Back

🧠 Why training AI to be too perfect may be holding it back… In my latest post, I explore why designing AI agents with human-like limitations is not a weakness - it’s a superpower. FPS games might be the key to developing AI that thrives under uncer

Can Al learn instinct - not just data? In my latest post, I explore how elite gamers might hold the key to training smarter, situationally aware Al. This isn't science fiction. It's the future of tactical decision-making - where machines learn not

🎯 Instinct vs Intelligence When milliseconds decide survival, intelligence alone isn’t enough - instinct takes the lead. Elite gamers don’t just react fast - they predict, adapt, and thrive under pressure. Now imagine AI that can do the same in t

Turns out Call of Duty might be more than just a game... We’re using FPS scenarios to train AI agents that can think fast, move smart, and survive unpredictable situations — just like humans do in real life. It’s not about reflexes. It’s about inst

When Nvidia validates your thesis… Jensen Huang just showed off a humanoid robot trained to walk in a metaverse-style simulation - not pre-programmed, but learning from 20M+ hours of human traces. That’s exactly the principle we’ve been building to

🎮 What if the best AI training ground isn’t a lab, but a game? From AlphaStar’s StarCraft victories to OpenAI’s Dota 2 team, real experiments show how games teach AI instincts: adaptability, teamwork, and survival under chaos. The lesson is clear -

🚨 SmolVLA is here — and it’s changing how we think about robotics AI. Hugging Face just released SmolVLA, a lightweight Vision-Language-Action model trained on community-shared datasets from their LeRobot platform. Despite being just 450M paramete

🚀 Just published a new piece on Medium: “Why First-Person Shooter Games Could Be the Future of AI Training” AI agents are getting better - but when it comes to tactical decision-making under pressure, traditional training methods fall short. What

FPS Games + UAV/UGVs = Next-Gen Soldier Training? From America's Army to VR drone duels, gaming sharpens tactics, reactions & teamwork - now with AI-powered sims for hybrid warfare. Check our updated Medium post on how Redoubt.in is pioneering this

sarvam just announced 30B and 105B models! 30B: + pre-trained on 16T tokens + context length of 32k tokens. 150B: + context length of 128k tokens + on-par with most other open closed source ones of its class Sarvam Chat app will be available fro