ByteDance, Tencent, Alibaba and Baidu are using red packets, subsidies and new AI apps to force habit formation. The real test comes after the holiday, when the giveaways stop
The framing of this as a gateway war rather than an app race is really sharp. The point about inference costs scaling with usage is something I don't see discussed enough — buying a massive user base that's economically negative to serve is a real trap. Curious whether the retention numbers post-festival will look anything like the WeChat Payments playbook from 2015, or if AI assistants are just too hard to make sticky without genuine utility loops.
Thanks for reading. Agree on the inference-cost trap.
On retention, I wouldn’t map this 1:1 to WeChat Pay 2015. Payments had a simple, repeatable loop and low marginal cost. AI is harder to make sticky unless it’s tied to real workflows. The post-festival test is whether usage holds once incentives stop, and whether engagement shifts from rewards to utility.
The framing of this as a gateway war rather than an app race is really sharp. The point about inference costs scaling with usage is something I don't see discussed enough — buying a massive user base that's economically negative to serve is a real trap. Curious whether the retention numbers post-festival will look anything like the WeChat Payments playbook from 2015, or if AI assistants are just too hard to make sticky without genuine utility loops.
Thanks for reading. Agree on the inference-cost trap.
On retention, I wouldn’t map this 1:1 to WeChat Pay 2015. Payments had a simple, repeatable loop and low marginal cost. AI is harder to make sticky unless it’s tied to real workflows. The post-festival test is whether usage holds once incentives stop, and whether engagement shifts from rewards to utility.