My Top 3 Takeaways From Code With Claude 2026 #ClaudePartner #claude #ai @claude

But we might be prompting our AI agents wrong rather than telling it what to do. Sometimes we need to be telling it what it looks like if it did the right thing. Here's my top three takeaways from Code with Claude 2026. So outcomes allows you to tell the agent this is what a successful outcome actually looks like and it's going to create itself based on that criteria you give it. One of the hardest parts about development with AI is getting designs to actually look how you want them to look and Doic has a pretty interesting solution to this. So this is claw design and it works sort of how you expect. You prompt it and then it generates a design and you can write comments on here and things and ask it to change the design and then once you are ready to actually build it, you can come over here to share and hand off to Claude Code and this would allow you to send this design over to Claude Code so they can actually build it for you. So essentially, dreaming allows you to tell Claude to go through recent transcripts to sort of organize what it learned into a single new memory store. That way, it's not going to make those same mistakes over again. And you could say, have this run every day or every week. And this is an example of dream memory updates. So essentially, it's going to create all these different files for you to ensure that Claude is making the best decisions for you in the future and learning from everything it's done in the past. And of course, a big thank you to Enthropic for flying me out here and sponsoring this