Inspiration

With so many startups emerging, especially around AI, it can be difficult to tell genuine progress from excitement. I’m building a startup myself, and I kept wishing I had something to help me understand how we were doing, what mattered, and what I was missing.

So, with two teammates, I built SignalShift: something I would have wanted beside me when starting out. Founder optimism is useful, but occasionally it needs a second opinion.

What it does

Enter a company’s name or website, and SignalShift brings together public evidence about its adoption, attention, financial health, competition and team. You can inspect sources, explore experimental signals and compare companies.

It helps you ask better questions about a startup, including your own.

How we built it

We combined a Python research pipeline, machine-learning experiments and Gemini-assisted evidence interpretation with a React interface.

You can trace findings back to their sources, check their dates and see what information is missing. That’s what makes me trust it more: I can question the answer and inspect what supports it.

Challenges and lessons

The biggest challenge was learning how easily a promising number can be misunderstood. Downloads aren’t customers, funding isn’t profit, and a confident explanation isn’t necessarily correct.

We also dealt with incomplete data, API limits and models that didn’t pass our checks. Those experiences shaped the product: experimental estimates stay labelled, and unsupported predictions stay unavailable.