Local AI gets a hardware push while cloud coding gets squeezed

Apple announced a new 14-inch MacBook Pro and a new iMac, both with the M6. A thread asks whether anyone has replaced Claude or GPT with a local model for daily coding. Answers: Gemma 4 on an M4 is too slow, one person runs DeepSeek V4 Flash on two RTX Pro 6000 Blackwell cards at 160 tok/s, and another says Qwen 3.6 27B dense is roughly Haiku 4.5 level. A new project called Survival LLM bundles an AI assistant and emergency library that runs entirely on your own computer.

On the cloud side, a thread about Anthropic no longer letting Claude Code subscriptions work with OpenClaw complains that "these tools put an outsized strain on our systems" means "when you fully use the capacity you paid for, that's too much." Others say they hit rate limits within 1:45 on afternoons.

The pattern: users are pricing the risk of someone else's quota and rate limits, and local models at roughly one tier below frontier are starting to look like a reasonable hedge.