Cursor

Founders note: I noticed that OpenAI and Claude are way too good at telling me how great I am at everything. Cursor doesn’t do that.

What Cursor is

Cursor was the first AI-first IDE and code editor. I use it as my Claude, Codex, or Windsurf solution — I do everything in it. It is not instant — expect a couple of weeks before a non-technical founder is truly at home — but the business models are not the same. An AI that makes you sound like the smartest, most handsome guy in the world is going to get you revenue. Cursor gets its revenue because I can deliver eight hours worth of code in an hour.

I thought: what happens if I can treat my content like code? I can search it. I can index it. I can build rules. I can do everything that I would do in Claude, but I can do it in a tool that sits on my desktop. I don’t have to be on the cloud. And I don’t have to give it all away. I actually control it. It’s like Claude or Codex or Copilot, but in code with a Git repo that tracks every change. That’s pretty powerful.

The assumption trap

Developers think Cursor is a great way to write code, but not much different than VS Code or any other IDE. I think it should be viewed differently. If you break the paradigm and blur your eyes, you’ll notice that it’s actually a tool that does way more than what developers normally think about.

One shared stage

Both a technical and a non-technical founder in the early days need to be talking about and working from the same stage. You could set up a non-technical founder with Cursor and a few small commands and MCP commands and level the playing field so your non-technical founder and your technical founder are having the same conversation with the same background, context, history, and archives that a developer would. There’s an inherent value in bringing a CEO into the code world and potentially having the CTO or the coder sit in the sales world.

The CEO/CTO divide

It’s kind of hard to regret what you don’t have. But almost every founder I know looks back and realizes they spent months or years doing what they should have learned faster. What we don’t know is what kills us. What we don’t know is what slows us down. In every founding team there is an inherent divide between the way the CEO and the CTO approach problems. The CTOs usually look down at code and the CEO is looking up at investors or sales. That’s a problem because that means only 50% of the team is in operation mode and 50% of the team is in build mode. And the bridges, while they’re there, they’re a little weaker than we might imagine.

The Meenta lesson

Meenta was a biotech startup — the Airbnb for genetic sequencers. If I could go back and get my CEO who was 100% non-technical talking to the code using AI and Cursor and agents and rules and structures, I wonder what Meenta could have done. I don’t think that code is purely a technical domain knowledge. Some of the concepts are universal — and if we break down the barrier, my last company could have done amazing things. It might be a very bad idea to keep the walls up and keep the two domains separate.

In retrospect, we probably should have spent more time in operations mode and less time in build mode. But we spent the first three years building the tech that we thought we could sell to core labs and genetic sequencing facilities.

Why investors might care

Investors are looking for scalar engines where cash goes in one side and an exit comes out the other side. They may say they want to do the Berkshire halfway approach of buy it and hold it. I think that’s all bullshit. Too many venture-backed startups fail. A good investor can make 30 investments and hope that one or two actually pan out. So they need a 5, a 10, or a 50x return. The faster a team can go from build to operation mode and the more effective they can move across those, I have to think — although I haven’t proven it yet — that this is a hugely important value to an investor. But I don’t know if investors can see that far into the minds of the people they invest in. What I’ve found so far is investors typically look at the founding team, but they look most closely at the CEO because that’s the neck they can wring.

Three steps for a non-technical founder

1. Install Cursor, get a paid package, set up a repo. You need a real workspace — not a browser tab that disappears.
2. Wire MCP to GitHub. Learn one or two simple prompts so you’re talking to the same history a developer would.
3. Set up rules and structures for how your company thinks. Externalizing your thinking in markdown is a huge gain. It lets you capture blind spots and keep yourself on track. I have a rule that says every time I come up with a new idea, I document it into an MD file and I run it against my ICP because there are so many signals in complex systems that are easy to miss.

At a glance

• What it is: An AI-first IDE on your desktop — rules, agents, MCP, markdown-as-company-brain, Git behind every change.
• Best for: Founding teams who want one shared stage — CEO and CTO in the same repo, same context, same archives — not parallel silos in Notion, Git, and ChatGPT threads.
• Not a fit: A quick chat answer or ego-friendly brainstorming in the browser; if you won’t shepherd a non-technical founder through the first couple of weeks, time-to-wow stays painful.

When to reach for it

Reach for Cursor when you want throughput, not flattery — and when you’re ready to treat content like code. If your co-founder stack is GitHub plus Notion plus ChatGPT, consolidate the thinking into markdown in a repo both of you can open. The Meenta lesson applies: don’t keep the walls up between code world and sales world.

See it in the stack

• Snowflake CRM — why I built a one-user CRM and what Cursor throughput made possible.
• Vibecoding Calendly? Really? — screenshot-and-rebuild workflow; Cursor did the UI in under five minutes.

Related playbook entries

• Astro — flat files in git are AI-ready infrastructure; Cursor edits the archive agents should copy.
• Inngest — background jobs your Cursor-built stack still needs orchestrated.
• Postmark — product email wired from the same MCP bench.
• PostHog — first-party analytics and identify/backfill on the Snowflake stack.

AI prompts for the first week

Drop these into Cursor after install. Tweak repo names to match your stack.

Read the stack first: Snowflake CRM tech stack — see Cursor next to Supabase, Postmark, and Inngest in production.

Tech Stack Clarity Check (15 min) — Book a slot if you want a second pair of eyes on shepherding a non-technical co-founder into Cursor, MCP wiring, or rules setup.