Every web scraping vendor claims a 98–100% success rate. We tested 15 of them against 90 real sites. Same harness, same URLs, 6,750 live requests The actual range: 33.3% to 95.8%. They can't all be right. So we open-sourced the whole thing - targets, harness, adapters 👇

2026 predictions: 1. agent swarms on hype 2. OSS LLM gpt5 level (deepseek please) 3. new standard coding + comp use benchmark (swe bench like) 4. eval building agent prototypes (we need it)

Pretty surreal to get a shoutout from @auren on his podcast with @yipitdata CEO @vacanti! Public web data is getting exponentially harder to access. At String, we're helping customers stay ahead of that curve so they don’t have to think about antibots, proxies, or really any of the complexity behind the scenes. This is also one of the most insightful alt-data conversations I’ve listened to. Vinicius’ emphasis on actual business value, not just selling data, is likely why Yipit is the only unicorn in the category. And Auren’s questions were incredibly useful from a founder perspective. Excited for the sequel!

Anyone else being incorrectly double charged by @cursor_ai? Second time this is happening, now I'm being blamed for subscribing to the same product twice (I didn't) and had my account suspended because I didn't pay for the same thing twice. emails in the thread

🚀 We just launched Zencoder on @ProductHunt! It's a coding assistant that works for engineers. Our deep understanding of the codebase, agentic pipeline, and 20+ integrations make Zencoder a true dev assistant, not just another AI tool. 👉Show some support and share feedback here: producthunt.com/posts/zencod…

dropped some post about measure theoretic foundations of policy gradients myxik.github.io/posts/measur…

We got 1st place in HP challenge at HackUPC 2023! 🙌 @Devpost devpost.com/software/hp-supp…

Audio is all you need?🔊 A lot of the latest AI models push the boundaries for music/audio generation. Here are 6 new models that were launched this week. 🧵

This audio generation papers this week make me wanna go full in academia after MSc

The Prismatic Desert

This is a really early preview version, but I was having too much fun with it to keep it from y'all. (All of it is subject to change) Disco Diffusion v4 Preview Version [w/ Zoooming and @chigozienri keyframes] colab.research.google.com/dr…

We can already emulate existing audio effects, but how can we create NEW and controllable effects with neural networks? We propose a simple method in our latest work to appear at the NeurIPS @ML4CDworkshop. Let's go over some of the details in this thread 🧵

Pre-NASH workshop was great, RAVE from @antoine_caillon is really impressive, @dadabots gave good insights on using NAS in industry, and also @Ilaria__Manco’s Word2Wace gave promising results. Big thanks to all the speakers!