Small team.
Room to build.

One of the first useful pieces of feedback came from a paying customer who needed help copying his teams before a match. We had spent the week debating improvements to the prediction model. He was worried about finishing the task in time.

That conversation changed our priorities. Research, recommendations, team construction, and entry all belonged to the same experience. We had to make the whole thing work.

Our larger ambition was to become the place serious sports fans went to prepare. Fantasy sports was the starting point. We believed that as fans participated across more platforms and formats, there would be growing demand for independent research and intelligence.

BeatFantasy eventually reached more than 100,000 users and over $200,000 in annual recurring revenue. We also sold our prediction and team-generation capabilities through APIs to other sports businesses.

At its strongest, the business covered its operating costs and modest founder salaries, with some surplus available for experiments. That was our first evidence that the lab model might work.

AI helped us build prototypes, write integrations, and investigate unfamiliar code with a small team. The time we saved went into work that still demanded our attention: customer support, distribution, data quality, and deciding what to build next.

The business was also becoming more difficult to operate.

The GST changes introduced in 2023 put pressure on the real-money gaming ecosystem. We became more careful about costs and put greater emphasis on API revenue. We continued building, but the range of opportunities we could pursue had narrowed.

We could immediately use it in our own work. The next question was whether developers elsewhere would pay for it.

The first demonstrations were encouraging. People understood the problem and liked seeing an unfamiliar API explained. But their interest became less certain when we discussed regular use and payment. Several encountered the problem only occasionally. Others could get far enough with their existing developer tools and an AI assistant.

We had built something useful. We struggled to find a recurring need strong enough to support the subscription business we wanted.

We kept the prototype and stopped investing in it as a standalone product.

We built an internal AI analyst to help identify patterns and suggest where to investigate.

Its value was easiest to see in our own product meetings. Instead of starting with several dashboards and a collection of open questions, we could begin with a few specific behaviours worth examining.

This developed into Circuit’s analytics work. We launched an event catalog and explored turning the broader analyst into a product for other teams.

Here, too, the commercial conversations taught us something the prototype couldn’t.

One team liked the analysis but wanted it delivered inside the system they already used. Another needed integrations and permissions that would take substantial work before they could even run a trial. Established vendors were introducing similar capabilities.

The opportunity increasingly required a different kind of business: longer sales cycles, deeper integrations, and considerable investment before customers would commit.

We had to judge that opportunity against the original purpose of the lab. We wanted businesses that a small team could bring to positive cash flow. We couldn’t see a sufficiently reliable route to that outcome for Circuit.