Boris Cherny, creator of Claude Code: "Opus 5.5 does in a day what used to take your team a month. And still most people keep using it wrong." In 1 hour minutes he explains why latest Claude Code models needs less prompting and why your old detailed prompts work... against you now. Watch it, then save the exact setup below 👇show more

saved this

you are welcome hehe

@zodchiii oh man, been there. tweaking the prompts endlessly when I really needed to understand how Claude's latest update handles context better. it's a game-changer.

The hardest part is deleting the prompt you spent a month building.

Setup threads are great. The people who win still ship the boring product loop the day after the fancy model drops.

We spent years learning how to tell AI exactly what to do. Now we’re learning how to get out of its way

Boris always knows where the friction is. The real bottleneck isn’t coding; it’s context window management and knowing when to let Opus handle a task without constant oversight. Smart devs are learning that restraint beats brute force every time

$4 in $20 out is still steep if it does a month of work in a day. the real comparison is cost per outcome, not per token, and nobody's posted that math yet.

🙏🏼🙏🏼🙏🏼

long prompts were a patch for weaker models. the part worth keeping now is the constraints: what not to touch and how to verify it's done, not step by step instructions

This interview must surely be hosted somewhere else as well..