SparkTank - Edition 16

What we talked about.

Edition 16 of SparkTank brings four founders before a panel of investors, hosted by Sunny Ray. This portion of the episode covers two pitches. Si Shnata, a founder from Odisha, India, presents an engineering intelligence platform that measures what happens to AI-assisted code after it ships, linking signals from tools such as GitHub and GitLab to rework, quality risk and cost. The product is metadata first, so it does not need access to proprietary source code. The MVP is live in pilot testing, and Si asks for $50K to fund twelve months of development. Panelists press on competition from large coding tools, defensibility and how accurately AI-written code can be identified. Kieran Skelly of Houston then presents the Primer Wand, a squeeze-bottle applicator for PVC primer meant to prevent spills on job sites. Developed with master plumber Sebastian Merrill, it has completed a design for manufacturing run and has a non-provisional patent pending. Kieran seeks $150,000 for 20 percent, and investors question the equity, tip drying, production capacity and distribution plans.

Two founders pitch SparkTank investors: a platform that audits what AI-written code really costs, and a squeeze wand that ends plumbers' primer spills.

The questions, and the answers.

What stops GitHub Copilot or similar tools from building a competing product?

I'm not competing with those tools, I'm measuring them. Teams can keep using whatever they like, and we show what the code base looks like, what came from AI and what went wrong. Our memory layer helps future developers understand why changes were made, so companies stop bleeding money on avoidable breakages.

Why not take a percentage of the money you save customers?

We're already doing that. If we help a company save $5,000, they pay us $1,000 of it. These companies spend enormously on AI tools and have little visibility into wasted seats or code base health. No AI vendor will report that, because it doesn't fit their business plan, and that's our leverage.

How accurately can you identify AI-written code without seeing the source code?

We can't identify it with 100% certainty, and we only derive signals. What matters more is whether changes are breaking the code base or its structure. We track warning keywords and patterns, but we also measure code base health and what is happening inside the organization at the back end.

Is there a moat, or can someone copy this with AI for less?

It isn't impossible to copy us, since we're small and others do similar things. Our leverage is the backend memory layer. Once a software company installs it and it learns their patterns, they won't throw it away after three months. How well we serve clients is our strongest hope.

Isn't 20% for $150K too much equity to give up?

For us, time is money. This gets us on a path where we're making a million dollars in two years, which beats pitching a hundred times to keep a bit more equity. Sebastian and I would rather move quickly, and we want to choose investors who bring real industry relationships.

Doesn't the applicator tip drying out make this a concern?

The tip only dries if you forget the cap. With the cap on, it has worked for weeks across dozens of job sites. The tip is natural cotton and easily replaceable, so you can pull it out, fold it over and put it back in, or swap in a new one.

What is your go-to-market strategy and production capacity?

We started on Instagram, and distributors from Arkansas and the Northeast have asked to stock it. Our factory contact already sells gloves to Home Depot, so we're sending samples there. We start with direct sales to smaller stores. Our Shanghai factory can make 10,000 units a month.