Kushal Chauhan

@kushalchauhan.bsky.social

Research Engineer at Google DeepMind

🥁Introducing Gemini 2.5, our most intelligent model with impressive capabilities in advanced reasoning and coding. Now integrating thinking capabilities, 2.5 Pro Experimental is our most performant Gemini model yet. It’s #1 on the LM Arena leaderboard. 🥇

BildBild

The worst part of working with Gemini is the incredibly tedious process of self-human-evaluating its performance for every early stage idea I have. I miss the simpler times where automatic evals were reliable.