Software is a Middle Manager

Why every modal sounds like Bill Lumbergh

It started with a prank modal because I wanted Maddi to roll her eyes.

I was building a little internal tool and I couldn’t resist: a deliberately irritating pop-up written in the same stuff I say out loud when I’m being “efficient” in the way that makes other people want to throw me into the East River.

“Quick thing.”

“Just gonna need you to...”

“Real fast.”

Maddi always tells me those phrases bug her because they’re not actually quick, and they’re never actually one thing. They’re the verbal version of a meeting invite with no agenda.

So I wrote the modal in that voice and shipped it. Then I realized I was seeing it everywhere.

The forced labor:

“Yeah, gonna need you to go ahead and sync your contacts real quick. Just everyone you’ve ever emailed. For your experience. That’d be great.”

The security loop:

“Yeah so, we noticed you’re logging in from a new device. Gonna need you to verify by clicking the link in your email. Which you can’t access. Because you’re not logged in. Yeahhhh. Soo.”

The redesign:

“Yeah so we moved everything. It’s more intuitive now. The thing you used every day is under Settings, then Advanced, then Advanced. The bathroom’s behind a hamburger menu now.”

You stop seeing these as prompts and start seeing them as calendar invites. Meetings you never scheduled with software that reports to a dashboard you’ll never see.

This was all bouncing around in my head when Kevin Roose tweeted something about the gap between SF and everyone else. People here running multi-agent swarms while the rest of the world is still trying to get Copilot approved.

Kevin Roose@kevinroose

i follow AI adoption pretty closely, and i have never seen such a yawning inside/outside gap. people in SF are putting multi-agent claudeswarms in charge of their lives, consulting chatbots before every decision, wireheading to a degree only sci-fi writers dared to imagine.

4:39 PM · Jan 25, 2026 · 2.51M Views

---

678 Replies · 459 Reposts · 5.94K Likes

I’m always interested in different ways that haves and have-nots emerge. They tend to rhyme. Something folks might not know about me is that I studied history. (I’m kidding. I never shut up about it.) This is why it was sitting heavy in my brain for a few days.

Then Kevin & Casey mentioned it on Hard Fork. Casey’s reframe was the thing that stuck with me: the skeptics aren’t saying AI is fake.

They’re saying, “I don’t see how this helps me.”

Then it clicked.

To understand the click, you need to know three things I’d been noodling on for the past couple weeks:

1. I used to be a good texter. Now I’m terrible. I leave people on read for days. Why do I hate my phone so much?

2. I’ve been very proud of myself that we canceled our CRM subscription. Or as I like to say, “I fired Affinity.” Canceling my cable subscription was awesome. Canceling a B2B SaaS contract is 100x that.

3. The Lumbergh thing. Funny observation, but something deeper underneath it I couldn’t name.

The click was pretty simple:

I realized I was reaching for my laptop way more than my phone because every app on my phone talks like Lumbergh. That's why I hated using it. That's why firing Affinity felt so good. Claude let me skip the modals and just do the thing I wanted to do. It just happened to live on my laptop.

And Casey’s skeptics? They’re not wrong. They’re just evaluating AI inside the old architecture. If AI lives in the UI, it’s just Lumbergh with better autocomplete and more ways to sell you stuff. But if AI is what lets you skip the UI entirely, that’s a different proposition.

Back to the Affinity thing. Why was I so proud of firing a CRM?

Honestly? It was helpful. They ingested my email, indexed my calendar, and surfaced patterns I wouldn’t have spotted on my own. “Last Contacted” told me who I was neglecting. The relationship analytics were real.

But one day I tried to articulate what I was actually paying for. They didn’t source my deals. They didn’t build my relationships. They didn’t close anything. They organized my data and showed it back to me.

That used to be worth $6,000 a year. It felt like a massive value add at the time. And maybe it was, when the alternative was doing it all manually or paying a developer to build something custom.

Then I realized I could just... do it myself. Describe what I wanted. Have Claude pull the data, run the analysis, surface the patterns. Same insights, no interface, no subscription, no meetings with modals.

The value didn’t disappear. It just got cheap.

So I stopped. Weirdly freeing.

But I still needed to do the things the CRM was doing. Track relationships. Surface patterns. Know who I was neglecting. That’s when I actually understood the two-door thing.

Most software has a front door and a side door. The front door is the app, the interface, the modals. That’s where Lumbergh lives. The side door is the API. Same data, same capabilities, none of the theater. You request something, you get it.

I’m not a developer. I was a banker. I spent years in Excel and I still think in spreadsheets. But I don’t need to write code anymore. I describe what I want and Claude pulls the data. No interface, no subscription.

The door was always there. It just used to require a skill I didn’t have. The learning curve went from “years of computer science” to “a few stubborn weekends.”

I don’t feel more productive. Honestly I don’t know if I am. But I feel less frustrated. Turns out I don’t mind talking to computers. I just don’t want to take meetings with them.