🚨 xAI’s Colossus 2 is getting absurd Elon says the cluster behind Grok already has: – 110K GB200s – 440K GB300s and plans: – +220K next week – +220K in November – +220K more by late December potentially 1.21 MILLION Nvidia AI chips powering xAI’s push to make Grok competitive with GPT-6 and Claude remember when Elon wanted AI development to slow down 😭 bro went from “pause AI” to “send Grok another 220,000 GPUs” what a timeline 💀

Google is literally sending Gemini to space next week 💀 Project Suncatcher launches October 1 on a SpaceX Falcon 9 – refrigerator-sized satellite – 4 Google Trillium TPUs – Gemini models running in orbit – powered by solar energy – testing radiation + cooling in space long term, Google wants entire AI compute clusters in orbit and here’s the problem: the TPUs can currently run for only ~15 minutes before they need to cool down 😭 AI data centers in space… genius or insane?

🚨 Claude helped make Claude 3x faster Anthropic gave an Opus 5.5-level model access to its own performance work for two weeks the result: – Claude web loads ~5x faster – Claude Code starts ~2.5x faster – Cowork starts ~3.5x faster – 3,000+ changes merged and somehow one of the bugs Claude found was an em dash making code highlighting slower 💀 AI optimizing the product running the AI is getting very real how long until Claude starts improving itself without us?

🚨 GEMINI 4 is getting close Google DeepMind just confirmed: – Gemini 4 is in early post-training – internal results already have the team excited – an early version could roll out ASAP – release is expected “much earlier” than year-end Google clearly doesn’t want to spend another cycle watching Opus and Astra own the conversation. the Gemini 4 leaks suddenly make a lot more sense 👀 Can it beat Opus 5.5 and Astra?

aventure.vc/news/2026-09-24-….

🚨 Gemini 4 Pro leaks are getting WILD Alleged checkpoints are showing a MASSIVE jump in 3D, SVGs + frontend coding. Google still hasn’t confirmed it. Best leaked demos below 👇