One tweet almost cost a founder a month of progress.
Pat Walls, the guy behind Starter Story, posted that Cursor was amazing but Claude Code changed how he works. Ross, who builds apps on the Starter Story Build channel, saw it and started spiraling. Should he rebuild his whole app? Learn a new tool? Was he already behind?
So they recorded a podcast episode about it. And the Claude Code vs Cursor debate they had is worth 32 minutes of your time, because the answer they landed on has almost nothing to do with code editors.
It has everything to do with why your SaaS isn't growing.
Here's the full episode:
The FOMO that started it
Ross is a beginner. He spent 40 to 60 hours learning Cursor from absolute scratch. Never used a code editor before. Didn't even know what a repository folder was.
Those hours paid off. He shipped a mobile app with Cursor and nothing else.
Then the timeline turned on him.
Pat's tweet. A Reddit post titled something like "Claude Code changed my life." Alex Finn posting about it daily. Everyone in his feed suddenly acting like Cursor was a relic.
Ross even posted about it himself: he feared he was about to waste dozens of hours learning a new tool. The comments split down the middle. Half said do it, half said don't.
If you've ever watched a new AI tool blow up while you're mid-build on the old one, you know exactly how this feels.
(I see this every week with founders. The stack changes faster than the MRR does.)
What Claude Code actually does differently
Pat walked Ross through his real setup, and the differences are concrete.
Cursor is an IDE, a fork of VS Code. Chat window on the right, code in the middle, folders on the left. Visual. Clickable. Beginner-friendly. That's exactly why Ross liked it.
Claude Code is none of that. It lives in the terminal. You type one prompt, and it goes off and does a bunch of things without showing you a wall of diffs.
The specifics from Pat's workflow:
- It plans on its own. Pat told it to run a script. It wrote its own to-do list first, then worked through it. He never gave it a plan.
- No context babysitting. In Cursor, Ross learned the hard way that you have to @-tag the right files or things get screwed up. Claude Code just finds what you mean, even in the big messy Starter Story repository. Pat one-shots it.
- The permissions flag. Pat runs it with a flag called dangerously-skip-permissions, which stops it from asking "can I run this?" every 30 seconds. His words: it's kind of unworkable without it. It's also, as the name says, dangerous. Beginners should keep the training wheels on.
- The screenshot trick. Command + Control + Shift + 4 copies a screenshot to your clipboard. Paste it into the terminal with Control + V, type "the submit button doesn't work, fix it," and walk away.
And here's the part that surprised me: Pat still opens Cursor every day.
Not for the AI. He runs Claude Code inside Cursor's terminal, purely out of laziness, because his files are already there. He prompts Cursor itself maybe once or twice a day now, down from using it for everything.
The walk-away workflow
The real shift isn't the tool. It's how Pat works now.
Old way: sit down, zone in, feed the AI one step at a time. Ross still works like this. He has to block out an hour, give Cursor step one, wait, give it step two, wait. Alex Finn taught him that trick because long prompts kept blowing up.
Pat's way: take a 10-step feature, hand Claude Code the first 8 steps in one prompt, and physically leave his desk.
He puts chicken in the air fryer. He walks his dog Bernie. A task takes 5 or 10 minutes, and the AI figures itself out while he's gone. Then he comes back, does the last 2 steps himself, and tests everything before pushing.
His framing: it's a junior programmer. You delegate, you review, you ship.
That's not a coding tip. That's a management skill, and it's the same muscle behind every one of the business automations that let a solo founder run something that looks like a 10-person company. Delegate the repeatable 80 percent, keep the judgment calls.
The founders I interview who get past $50K MRR all make this jump eventually. The tool just forced Pat to make it sooner.
The Lovable reality check
Mid-episode, Ross brought up Lovable. Reportedly at $100 million ARR, called one of the fastest companies ever from $1M to $100M. And yet Noah Kagan and half of tech Twitter keep saying they've never met a single person who uses it.
