3 ms·Claude Code devs: use previos model, compact at 200k tokens to avoid usage limit5 points by troupo 7mo agojazz9k 7mo agoI see token burn as the limiting factor with AI.