Thank you both Daria and Karen! I love how you two collaborate and wrote this piece so cleanly!
Like you Karen, I'm not a design person, so having a dedicated agent to control styles save a lot of headache. Thanks for the generous sharing of the agent prompt, definitely stealing it :)
The subagent architecture you're describing is exactly how I structured my own automation system. The real breakthrough isn't parallelism for speed—it's specialization for accuracy. When I need to search 5 job boards simultaneously, spawning parallel agents means each one has focused context instead of trying to juggle everything.
What's less obvious: subagents aren't just about delegation. They're about isolation. If one agent hits an error or goes off-track, it doesn't contaminate the parent context. That architectural separation is what makes complex multi-step tasks actually reliable.
Agree subagents mimic human “teams” by working in parallel…and the more the merrier, generally speaking! So glad you found a framework that works for you.
Super helpful and informative. What I enjoy most is being able to ‘park’ ideas until pros like you, Karen, and Daria write about them, and then the divine timing arrives to send them over to Claude Code. It’s really helping me grasp how powerful Claude Code is and making me feel great about the max subscription.
Awesome post Karen and Daria! I once tried spawning multiple Claude agents in CC once through a prompt, it opened multiple terminal instances and was a bit chaotic (also maxed out my pro subscriptions in 5 minutes). I think I need to try this approach with the /agents command.
Also congratulations on the YouTube channel Karen, just subscribed, videos were so helpful.
I think you’ll be pleasantly surprised by /agents! They work in a single terminal, and each one gets its own context window, so you have more time before the dreaded “compact.”
Thank you Daria and Karen for providing a great template to learn more about Claude Code and, especially, its subagents. I learnt quite a few things from one of Karen's previous articles on Claude Code subagents (if I remember correctly). Also, thanks for the shoutout to our Claude Code article 🤗
Thank you! Karen really is the person to follow on Claude Code building, couldn't agree more.
PS: the shoutout was well deserved! your article with Ashwin is a great resource for anyone getting started with Claude Code who wants to see the possibilities.
Thanks Daria and Karen, I am saving this post for later because, having read through it once, it's already given me exactly what I need as I continue my Claude Code journey. Also, a continued shout out to Karen for being such a helpful person in terms of helping me get back on the coding bandwagon.
Thanks Daria. Me and Karen are working on something at the moment, but in the interim I also built myself an email scraper that goes through my entire Gmail and looks for items that are tax deductible that I might have missed. Not something I'd necessarily ship, but something that's been incredibly useful for me.
I love that Claude Code makes it so much easier for domain experts (like you!) to build really cool things. ✨
Sometimes, it can feel like magic elves are building your software…a feeling that dissipates when it builds the wrong feature or breaks something that’s already working. 😆 But patience + a willingness to test and correct Claude goes a really long way!
Thank you both Daria and Karen! I love how you two collaborate and wrote this piece so cleanly!
Like you Karen, I'm not a design person, so having a dedicated agent to control styles save a lot of headache. Thanks for the generous sharing of the agent prompt, definitely stealing it :)
Using subagents with Tailwind has drastically improved my project UIs!
You know something’s good when it gets stolen haha. But honestly - I had zero doubts with anything Karen creates… or anything you build either, Jenny.
Thank you 🤗
The subagent architecture you're describing is exactly how I structured my own automation system. The real breakthrough isn't parallelism for speed—it's specialization for accuracy. When I need to search 5 job boards simultaneously, spawning parallel agents means each one has focused context instead of trying to juggle everything.
What's less obvious: subagents aren't just about delegation. They're about isolation. If one agent hits an error or goes off-track, it doesn't contaminate the parent context. That architectural separation is what makes complex multi-step tasks actually reliable.
The challenge is coordination overhead—when do you parallelize vs. sequence? I documented 10 real-world patterns where this architecture proved essential: https://thoughts.jock.pl/p/ai-agent-use-cases-moltbot-wiz-2026
Agree subagents mimic human “teams” by working in parallel…and the more the merrier, generally speaking! So glad you found a framework that works for you.
I access Claude 4.5 through Perplexity as my default LLM. Is this different from accessing Clause directly?
Yes, accessing Claude via Perplexity is a little different.
Using Claude in Perplexity = you’re chatting with Claude through Perplexity’s interface and tools (search, citations, their UI, their limits).
Using Claude Code = you’re using Claude through Anthropic’s own app, with full access to coding/agent workflows.
Hope this helps!
Amazing stuff here. Amazing explanation of one of the most important tools to know how to use in 2026 and beyond! Everyone should be saving this one.
Thanks Ilia - indeed, sooner or later more and more of us will jump on the CC wagon
This is such a clean explanation of what people keep missing about Claude Code.
Subagents aren’t about “more AI.”
They’re about separating concerns so thinking doesn’t collapse under one giant context window.
Awesome walkthrough 🔥
Thanks so much! 🤗
Super helpful and informative. What I enjoy most is being able to ‘park’ ideas until pros like you, Karen, and Daria write about them, and then the divine timing arrives to send them over to Claude Code. It’s really helping me grasp how powerful Claude Code is and making me feel great about the max subscription.
This really made me smile💜 So glad the timing lined up.
Karen is the pro here!
So happy to hear that! ✨🤗
Awesome post Karen and Daria! I once tried spawning multiple Claude agents in CC once through a prompt, it opened multiple terminal instances and was a bit chaotic (also maxed out my pro subscriptions in 5 minutes). I think I need to try this approach with the /agents command.
Also congratulations on the YouTube channel Karen, just subscribed, videos were so helpful.
I think you’ll be pleasantly surprised by /agents! They work in a single terminal, and each one gets its own context window, so you have more time before the dreaded “compact.”
So glad you enjoyed the videos! 🤗
A great guide to getting started with Claude code. I was so scared at first to use it. Once i have been in, i ve just loved how cool claude code is!
💯 It’s really easy to use!
"building in parallel" 🎉
The only way to build! 😁
Not “AI helps me do tasks,” but “AI helps me think and build in parallel.”
💯
Thank you Daria and Karen for providing a great template to learn more about Claude Code and, especially, its subagents. I learnt quite a few things from one of Karen's previous articles on Claude Code subagents (if I remember correctly). Also, thanks for the shoutout to our Claude Code article 🤗
Thanks so much! Yes, I tackled subagents before, they’re one of my favorite features!
Thank you! Karen really is the person to follow on Claude Code building, couldn't agree more.
PS: the shoutout was well deserved! your article with Ashwin is a great resource for anyone getting started with Claude Code who wants to see the possibilities.
Thanks Daria and Karen, I am saving this post for later because, having read through it once, it's already given me exactly what I need as I continue my Claude Code journey. Also, a continued shout out to Karen for being such a helpful person in terms of helping me get back on the coding bandwagon.
This is exactly why I wanted to bring Karen in. She's the best! love everything she's building and glad it made it into your bookmarks.
(Also super excited to see what you build next!)
Thanks Daria. Me and Karen are working on something at the moment, but in the interim I also built myself an email scraper that goes through my entire Gmail and looks for items that are tax deductible that I might have missed. Not something I'd necessarily ship, but something that's been incredibly useful for me.
Building things to solve your own problems is a big win!
I love that Claude Code makes it so much easier for domain experts (like you!) to build really cool things. ✨
Sometimes, it can feel like magic elves are building your software…a feeling that dissipates when it builds the wrong feature or breaks something that’s already working. 😆 But patience + a willingness to test and correct Claude goes a really long way!