trade, lend, or rebalance instantly in DeFi. Launching with Uniswap & Aave on Ethereum + Arbitrum. Agents can finally ACT, not just talk. API reference: API App: Check out more & build with us: Stay tuned because we’re going fast. With many new integrations launching over the next weeks! #DeFi #Trade #DeFAIshow more

Scroll-triggered animations are coming to Chrome early next year... (and we've been really excited about this one for a bit!) 🥳 Unlike scroll-driven animations (already in Chrome & Safari), this API introduces "timeline-trigger" and "animation-trigger" to start an animation within a specific range, rather than relying on the scroller completion *as* the animation timeline. I got nerd-sniped and wanted to try it out so I built this shape morphing demo to show how you can apply multiple animations to the same element over with different triggers. PRETTY COOL IMO! Bramus has a post with more info ⬇️show more

The SF5 (3D printable MP5-based build) is compatible with... all combinations of MP5 parts - full size, K-size (as seen here), RS, even 40 and 10mm. It's still in beta, though testing has been going great - waiting on just a little more testing before release.show more

The new Hellfire Warlock Hardcore build is popping off,... actually quite fun And I managed to make a build with even more ocular cancer than before, who knows what's under you, it's a surprise every time!show more

Seedance 2.5 is here The model, plus everything it... was missing only at Magnific → NEW 3D motion control tool by Magnific → Up to 50 references, and 30 second generations → Edit, extend and upscale to 4K in one workflow Available on Magnific, MCP, Spaces, Plugins and APIshow more

You can now use GPT 5.5, Gemini 3.7 Flash,... Kimi K3 and 47 other AI models completely free😱 No subscription. No credit card. Even the API usage costs $0. AIHubMix just opened a free catalog with 50 AI models. Some of the available models: • Ox Alpha • Gemini 3.7 Flash • GLM 5.2 • Kimi K3 • MiniMax M3 • GPT 5.5 • 40+ more And you don’t need separate API keys for each model. Setup takes 2 minutes: > Step 1: Go to > Create an account using your email or OAuth. No card needed. Step 2: Create one API key > The same key works with every free and paid model. Step 3: Add it to any OpenAI-compatible tool Base URL: Then choose any model ending in -free, such as: coding-glm-5.2-free gpt-5.5-free That’s it. One API key. 50 AI models. $0 for both input and output. Save this. You might need a free multi-model setup later.show more