This Is How I Bullied ChatGPT Into Building Lean Products With Me

Today’s theme: how I argue with ChatGPT about features until one of us gives up. Spoiler: it’s never me.

I’m shipping the next version of indie10k’s main dashboard. Beta folks are moving closer to $10k MRR (🔥), but I noticed two pain points:

Some people stall and stop shipping.

Others keep clicking next without proof they actually did the thing.

That breaks the TenK 6 loop. No ship = no momentum. No evidence = no way to fine-tune later.

Step 1: Throw spaghetti

I tossed ChatGPT a list of ideas: daily hacks, weekly retros, better Arc loops, journaling small wins. Then told it: “use first principles and tell me what’s actually worth building.”

It spit back a decent analysis. But decent isn’t good enough. So I kept poking:

“Nah, too bloated.”

“That’s someone else’s product.”

“Explain it like I’m five.”

“Cut harder.”

Step 2: Trim the fat

One by one, we killed ideas:

Public feed? Backlog.

Experiment ledger? Nope, duplicate.

Header counters? Cute, but bloat.

What survived? A dead-simple private feed. Just a place on the dashboard where you log today’s progress, drop a link/metric/note, and maybe use it as a mini retro. While all feed item are TenK 6 loop data. No break of core flow. That’s it. No fluff.

Step 3: Momentum check

So far in TenK 6:

List 5 ✅

Pick 1 ✅

Ship 1 🚧 in progress

Momentum feels good. ChatGPT plays the overeager intern; I play the grumpy boss with scissors. The combo works.

Next step: ship the private feed, then move to Ask 3.

Lessons Learned

ChatGPT is great at brainstorming, but terrible at saying no. If I don’t play the grumpy boss, I end up with 10 shiny features I’ll never ship.

First principles cut deeper than feature lists. Every time I asked “does this actually help someone hit $10k MRR faster?” another feature died. And that’s good.

Plain English saves the day. Fancy words = easy to fool yourself. When I forced it to explain things simply, the real winners (and losers) showed up.

Momentum > perfection. The point wasn’t to design the “perfect dashboard.” It was to keep shipping lean slices that push indie devs forward.

At the end, what stuck was a tiny private feed. Not sexy. Not overengineered. But exactly what helps the loop stay alive.

Here’s the messy transcript if you like feature autopsies: link

Just in case you don't know what is TenK 6 Methodology.

You can even read a Pocked-size Handbook about this methodology.