Very basic prototype for ProposalGPT

*Note: I’m mad busy, so I just voice dumped thoughts into ChatGPT and asked it to write this newsletter. That’s why it’s a bit “AI-y”, I’d usually write more of it myself. It did a better job than I expected though tbf.

I’m way behind. But I’ve got a prototype (sort of).

So… I’m almost at Week 6 of my six-week challenge. And I’ve only really started building in the last few days 😬

I’ve got one week left. It’s going to be tricky to get it done, but I’m going to try.

I started out building in Replit. Things went okay at first, but I quickly got stuck — mostly with things like templates, saving to a database, and all the little wiring pieces. And I think I’ve figured out why.

A big learning for me has been this: When working with AI dev tools, you really need to break things down into very specific milestones — and only give the tool context around that milestone.

That’s almost the opposite of how I’d work with an engineer. With a person, you want to give them the full product vision so they can scope things out, build good schemas, and plan properly. But with AI, that actually seems to make things worse.

With AI dev tools, scoping out a small milestone, making that work, and then moving on has been a much better approach.

The second thing I’ve learned: I really don’t like the designs Replit produces.

So I found a bit of a hack using a tool called PolyMet. PolyMet is like Replit but focused on front-end — it doesn’t generate fully functional apps, but it does produce really nice React code and solid designs.

Here’s what’s been working:

1. I design the UI in PolyMet.

2. I download the React code.

3. I upload that to Replit and ask it to wire it up.

That combo has actually worked quite well for me so far.

For this first milestone, I scoped things really tightly:

Just a tool that can generate a proposal based on a short chat with the user. That’s it.

No template saving, no company setup, no database work — that’ll come later.

I’m going to share two short videos in this email:

📹 Designs in PolyMet

📹 Semi-functional prototype in Replit

To keep myself accountable for this last push, here’s my rough plan:

• Friday: Build a landing page with a discounted “buy now” button, and start sending it to a few people for early interest.

• Monday/Tuesday: Flesh out the MVP into something people could actually use.

• Wednesday: Tidy everything up and do a small push to potential users.

• Thursday: Try to launch.

That’s… tight.

Work has decided to get very busy just as I started this. But I’ll give it a go and see how far I can get.

I’ll try to send short updates every day. Let’s see what happens.