Hey Reader,

Quick update from my 30-day Cursor-to-Claude Code challenge.

I found something that might change everything.

Is This The New Cursor?

Remember when I said I'd obviously go back to Cursor after this experiment?

Yeah... I'm not so sure anymore.

On day 3, I re-tested a tool called Conductor.

I don't want to overhype this, but... this thing is the real deal.

It wraps Claude Code AND Codex into one clean interface.

And it solves literally EVERY frustration I've had with using Claude Code on its own.

Think of it like this:

Claude Code alone is like texting the smartest person you know... but they can only reply one message at a time and you can't see what they're typing.

Conductor lets you watch them work, open multiple chats, and actually see the files they're editing.

Here's what it gives you:

→ Great integration with Claude Code's plan mode
→ Multiple tabs with multiple agents at the same time
→ Can use multiple models, not just one
→ Best git worktree support I've ever seen
→ Dedicated review buttons paired to any model
→ Buttons to create PRs, launch dev servers, etc

...and the wildest part... it's free. 🤯

IT'S FREE!!!

I'd literally pay for this UX. But I don't have to (because it uses your existing Claude / ChatGPT subscriptions).

I'll continue testing it deeply, but love the team (they're very responsive) and the workflow so far.

Sub-Agents Are Insane (Kinda)

On day 4, I discovered something Cursor literally cannot do yet.

Sub-agents.

Imagine launching 5-10 specialized AI agents to review your code.

They work simultaneously, but focus on specific parts of your code.

And you don't have to decide which parts, Claude will do that for you.

Not generic "check my code" requests.

Agents that specifically hunt for:

→ Unused code and libraries
→ Duplicated logic across your app
→ Broken best practices

I built my own agent and ran 5 of them for about 5 minutes.

They found issues that a single Cursor chat completely missed.

Here's why this matters for non-technical builders like you:

I haven't written actual code in 10 months. I ship everything through AI.

That means a solid code review is EVERYTHING.

Nothing has been as thorough as running multiple specialized agents at once.

Cursor is working on sub-agents, but they're not here yet.

As of right now, Claude Code is ahead here.

The Harsh Truth About Skills

Now here's where I need to be brutally honest.

I spent three days diving deep into this whole ecosystem of skills, agents, marketplaces, plugins, rules, and commands.

And the vast majority of it is genuinely useless.

File organizers. Video downloaders. GIF creators. PDF readers.

Literally nobody needs them, and if you do once in a blue moon, you just ask Claude for it.

Just a few words can replace the entire skill. They're literally useless.

It's just like MCP servers.

Tens of thousands of them exist.

But the ones that actually matter? I can count on one hand.

Some of these skills can even be DANGEROUS.

Just like insecure MCP servers, some can actually put your app - or even your entire system - at risk.

Sorry to be the voice of reason here.

Plus, they absolutely kill your context window if you're not careful.

What's Actually Worth Using

After all that testing, here's what I found useful:

1. dedup-code-agent

This is my own agent that has been surprisingly helpful.

It cleaned up hundreds of lines in my Prompt Coach app (dead/duplicate good).

2. skill-creator metaskill

There's plenty of them, just google this.

They help you build skills for YOUR specific needs when you actually need one.

Think: build a specialized shadcn designer, or a stripe implementation setup skill.

Or maybe your app has an API, or your company does - you can teach Claude that with skills.

3. frontend-designer

Probably the most famous Claude skill, for a reason.

It's just a simple prompt, but it works, and leads to great designs.

For everything else, I'd just ask Claude directly, no skills needed.

I even built my own marketplace this week. Two plugins live right now that auto-update as I improve them.

But honestly, this was more of a learning exercise than anything else.

Don't expect game-changing agents that will 10x your workflow.

Not yet anyway.

Will I Return to Cursor?

I'm still on my 30-day challenge.

Just about a week in now.

Last email I told you I'd obviously go back to Cursor when this was over.

But now, with Conductor, I'm honestly very happy and the token anxiety is GONE.

This all had me so excited that I even recorded a YouTube video for you.

It will be on the channel next week.

It's the closest thing to a "perfect workflow" I've seen in a long time.

It's minimal. It's the polar opposite of Cursor. But it has all the essentials.

I could genuinely see myself using this full-time.

What are you using right now? Hit reply - I read everything.

Talk soon,
Rob

P.S. The 20% OFF Christmas Sale is still on. Grab your Blueprint now and start 2026 with my support.

​