Ethereum co-founder Vitalik Buterin stated on October 6 at the OKX NOW Global Product and Ecosystem Conference in Singapore that Ethereum transaction costs will further decrease in the next two years, privacy capabilities will improve, and the integration of artificial intelligence with on-chain applications will accelerate. He believes that blockchain is far from mature, and artificial intelligence represents both the greatest opportunity and the greatest risk.

He suggested changing the interface from applications to agents. Users will not need to open traditional applications or websites; they only need to specify the desired outcome, and the agent will write scripts, execute on-chain operations, and explain transactions. About a month ago, he updated Ethereum Name Service records using a locally running agent without going through the traditional interface, completing the entire process in about five minutes.

The scenario two years from now is described by him as a three-step process. Today's trading counterpart may already be a robot, while applications are still written by humans. In two years, applications themselves may also be written by robots. Further down the line, the concept of "applications" as understood today may not exist, with different robots continuously posting quotes on-chain, and users' robots reading the quotes, interpreting their meanings, and deciding which ones to accept.

Apart from payments, he pointed out existing uses: helping users pay for artificial intelligence reasoning while enhancing the privacy of these payments. The next step is for artificial intelligence to become an on-chain participant itself, rather than just a service being paid for. He estimates that changes in tools occur approximately every three months, sometimes even faster.

Risks are listed separately. As agents take on more operations, Ethereum's original pseudonymity may be weakened. Another risk is prompt injection: malicious instructions mixed into the data read by the agent, enticing it to perform actions not requested by the user. Cheapness and privacy are stated as prerequisites, with agents described as interfaces, while security is not claimed to be resolved.

The buyer is the user who hands over intentions to the agent, while the seller is the robot posting quotes on-chain. Events are driven by the judgment of interface migration, not by the launch of new protocols. Funds shift from in-application click paths to transactions filtered by agents. The beneficiaries are the agent layer that can read quotes and execute transactions, while traditional applications that charge fees based on front-end pages are under pressure.

The five-minute name service update only proves that a single operation can bypass the interface. The inference that applications may disappear in two years is not based on already effective network rules.

Source: Public Information