Husky AI

Everyone posts about building entire apps in a day with AI but I rarely see anyone talk about what it actually looks like when you have paying customers, medium complexity infra, and expectations around quality.

So here's my breakdown:

~1h - Standup and feature discussion

~2h - Creating an implementation plan (back and forth with Claude Code on architecture, what scales, what we can reuse, infra changes)

13 min - Claude's actual implementation

~1h - Updating infra: secrets, new queue, permissions etc.

~1h - Fighting with Claude to follow our Figma design

~30 min - Adding missing functionality

~2h - Testing edge cases and fixing as I go

~1h - Testing on staging with the team and deploying live

Given that this same feature would've taken an engineer about a week before AI, it's a huge improvement. But I'm still spending around 5 hours on infra, testing, and QA and I feel like that can be improved even further.

Curious what others are doing to speed this up?

I’ve been building an AI app/website builder aimed at non-technical people. My experience is that most no-code/AI tools still assume you understand databases, APIs, UX, etc.

So I’m trying a different angle: instead of giving users a blank editor, the AI guides them step-by-step (similar to working with a small agency).

Where I’m at: Early stage. It can generate and publish production ready websites today.

I’d love feedback on whether guidance is actually the missing piece vs just better generation - and where people usually get stuck with existing tools.

Link if you want to poke around: https://www.huskystudio.ai

Very open to blunt feedback 🙏