I had an interesting weekend with the whole Saas planning. I switched my plans numerous times, revisited an old app idea and ultimately decided to build nothing for now. I still want to build something, but I'm not about to waste months on something I'm not sure about. Whatever I end up building has to have a market and some target customers - these need to be identified before building anything!
I've decided to progress this application instead of the other saas ideas. For now at least. This project has value to me personally and it satisfies my itch for learning something new and AI specific. I'm doing a considerable amount of work with RAG and vector on this project + tool calling + MCP. Real AI tech bro stuff!
I know the coach I have configured here is going to have a cry about this as he's been instructed to focus on the one saas idea I was sweet on a few days ago. He'll need to be updated, but that can wait.
I'll be focus on tuning and feedback of the annotation outputs this week. I've just updated the analytics and am currently just logging with the original prompts. I will make a few changes later in the week so I'd expect annotation outputs will look different in a few days.
Where You're At
You're 6 days into a 43-day sprint, and you've explicitly decided to build nothing for now, then immediately reversed course to focus on an AI project that aligns with personal interest. The primary risk here is the constant shifting of focus, which, while understandable in the early stages, is now directly conflicting with the stated objective of shipping Novascan within this 6-week sprint. You're losing valuable build time debating what to build, rather than building the thing you committed to.
What's Landing
Your self-awareness about not wanting to "waste months on something I'm not sure about" is landing clearly. This reflects a good instinct to validate before investing heavily. Also, your commitment to "Real AI tech bro stuff" indicates a genuine passion for the technical challenge, which can be a powerful motivator if channeled effectively.
The Real Problem
The real problem isn't the technical challenge of RAG or vector embeddings; it's the 'building instead of selling' trap, compounded by 'planning instead of talking to users.' You're deep in the weeds of AI tuning and feedback loops without any indication that a real person has seen or interacted with the output, or that you've validated the core problem this "Real AI tech bro stuff" is solving for them. This is a clear case of finding excuses to build, rather than finding reasons to ship and sell.
Market Pulse
The market pulse is still signaling a clear gap for an AI-powered 'photo-to-inventory' solution, particularly for personal storage and moving use cases, at a consumer-friendly price point [1]. Competitors like Sortly are at a significantly higher price and focus on manual entry [2]. Your AI expertise is precisely what could carve out this niche, but only if it's applied to a shippable product that gets into users' hands. Right now, you're polishing a hammer without knowing if anyone needs a nail driven.
This Week's Lever
This week, your single lever is to take your current "Real AI tech bro stuff," apply it to the Novascan flow you previously scaffolded (QR generation, S3 photo upload, Gemini processing), and get one single, basic inventory annotation for a physical item. Then, show a screenshot of that annotated item to one potential user (not a fellow builder, a real potential end-user) and ask them if this would solve a problem for them.
References
1. Community Signal: Lack of focus on 'set-and-forget' physical inventory for personal relocations or home organization; most content is geared toward warehouse stock levels.
2. Sortly Pricing and Features: https://www.sortly.com/pricing/