Everyone is talking about AI, so I'll make a prophecy.

Tools like OpenClawd(ex Clawdbot) are essentially unstable solutions built by reverse-engineering APIs and CLI. In the long run, more and more apps will provide standardized interfaces for AI to call (MCP). But this also raises another issue: entrusting user interaction to AI will inevitably weaken certain types of applications(especially toC apps)'s control over user attention, behavioral data, and so on. Therefore, in the short term, these apps will tend to avoid or actively create obstacles.
Regarding the development of agents, there are two possible directions:
1. For the issue of standardized interfaces, a supplementary solution is to enable agents to control GUIs, similar to VLA in robotics, mapping visual observations and language instructions to action sequences, allowing agents to click, move the mouse, take screenshots, type, and so on. Therefore, I believe that without considering costs, it will still take some time for robots that perform on the grand stage to enter ordinary households, but the related technology for GUI control is much more feasible.
2. Giving agents payment capabilities is another possible addition I can think of. As for the currency used for payments, it could be fiat currency, stablecoins, or even BTC. But regardless of the method, it will definitely involve KYC, compliance, and other issues. Overall, I think companies like Stripe that provide payment infrastructure are good growth points for this need (by the way, they are already working on it but haven't gone public yet, and recently there are rumors of acquiring PayPal).
Combined with the recent flood of OpenClawd deployment tutorials and the sold-out Mac mini, aside from hype factors, this phenomenon to some extent reflects people's strong expectations for personal agents. However, local deployment solutions face hardware barriers, maintenance costs, stability issues, and other problems that individual users find difficult to solve. So I believe that the true scalable end state of personal agents will most likely be cloud-based. AWS, Azure, and Google Cloud already provide computing infrastructure and have cloud desktop products. Adding an agent orchestration layer on top seems like a very natural product development logic. Cloud providers have additional advantages in this business because their long-standing infrastructure can support a complete set of capabilities around agents, similar to observability—logs, permissions, circuit breakers, and so on. Overall, I think this kind of business will be one of the future narratives for cloud vendors.