Zoom in for product fit
The primary challenge of an early-stage startup is always the product-market fit. You need evidence of a real problem you can solve in a novel way before you can scale.
When I joined UpToSpeed in June 2024, we were selling the idea of efficiently onboarding developers to a team. Our product—AI for code documentation—was developed to meet the needs of a new developer joining a team. We had to support a wide variety of languages, and each new codebase was a stress test for our tools.
We interviewed many developers, CTOs, and VPs of Engineering, and while the problem was real (delay in the ability of Software Engineers to be productive costs companies millions of dollars), there was very little traction for our solution. We heard that it is difficult to allow an AI tool to access the codebase; that problem, while real, is habitually ignored as unsolvable: spending 3-6 months onboarding Engineers to new codebases is perceived as just the cost of doing the business.
We saw a real problem. We had a reasonable solution. We could not bridge the gap.
Instead, we decided to look around and find a more robust product-market fit where we could make a difference faster. We found one in Web3 security, specifically smart contracts audit contests.
You see, the smart contract codebase is open, so we can run our AI tools freely. The contests are time-boxed, so there is a high pressure to learn a codebase on auditors. And the payoff is significant. Speeding up the auditor process by even a few hours may help to develop an edge and find a critical bug first.
With smart contract audits, we zoomed in on a highly specific problem. I created a new website, AuditLabs.ai. We opened the shop in October, targeting the Uniswap 4 big contest with our first “pack of code tours for a contest.”
In November, we attended the DeFi Security Summit in Bangkok, where we connected with dozens of security researchers, pitched our product, and received direct feedback. It was incredible: we had a room full of our potential customers open to try the product.
By early December 2024, we had a handful of customers and improved our system a lot to be able to target all the ongoing contests semi-automatically. We built a “AI machine” of highly specialized content creation and were ready to scale.
Zooming in on specific niche in the market allowed team to focus on one of few languages, improve the tools, and develop in-depth features. It also simplified outreach, sales, and overall business development.
In five months after “zooming in”, we had a robust sales pipeline with a few major auditing firms interested in using our products both for in-house private audits and as a way to support their public contests.
