Introducing Code Review, a new feature for Claude Code. When a PR opens, Claude dispatches a team of agents to hunt for bugs.

🇧🇷 Brazilian founder @luanalopeslara built the largest prediction market in the world: Kalshi When she was a kid, she would spend 8 hours a day not coding but training to become part of the top 1% of ballerinas in the world. At 18, she quit ballet. She decided it was time to be where the brightest minds in the world were. So she went to MIT and went full grind mode. Machine learning, statistical theory, algorithms, learning AI before it was cool. She then started her career on Wall Street working for legendary names like Bridgewater, Five Rings, and Citadel. Those funds were obsessed with predicting the future, and Luana thought: “why not just trade on binary events directly?” In 2018 she met @mansourtarek_ at MIT and the two came up with the idea of building a market for EVERYTHING. That’s how @Kalshi was born. (Kalshi means “everything” in Arabic) They also kicked off a five-year battle with US regulators, who had no idea what to do with a startup that wanted to let people trade outcomes without turning it into gambling. By 2020 they won. Kalshi became the first regulated prediction market in the United States. A year later, the platform went live. They started with economic events, kept building, kept pushing. Then came the US elections, which ended up being the moment prediction markets finally went mainstream. Number one in finance apps on election night. Nearly two billion dollars in total volume. Now, they've raised 300 million dollars from Sequoia and a16z at a 5 billion valuation. Still, not enough people realize just how exceptional Kalshi’s founders really are. I wanted to share Luana’s story as a reminder that Latino/as founders DO build global tech companies. Long LATAM 🫡

Go listen to the demo for yourself👇

how to get rich

Introducing the compact, dense versions of Qwen3-VL — now available in 4B and 8B pairs, each with both Instruct and Thinking variants. ✅ Lower VRAM usage ✅ Full Qwen3-VL capabilities retained ✅ Strong performance across the board Despite their size, they outperform models like Gemini 2.5 Flash Lite and GPT-5 Nano, and often beat them on benchmarks spanning STEM, VQA, OCR, video understanding, agent tasks, and more. In many cases, they even rival our flagship Qwen2.5-VL-72B from just six months ago! Plus, FP8 versions are also available for efficient deployment. Hugging Face: huggingface.co/collections/Q… ModelScope: modelscope.cn/collections/Qw… Qwen3-VL-8B-Instruct API： modelstudio.console.alibabac… Qwen3-VL-8B-Thinking API： modelstudio.console.alibabac… Cookbooks: github.com/QwenLM/Qwen3-VL/b…