Launch of Kiro, which can autonomously program for several days after learning from humans; Amazon Cloud Vice President: AI Agents will be the biggest technological transformation "since the birth of cloud computing."

Amazon is pushing the AI arms race into a new phase, with the recent launch of its “frontier AI agents” not just a technical showcase but positioned by company executives as the next major technological transformation since the advent of cloud computing. At this event, the AI agent Kiro, capable of autonomous programming for several days, took center stage. At the AWS re:Invent conference held on December 2, Amazon’s cloud computing division officially previewed three brand-new AI agents. AWS CEO Matt Garman and Swami Sivasubramanian, Vice President in charge of agent-oriented AI, emphasized that these tools are designed to go beyond the traditional chatbot Q&A format and instead represent users to proactively “get work done.” The most notable among them is Kiro, an autonomous agent specifically designed for software development. Reportedly, **after learning from human developers, Kiro can work independently for hours or even days, handling complex programming tasks. This ability is attributed to its “persistent contextual memory,” which prevents it from “forgetting” instructions during long-term tasks.** This move directly intensifies Amazon’s competition with technology giants like Microsoft, Google, and OpenAI, who are also actively pursuing AI agent technology. It marks a more aggressive strategy from Amazon to consolidate its leadership in the AI-driven cloud services market. Amazon has already demonstrated its potential using internal data, claiming that its self-developed AI agents have saved $250 million in capital expenditures. **Kiro: An “AI Colleague” Capable of Autonomous Programming for Days** Kiro’s design goal is to become an “AI colleague” for development teams. According to TechCrunch, Kiro is based on the concept of “spec-driven development.” It learns by observing how human developers give instructions, confirm, or correct its assumptions, and by scanning the existing codebase to understand the team’s workflows and standards. “All you need to do is assign a complex task from your to-do list, and it can independently figure out how to complete the work,” AWS CEO Matt Garman said in his keynote speech. He added, “It can genuinely learn your preferred way of working and, over time, deepen its understanding of your code, products, and team standards.” Amazon provided an example where developers can instruct Kiro to update a critical code snippet used by 15 different company software products, without having to assign and verify each update separately. Reportedly, Kiro can maintain “persistent context across sessions,” meaning it doesn’t lose memory during long-term tasks, allowing it to work autonomously for several days with minimal human intervention. In addition to Kiro, Amazon has launched the AWS Security Agent and AWS DevOps Agent. The former can independently identify security issues during coding and provide fixes, while the latter is responsible for automatically testing the performance and compatibility of new code. Together, these tools form a suite aimed at automating software development processes. **Transformation Comparable to Cloud Computing, Saving Hundreds of Millions Internally** Amazon executives have elevated the emergence of AI agents to a strategic height comparable to the birth of cloud computing. AWS Vice President Swami Sivasubramanian has explicitly stated: **“The era of AI agents can be said to be one of the greatest technological transformations since the advent of cloud computing.”** He **believes AI agents will shift from “merely answering questions in chatbots” to “actually doing things for us.”** This confidence comes from successful internal practices at Amazon. According to Yahoo Finance, Amazon has saved 4,500 developer years and $250 million in capital expenditure by deploying AI agents to upgrade its internal software development. Sivasubramanian also shared a specific case: AWS’s Bedrock team used AI agents to rebuild its inference platform in “just under a few months” with a team of “six to eight people.” He stated that such work “might have taken a year” in the past. These figures provide early but strong evidence of the tremendous potential of AI agent technology in boosting efficiency and reducing costs, and also demonstrate its broad commercial prospects to the market. **The AI Agent Arms Race Heats Up** Amazon’s high-profile entrance marks the rapidly intensifying competition in the AI agent field. Major competitors such as Google, Microsoft, and OpenAI have already devoted significant resources to this area, aiming to turn AI chatbots into utility tools that can perform tasks for users with minimal intervention. For example, **OpenAI announced last month that its agent-based coding model GPT-5.1-Codex-Max is also designed for prolonged operation, up to 24 hours.** Despite promising prospects, widespread adoption of AI agents still faces challenges. Some argue that current large language models (LLMs) suffer from “hallucinations” and accuracy issues, which could force developers into a “babysitter” role, constantly supervising and verifying AI-generated results. Consequently, many developers still prefer to assign short, easily verifiable tasks. Nevertheless, expanding AI’s context window—enabling it to handle longer and more complex tasks—is widely regarded as a key step toward making AI a true “colleague.” Amazon’s newly released technologies are undoubtedly a major step in this direction, further intensifying the race around next-generation AI productivity tools. Risk Warning and Disclaimer The market has risks, and investment needs to be cautious. This article does not constitute individual investment advice, nor does it take into account the specific investment objectives, financial situations, or needs of particular users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment made according to this is at your own risk.