GROK 3: ANOTHER WIN FOR THE BITTER LESSON

Summary from the TLDR Founders newsletter

xAI's new Grok 3 model—trained at a huge scale—has vaulted to state-of-the-art territory. This leap underscores a “bitter lesson” in machine learning: more compute and data often beat smaller, clever hacks. While some point to streamlined approaches like DeepSeek as an alternative, raw scale still wins long term. Once again, bigger clusters and massive training runs seem to outpace all else.