About CostFunction

The Origin Story

CostFunction was born from a simple frustration: after decades of working with images at scale, founder Broc found himself spending hours running the same tedious tests every time a new LLM model dropped. Which one handles product photography better? Which gives the most accurate descriptions? What's the actual cost when you're processing 10,000 images a day?

The answer shouldn't take a week of manual testing. So he built the tool he wished existed.

30 Years in the Making

Broc's journey to CostFunction started in the early '90s behind a camera. As a photographer, he learned firsthand the chaos of managing thousands of images—the endless folders, the missing metadata, the nightmare of finding that one perfect shot from three shoots ago.

That pain led him into digital asset management before most people knew what DAM meant. He spent years building systems to organize, tag, and retrieve visual content for enterprises drowning in media files. Along the way, he discovered that the hardest problem wasn't storage—it was understanding what was actually in the images.

Enter NLP and machine learning. Broc dove deep into natural language processing, exploring how machines could describe and categorize visual content. He worked on early AI systems that could auto-tag photos, generate captions, and extract meaning from pixels—long before this became mainstream.

His work in film and television production added another dimension. Managing media assets for productions taught him that every second counts, every dollar matters, and creative teams need tools that work at the speed of thought—not tools that require a PhD to operate.

Why LLMs Changed Everything

When large language models gained vision capabilities, Broc saw both the promise and the problem. Finally, AI could truly understand images—describe them, analyze them, extract insights that would have taken humans hours. But with GPT-4, Claude, Gemini, and a dozen other models competing for attention, how do you know which one to use?

The benchmarks published by AI companies are impressive but generic. Your use case isn't generic. Your images aren't generic. Your budget isn't infinite.

CostFunction exists because the right model for analyzing medical imagery isn't the same as the right model for e-commerce product photos. And the only way to know which is which—for your specific needs—is to test them side by side, with your actual images, and get real numbers.

The Philosophy

Get In Touch

CostFunction is built for developers, product managers, and anyone who needs to make smart decisions about LLM model selection. Have questions? Ideas? Feedback? We'd love to hear from you.