Replit has recently changed its pricing model, reflecting broader trends in AI product pricing. As AI agents become more autonomous, pricing becomes less predictable. Traditional usage-based billing causes anxiety for users since they can’t forecast their costs accurately. Instead, many companies are shifting toward prepaid credits, which allow for a more controlled spending experience despite not fundamentally altering economic realities. The transition to "effort-based" pricing complicates matters further. With Replit, a task that previously cost a flat $0.25 now varies based on the complexity and duration of the request. This lack of transparency can frustrate users; they might see a credit balance but struggle to understand how that translates to actual costs. The uncertainty in pricing impacts user experience, making it feel more like a gamble than a straightforward software interaction. Looking ahead, credits will likely continue to dominate because they address some of the unpredictability in AI costs. However, there's a clear need for better billing transparency. Suggestions include fixed costs for simple tasks, upfront estimates for more complex jobs, and post-task receipts explaining costs. Such measures could demystify spending in AI products, making them feel more reliable and user-friendly.