Product build for a purpose.
I believe the UCP design is completely flawed.
//This article wasn’t written by AI; only the images used for comprehension were AI-generated.
In 2025, I participated in numerous e-commerce-related projects, including cross-border e-commerce.
Then I saw Google release the UCP protocol and decided to try it out.
First, UCP shouldn’t require a traditional e-commerce platform as a mandatory entry point for authentication or trust verification.
The UCP design is very similar to a traditional e-commerce platform, but such platforms are no longer necessary. Consumers searching for products and constantly swiping down the screen is an inefficient way to select products and interact with the system.
Now, searches don’t even need keyword or fuzzy queries. Why not directly perform multi-dimensional vector matching? This would be faster and more accurate in AI’s memory, and it could discover and match implicit factors.
AI has reached 2026, and many functions are already achievable. I don’t really understand why UCP is designed with a standard product description + image display—such a traditional presentation method. In the AI era, products can be physical goods, services, industry knowledge, or industry guidance, etc. Before AI matures, the capabilities of sellers outside of traditional businesses are difficult to categorize and assess, relying primarily on certificates, degrees, and so on. Now, these skills and knowledge can be learned by an individual’s AI agent, much like how HR professionals in ancient times used resumes to learn about a candidate’s background.
In the future, skill providers (also known as job seekers) won’t need to condense all their experiences and projects into a single A4 page just for HR convenience. Instead, job seekers can compile all their experiences, materials, project documents, details, etc., into a chatbot for HR or buyers to inquire about and negotiate. HR will also find it easier, directly requesting relevant project information and conducting offline capability assessments. HR can even communicate their needs to an intermediary agent, allowing for screening within A2A or UCP networks.
In the future, merchants won’t need to categorize, package, describe, and upload images of all their products like on Amazon/eBay. Instead, they can simply create a knowledge base containing all their products, photos, manuals, shipping and return policies, etc. Enabling AI to communicate and summarize with each other greatly improves purchasing efficiency and the purchasing experience.
Do these technology stacks sound familiar? Yes, AI can perfectly replace commercial search.
Google has a very important software called NotebookLM, but many of its functions are not being fully utilized. NotebookLM is a great product, currently the most perfect knowledge base product. If businesses could build their own external product knowledge bases, and job seekers could build their own resumes, skills, and project experience knowledge bases, then the communication efficiency of the entire internet would be greatly improved.
Many needs that couldn’t be met previously through the search boxes on Amazon, LinkedIn, eBay, etc., including follow-up questions on candidates and simultaneously filtering dozens of candidates, can now be accomplished very easily.
//For example. If I’m a pet owner, I need to buy pet supplies, food, consumables, and occasional boarding, etc. Then I would need to open eBay to see what supplies are available and what prices they are, then open Amazon to see if my favorite brands are on sale, and then open various price comparison and group-buying websites to find the latest coupons, etc. Then, you pick up the calculator on your phone, do the calculations, figure out how to combine different packages, and pay attention to the weighted average of postage and taxes. This is a very long decision-making chain, influenced by many factors. Why not directly use search + AI to help users complete these tasks, automatically providing the optimal solution for several package combinations, and explaining the performance or other differences brought by different prices, offering one-stop guidance for purchasing and ordering? Users don’t actually need traditional websites like Amazon or eBay; these websites are just sources of information. Abandon GEO, SEO, and other optimization strategies; let these algorithms and platforms become history.
// Imagine I’m the head of a small electronics design outsourcing team. Designing a dedicated programming keyboard, then procuring the electronic components for this sample wouldn’t be suitable using ordinary electronics procurement platforms. There’s a lot of background knowledge in electronics to consider as constraints and selection criteria. For example, whether to add memory, and which memory chip to add. Many types can be used in the design, but memory market prices fluctuate greatly. Many components have many alternative categories and non-parallel alternatives from other brands, such as core MCUs and SoCs. (Thanks to the development of AI coding, migrating programs to other platforms is now very smooth.) With numerous constraints and open conditions, many potential replacement models, and considerations such as availability, shipping times, platform discounts, and combinations, user choices increase exponentially. This is where AI becomes crucial. AI can help you sift through massive amounts of information, selecting the optimal combination based on your project needs, and identifying potential, cheaper alternatives. (Future quantum computing will also be very important, solving optimal and suboptimal solutions to long-constrained equations, etc.)
Most of the technology stack for these systems is already available from Google, requiring minimal development.
Some backend processes require more sophisticated design, such as agent-to-agent authentication, credit rating exchange, and authorization.
Current CLI tools’ skill designs are not ideal, leading to a proliferation of individual and company developers designing skill tools, but making it difficult for application users to assess actual performance. Future solutions should include an “experience” module, corresponding to the practical experience of each design process developer and lessons learned from repeated trials.
Google currently possesses the strongest toolchain, ecosystem, and hardware. But why wasn’t it done?
NOW, it’s the time.