Introducing Underdog Images: on-device Image Gen Generate any image, or drop in a photo and say what to change. It all runs on your device. Your photos never leave your computer Made with Qwen-Image Flash from @NVIDIAAI and Qwen-Image-2.1-Turbo from @Alibaba_Qwen

Anyone can get my friend’s real SSN from Claude I'm Terrified of my data in training sets. Emails in databases. iMessages harvested. Cant trust the cloud as AI is crazy good at hacking So I built Underdog for myself: on-device AI OS. Capable & 100% local. now my friends love it

10 months ago, we started Octen with one belief: search infrastructure built for humans won't be enough for AI. Most of the industry saw the same shift, but kept building with a human-search mindset. AI doesn't search like humans. It fires hundreds of queries in parallel, reads everything it retrieves, and needs clean, grounded content back in milliseconds, at a cost that still works at scale. So we didn't adapt human search for AI. We rebuilt it from first principles. Today, Artificial Analysis published its independent Search API benchmark. Octen is the only major provider ranked Top 3 in all three dimensions that matter: #1 in speed #2 in cost efficiency Top 3 in search quality We're also the only provider in Artificial Analysis' "Most Attractive Quadrant" for search quality vs. cost. What makes me proudest is how we got here. Many companies on this leaderboard have been building for 3–5 years and have raised hundreds of millions of dollars. Octen is 10 months old and has raised just $10M. To me, product built per dollar raised is one of the most honest measures of technical strength in this industry. We've kept the team small but exceptional, moved fast, and put nearly everything into the core technology. People used to assume search had to trade off speed, cost and quality. We've shown it can be faster, cheaper and better at the same time. The product engine is working. Next, we're bringing Octen to every model, agent and physical AI product that needs to search. Humans search with Google. AI searches with Octen. We're just getting started. 🚀

❤️ The most beautiful art is created at the research frontier. This new AI foundation model (JETS) detects high blood pressure (87% accuracy), atrial flutter (70%), and more. Save hundreds of millions of lives, please. ❤️

Proof small teams can win in AI: We trained a health foundation model on 3M person-days of data, comparable in scale to the larger labs. It detects high blood pressure with 87% accuracy, alongside atrial flutter (70%), ME/CFS (81%), and sick sinus syndrome (87%). The model was inspired by Yann LeCun's JEPA architecture, but adapted to multivariate irregular timeseries. We called it JETS: Joint Embedding for Time Series. The input is 63 channels of sensor data: heart rate, oxygen saturation, sleep stages, VO2Max, etc. The model was tested on both diagnosis and biomarker prediction tasks. The “AI revolution” won’t be limited to chatbots, images, or code assistants. We think physiological ground truth is the next frontier for health superintelligence.

Direct link: empirical.health/program/gif…

Forget the stuff. The most meaningful gift is helping someone you love live healthier, longer. Empirical Health memberships now available — includes 100+ biomarker tests, a doctor consult, and a tailored plan.

I saw a viral post that said “MCP is just an API” That is wrong and misleading ❌ I recorded a highly-detailed video to explain the difference and why it matters