Why AI application teams are validating profit before growth

What the interview reports

2.5 lab, a Jike unit focused on AI innovation, has incubated products including the open-source AI client Chatbox, ChatHub, kulikuli, and a self-discipline tool called Self-Discipline Stone. The interview says these products had generated profit even though their user scale was not large.

Xia Junchen describes a market in which each major model upgrade can weaken applications built around a single capability. In that environment, product teams need a defensible use case and an operating model that can support ongoing service costs.

Why monetization comes before expansion

The interview contrasts the older internet growth formula—free acquisition, rapid DAU expansion, and monetization later—with AI services that incur real costs for model calls and service delivery. As usage rises, token spending can become a major expense.

Xia Junchen says token costs often account for about 70% of total revenue in AI applications, based on his experience. His stated sequence is to validate commercialization first and then pursue growth. The percentage is an interview claim, not an independently verified benchmark for every product.

The kulikuli example

kulikuli is described as a tourism translation application, with Google Translate identified in the interview as its most concerning competitor. The product was free for its first three months. Cost pressure led the team to consider stopping it, but a later paid version found users willing to pay.

The interview says kulikuli now has more than 3 million total users, mostly overseas users, a team of about 10 people, and profitable revenue. Those figures are attributed to the interview and do not constitute a forecast or guarantee of future performance.

What cannot be concluded

The interview says that services similar to Tencent Cloud TokenHub could cut costs for mature use cases by 50% to 70% while maintaining quality. This is Xia Junchen’s statement in the source; it is not evidence that every AI application can achieve the same reduction.

The event does not establish a permanent product advantage over Google Translate, a market ranking, an investment return, a regulatory conclusion, or an official relationship with any exchange. This article is informational, not financial advice, and it does not promise profits.

A careful assessment would also separate total users from active users, revenue from gross transaction value, and token expense from total operating cost. The event does not provide a complete operating statement or an independent comparison with other translation products, so those conclusions remain outside the available facts.