
Claude in JetBrains, MiniMax M3's Sparse Secret | 25 Jun 26
GitHub Copilot now lets JetBrains users switch to Claude as their coding agent with a two-step setup. Claude Code hits GA on Dynamic Workflows and ships a new sandbox credential-blocking setting. MiniMax M3, an open-weight model beating GPT-5.5 on coding benchmarks, gets unpacked — including the Mixture of Experts architecture that makes a 1M-token context window computationally feasible. Plus: OpenAI acquires Gitpod to give Codex persistent cloud memory. Today's teaching concept: Mixture of Experts (MoE) — why activating only 23B of 428B parameters is the key to M3's speed and scale. Hosts…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.