Rate Limit Increases Unlock Reliable Claude Usage
Claude Code's 5-hour session limits doubled immediately for Pro, Max, and Team plans, extending usable time from 5 to 10 hours regardless of peak periods. Peak-hour throttling—previously accelerating limits during weekday mornings—eliminated entirely. API rate limits rose significantly across tiers: input tokens per minute jumped (lowest tiers by 16x, tier 1 to 500k or ~370 pages of context); output from 8k to 80k tokens/min, enabling parallel agent runs without frequent throttling. These changes address months of outages and complaints from high demand outstripping compute, making 1M-token context windows production-viable via API.
SpaceX Partnership Fuels Long-Term Compute Scale
Anthropic secured 300 megawatts and 220k Nvidia GPUs from SpaceX, deployed rapidly to handle surging demand from Opus/Mythos releases. This joins deals with Amazon, Google, Broadcom, Microsoft, Nvidia, Fluid Stack, plus a Goldman Sachs/Blackstone JV for enterprise expansion. Closed-source models like Claude outperform open-source (DeepSeek, Gemma) due to massive backend compute—users rely on Anthropic's infrastructure for speed without local VRAM needs. Future: interest in gigawatt-scale orbital AI compute in space to bypass terrestrial limits like power, water, cooling, and community backlash; Anthropic commits to covering consumer electricity hikes near data centers for faster builds.
Five Builder Actions to Leverage Changes
Retest failed workflows: Rate walls that killed Opus agents 6 months ago (e.g., LinkedIn automation with AI infographics) may vanish—new models/limits make them feasible. Shift from Haiku/Sonnet to Opus more freely, easing context management while experimenting boldly. Run production Claude Code behind infrastructure: Double limits allow routine automations/agent loops without exhausting daily knowledge work sessions. Deploy multi-agent setups: Five sub-agents each processing 50k tokens now viable in parallel. Claude Code emerges as flagship product, signaling 5+ year enterprise focus over co-work tools.