🧙🏼 OpenAI's enterprise takeover begins

SUBSCRIBE | ADVERTISE

Howdy wizards,

Welcome to the 270 new subscribers that joined last week.

In this edition:

• ChatGPT plugs into your stack with MCP connectors, Record Mode and flex pricing

• I spent yesterday building an analytics app with Cursor and Claude

Treat yourself to some damn good coffee (or herbal tea if it’s after 4pm) and recline in your best chair.

Here’s what’s brewing in AI.

DARIO’S PICKS

1. ChatGPT’s custom connectors, record mode and flex pricing

OpenAI just entered the enterprise market—in a BIG way. Here’s everything they announced for ChatGPT this week:

• MCP connectors: ChatGPT can now pull in real-time context from your company’s tools. That’s right—Deep Research now gets to feast on your business docs, data nightmares and black holes via pre-built connectors like SharePoint, Outlook, Teams, Google Drive, Gmail, Dropbox, HubSpot, GitHub and more. It’s also possible to build custom MCP servers for other third-party tools and connect it to ChatGPT.

  • Availability of these connectors varies: some are available only to Team and Enterprise, while others are available to Plus and Pro users as well.

• ChatGPT now has a Record Mode on the MacOS desktop app, but it’s playing hard to get (only available on the Teams plan). It's basically every AI meeting note-taker you've tried, but this one can seamlessly refer to the transcripts in your new chats with ChatGPT afterwards.

• Flexible pricing: Enterprise plans (and soon also Team) are getting credit-based access for fancy models like o3. Think of it like a shared pool of AI juice that your team can collectively drain. And if the pool gets exhausted companies can just throw money at the problem (as is tradition in enterprise tech) and buy more credits.

‎ Why it matters‎ ‎ ChatGPT wants to be a surface that you work from, not just a tool on the side.

It can now draw context from across your business data. Out of the public MCPs so far, I think Hubspot deserves extra consideration. The platform is widely used but difficult to navigate—huge mess in most firms. If Deep Research can make GTM teams spend less time in there and more time selling your product, that’s good news for your bottom line.

At the same time—while there’s no doubt that MCPs are promising—I’m a bit wary of hype, having experienced the buzz around custom GPTs connected to third-party APIs. Sounded great, but ended up kind of a flop.

The Record mode? Nothing revolutionary here—but yet another compelling reason not to context-switch away from ChatGPT.

The Enterprise credit-based pricing makes sense. Big companies want to democratize AI access but panic about adding 30,000 users @ $25/mo. This lets everyone experiment without casual users making the math look embarrassing. Plus, AI usage is addictive: start with one workflow, end up automating your entire existence. As adoption increases, the credit pools will vanish—and that’s exactly what OpenAI is counting on.

IN PARTNERSHIP WITH GUIDDE

Create how-to videos fast and easy with AI

Tired of explaining the same workflow to your team for the fifteenth time this month?

Stop being a human help desk—delegate that work to AI. 

Guidde is an AI tool that automatically generates polished step-by-step video guides with visuals and voiceover.

How it works:

1. Click capture on the browser extension

2. Do your normal workflow—Guidde's AI tracks every action

3. Get a complete video guide that your team can actually follow

Easily share or embed your guide anywhere.

Making how-to videos for your team with AI is one of those use cases that’s incredibly simple to get started with, and saves you time from the first minute.

The extension is 100% free.

Check out Guidde

UP CLOSE

In this mini-series I share different ways I’m using AI from week to week, as well as practical tips & tricks I discover and actually use.

Building my first web app to solve a real business problem with AI

I spent my Saturday having a solo hackathon building a tool I couldn't find anywhere else. Why? I really wanted to learn Cursor. That, and I saw a LinkedIn post about a marketing metric that shows how profitable your advertising really is, wanted to calculate it for my newsletter, but couldn't find any analytics tools that would actually work with my subscriber data.

Everything was either too simplistic or built for different business models. Nothing let me use the vast amount of data I had sitting there.

So I decided to build my own tool using Claude Sonnet 4 inside Cursor.

Quick note: The screenshots below show a demo version using fictional data and abstracted terminology. I created this demo mode in case I want to turn this into a product later - basically obscuring my specific use case while showcasing the same analytical capabilities.

This rewired my brain about what’s possible

I have some basic coding knowledge (very basic) but my background is more in marketing/product, so I expected to hit technical walls constantly. Instead, I discovered something incredible: you can build actual software that's useful without even looking at the code. I'd describe what I wanted—

"build an intuitive UI with configurable parameters. Then, group my data into cohorts by acquisition channel and create an interactive chart showing performance trends over time. Add the option to forecast data where it’s insufficient"

—thinking no f’in way that’s gonna work.

Then a couple of minutes later I was looking at it in my browser.

I’ve heard Cursor can start messing up at some point when your code reaches a certain size, but frankly I’m mindblown by how well it’s able to handle complexity, and it didn’t happen to me. To be on the safe side, though, I had Claude set up GitHub for my project. Never done that before—frankly didn’t even understand what GH was for. But it did the whole thing for me in 10 minutes. Turns out it’s like passing a checkpoint in Super Mario—you can keep adding layers of complexity on what you've already built with zero fear that it’ll break.

One key thing about AI that keeps standing out to me is that to get good results you need to actually understand the problem you're trying to solve. If not, you simply get the solution to a problem that looks similar, but it's not yours. That's why I spent most of the time reasoning—and asking Claude questions about—why we're implementing different methods and how. Then having Claude refine its approach. "Why do I need a cohort breakdown per channel?" "Why does the table show slightly different figures than the chart?" "How are we using historical data for future projections?".

This is the essential skill now. AI will readily implement something that might look like it's solving your problem, but there's probably a better way. The more you refine AI's understanding of your actual problem, the better the outcome.

Before I knew it, I had a tool that automatically and in a nuanced way analyzes my marketing performance and tells me exactly where to focus ad spend. The tool has already paid for itself in clearer decisions.

What’s breaking in the matrix

Most people and companies haven't grasped this yet: if you can clearly articulate a problem and reason through solutions, you can now build professional-grade tools to solve it.

With a $20/month Cursor subscription, I built something that would have cost thousands to commission. That math is absolutely ridiculous.

The gap between "this would be useful" and "working software" has never been smaller. The only bottleneck now is your willingness to experiment. That, and the temptation of an unhinged Netflix binge with a side of Doritos.

Look, I know the "where do I even start" paralysis is real. But here's the thing: you already know your business problems better than any developer you'd hire. Start there. Pick one annoying thing. Open Cursor. Just... try stuff.

PS these articles aren’t tutorials or walkthroughs—they're meant to inspire a mindset of augmenting yourself with AI. Because understanding how to break down your actual problems, articulate them and iterate your way to solutions is going to outlast any prompt I can give you.

VERY IMPORTANT AI TRAINING

This might’ve sounded like an advertisement for Cursor, but the real advertisement here is whatplugin FM - the playlist that powered my entire weekend hackathon.

Trust me, nothing brings Dexter’s Lab vibes more than 'Inspector Norse' by Todd Terje.

My Auditory Research Division reports increased coding confidence of an impressive 73% in a recent, double-blind study.

Listen to whatplugin FM here

THAT’S ALL FOLKS!

If you enjoyed—or didn’t enjoy—this edition, I’d be super thankful if you took a minute out of your day to write me feedback. What do you like, what do you want more of, less of, etc. Hit reply and let me know (I read every email).

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.