Last week, I missed the newsletter for 2 reasons:
First, I'm admittedly behind.
Three weeks behind, if I’m honest, yet I've never felt calmer about a blown deadline in my life.
I wanted CHUMP live by now, but Father's Day, my Father’s birthday, the Fourth of July and the World Cup have had me focusing on community and family as of late and that’s okay. If you didn’t know, I’m a BIG soccer fan, so the World Cup was always going to be a busy time for me, so the timeline slipped.
And, here's the thing about building solo: I make my own timelines, which means I also make my own stress and that’s reason one. I'm behind, and I've made peace with it.
Reason two is that I'm about to slingshot myself straight into the future.
A few days ago, I did something I hadn't tried before. I opened Claude, and instead of asking it to build the next thing, I asked it to finally boost up to Fable 5, its newest and best model to date, and grade everything it had already built on lesser models.
Fable 5 is a real step up from the 4.6 and 4.8 models I'd been building CHUMP with for months, so I gave Claude a simple assignment: You're the senior. The freshmen built this. Go read their work and tell me everything that's wrong with it.
I can't take full credit for the idea — I saw a version of this prompt on LinkedIn and I had to try it out. My older models were freshmen, heads-down, building fast and building a lot. Fable 5 is the senior who's seen more, knows more, and can walk through the whole codebase with fresh, sharper eyes, so I pointed it at CHUMP and asked for three things:
Redundancies to remove. What's duplicated, tangled, or doing the same job twice?
Things to improve. What works, but works worse than it should?
Things to delete. What's dead weight? What's actually holding this thing back?
The leverage here isn't that AI builds fast. The leverage is that when a better model shows up, you can have it easily review months of previous work and receive a senior-level audit of your entire product in an afternoon all for the price of a good prompt.
I’ve said this before here in this newsletter and a lot lately in IRL:
The superpower of building with AI is its speed of iteration.
And, what do you know? Fable 5 handed me a huge list of things.
Not vague notes. Specific, actionable, "here's the redundant logic on this screen and here's the cleaner version" kind of feedback. I took the whole thing and organized it into a checklist, and now I'm just working the checklist, top to bottom. It is, without exaggeration, one of the most efficient stretches of building I've had since I started creating CHUMP.
So, here’s some unsolicited advice:
If you're building anything with Claude, steal this. Open your project and say some version of:
"We've been building this with 4.6 and 4.8 for a while. Fable 5 is newer and stronger. I want you to treat everything we've built so far as a first draft written by a junior version of you. Review all of it as the senior. Give me three lists: redundancies we can remove, things we can improve, and things we can delete because they're holding us back."
Then watch what it finds, as it’ll likely be both humbling and incredibly useful.
I'm running that checklist right now. I'm also still wrestling the Plaid API into submission — connecting real bank data is exactly as finicky as it sounds, and it's been the last real boss standing between me and launch, and I'm closer today to launch than I’ve ever been.
Behind isn't dead and the tools keep getting better while we sleep, so being a few weeks late has never cost less than it does right now.
If you're building something today, and a stronger model drops, what’s the downside, if any, of letting that new model audit everything you made with the old one?
As always – nothing replaces action.
See you next Tuesday.
CHUMP is almost live. To be there on day one, visit hichump.com today.
If you want to follow the journey building CHUMP solo with AI, please subscribe.