Pat's response was the best moment of the episode.
People need to get out of their little bubble. Twitter is an algorithm that serves you, and you think it's the whole world.
His example: HubSpot. Pat uses Notion. Everyone he knows uses Notion. He knows almost nobody on HubSpot. HubSpot makes billions per year.
Your feed says everyone builds with Cursor or Claude Code and nobody uses Lovable. The revenue numbers say your feed is a tiny, weird corner of the actual market.
Two lessons hiding in there:
→ The space is enormous. Multiple billion-dollar companies exist in AI coding alone, serving audiences that never overlap.
→ If your sense of "what everyone uses" comes from your timeline, your sense of what customers want is probably just as skewed. Get into your customers' world, not your bubble.
The answer nobody wants to hear
So. Should you switch from Cursor to Claude Code?
Pat's honest answer, as the guy who switched and loves it: do what works for you.
Because he asked Ross a better question. Which feels scarier: rebuilding your app in Claude Code, or figuring out what to do next with your app?
Ross didn't hesitate. What to do next. Where to post. Who to talk to. How to get users. Rebuilding felt intimidating but knowable. Marketing felt like fog.
And that's the tell.
Tool FOMO is procrastination wearing a productivity costume. You could spend a week learning Claude Code, or a week figuring out how to get 100 customers. One of those moves your business. You already know which.
Pat put a number on it: talking to one customer is worth more than two weeks learning a new tool. A couple of marketing posts beat a marginally faster dev workflow every single time.
Ross's final call: he'll mess around with Claude Code a little, but he's not switching. Because if he goes deep, a month disappears and the app still has no users.
I've watched founders do the month-long version of this with new frameworks, new analytics stacks, new AI agents. The rebuild always feels like work. It almost never is.
What this means for your SaaS
Three things to steal from this episode:
1. Adopt the delegation workflow, whatever your tool. Batch the instructions, hand over the first 8 of 10 steps, review the output, test before shipping. Works with Claude Code, works with Cursor, works with a VA. If you want to see how far this goes, read how one founder turned Claude Code into a full AI employee.
2. Cap your tool-learning budget. New tools are easy to learn and easy to hide inside. Give a new tool an afternoon, not a month. The lazy approach wins more often than you'd think, and I've written about why building apps the lazy way outperforms grinding.
3. When two tasks compete, do the scarier one. The scary-and-vague task (distribution, pricing, talking to users) is almost always the one that grows revenue. The intimidating-but-knowable task is comfort work.
Your editor doesn't care about your MRR. Your customers do.
This is the exact kind of unglamorous, numbers-first conversation I dig into every week on the Profitable Founder Podcast, with bootstrapped founders doing $100K to $10M a year. Listen to the latest episode here.
FAQ
Is Claude Code better than Cursor?
For experienced builders, Pat found Claude Code faster and more accurate: it plans its own steps, needs no manually attached context, and handles multi-step features in one prompt. But Cursor's visual interface is friendlier for beginners, and Pat still runs Claude Code inside Cursor's terminal. Neither is "better" in the abstract. It depends on how much hand-holding you want to give the AI.
Should a beginner switch from Cursor to Claude Code?
Probably not right away. Ross spent 40 to 60 hours learning Cursor and shipped a working app with it. Pat's advice to him was to keep using what works and spend the saved time on customers instead. If your current tool ships product, the bottleneck isn't the tool.
Can you use Claude Code inside Cursor?
Yes. Claude Code runs in any terminal, including the one built into Cursor. That's Pat's exact setup: Cursor as the IDE for files and terminal, Claude Code for the actual AI work.
Is switching dev tools worth the time for a founder?
Run Pat's test: compare a week learning the new tool against a week spent getting 100 customers, then notice which one you're avoiding. Tool-switching usually delivers marginal speed gains while the scary work (distribution, pricing, user conversations) sits untouched. Switch when your tool blocks shipping, not when your timeline makes you feel behind.
