Claude cuts latency 35% with threads

🔥 What happened Anthropic launched Message Threads in public beta on the Claude Platform. Agents now send only new messages instead of resending the full history each turn. For 500k to 1M token requests, median time-to-first-token drops by roughly 35%. 💡 Why it matters Long coding sessions with hundreds of requests were throttled by constant re-uploading. Threads sit encrypted in memory and expire after 5 minutes idle. The catch: token costs stay identical, since Claude still reads the full conversation every turn. ⚡ Our take A clean latency win with no price cut — Anthropic is optimizing UX, not your bill. Anyone hoping for cheaper agents will be disappointed.