
Building Mode: Adventures in Founder Land
I predict this article will be shorter than my last post. Not because I don’t have much to write—I have
too much
. My bandwidth is shrinking as we’re now in full building mode.
Weeks 2 & 3: The Expected Team Breakups
As anticipated, weeks 2 and 3 saw more team breakups. I’ve been providing additional support to teams I’ve been working and volunteering to work with. One team paused our conversation, while another joined one of our calls in exchange for my help building their product.
The AI Dress Transformation Project
This founder had a fascinating idea: helping boutiques sell their rumpled, basic-looking dresses by running them through an AI pipeline and producing stunning images on the other end. Immediately saw the potential—I know many
sellers struggling to present unattractive inventory that could benefit from better presentation.
It’s an exciting concept that doesn’t require deep tech infrastructure, though it’s probably not very VC-backable (there’s virtually no technological or commercial moat). Still, I enjoyed building part of it with the founder.
The Technical Challenge
The workflow seemed straightforward at first:
-
Get users to upload dress images
-
Run reverse image searches
-
Push product data to Shopify
However, we hit roadblocks:
-
The Shopify plugin needed actual website links but had no programmable API
-
Reverse image searches yielded untrustable websites with unreliable information
Rethinking Our Approach
We took a step back and reassessed the core problem:
Recommended by LinkedIn
[Constraints
Dunni Abiodun
7 months ago](https://www.linkedin.com/pulse/constraints-dunni-abiodun-zzfoe)
[Tech recession or reality check?
Temitayo Diadem Adegoke. DBA.
3 years ago](https://www.linkedin.com/pulse/tech-recession-reality-check-temitayo-diadem-adegoke-)
[No One Cares About Your AI Feature Until It Saves…
Jephtah Uche
1 year ago](https://www.linkedin.com/pulse/one-cares-your-ai-feature-until-saves-time-money-jephtah-uche-l9ele)
Original pipeline:
Dress → Accurate link → Shopify
What if we replaced the middle step with something less deterministic but good enough?
Dress → AI web search → Shopify
Dress → Perplexity → Shopify
We tested this with direct image uploads and implemented it using Zapier. When Perplexity’s Zap had issues handling images, we fell back to ChatGPT with decent results. I’d eventually prefer to switch back to Perplexity with a small Node.js API since I have Perplexity Pro access.
The solution works! We upload a dress, and it appears on Shopify. We successfully broke down the AI-generated description into fields for Shopify’s APIs. Next, the founder will apply her proprietary image refinement techniques using AI editing APIs.

What we endend up with
My Current Focus
I’ve decided to pursue something similar to what I was previously building for artists—a project where I felt genuine market pull. After working with various founders, I’ve settled on a consistent mentality: if an idea feels like pushing a boulder uphill without opening exciting rabbit holes of possibilities, I am more inclined to close that chapter. I felt this way about a couple of ideas recently even when the founders were “great peoples”.
Scrappy Building
I collaborated with a founder on a second-hand furniture concept, ideating around a flow to ingest millions of Facebook Marketplace listings and index them in our vector database to capture long-tail keywords then handle inbounds manually while we onboard sellers, quite unscalable. Many roadblocks on this one especially when it came to logistics and executing “last mile” properly because … furniture. In my opinion I will not bother with bundling logistics into this and would rather focus on ideas that capture value earlier in the purchasing lifesycle like
This week is about to get serious
For the coming week, I’m focusing on this product for artists—shipping experiments and using the experiments to validate our assumptions about the market.
Originally published in Blackbox Thinking on LinkedIn.