{
  "id": 20020489,
  "title": "AWS Emulator proves agents need a verification loop. Realism is the hard part.",
  "url": "https://urgent.news/2026/11/14/aws-emulator-proves-agents-need-a-verification-loop-realism-is-the",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-14T17:30:00.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/aws-emulator-verification-loops/"
  },
  "original_language": "en",
  "account": "Amazon Web Services has introduced AWS Emulator, a local replication of its AWS services, designed for coding agents to test against. AWS Emulator covers over 200 AWS services and was developed entirely by agents. The key takeaway from the release is the importance of a verification loop, as explained by Swami Sivasubramanian, a key contributor to AWS systems such as S3 and DynamoDB.\n\nIn the past, building a replication engine for Alf, a Paxos-based lock service, took about four months. However, when AWS's agentic coding tool, Kiro, was asked to rebuild it, it took only two hours due to the introduction of a feedback loop that allowed an agent to check its own results. This highlights the critical role of a verification loop in ensuring the effectiveness of coding agents.\n\nA coding agent functions in a loop: it writes code, runs it, reads the result, and decides what changes to make next. However, in distributed systems, this loop runs inside the infrastructure. If the services the code calls do not behave as real ones, the agent learns incorrect lessons and continues making the same mistakes. AWS Emulator addresses this issue by integrating AWS into the loop on a developer's laptop. By pointing existing AWS tools and code at AWS Emulator with a single environment variable, tests can run against emulated services without any modifications.\n\nAWS Emulator was written by autonomous agents using AWS's public API models. Each behavior is validated by a fidelity test that initially passes against real AWS and then against the emulator. Any changes undergo a merge gate where security agents review them. The key message is that when agents write the code, the engineer's responsibility is to create the specs, tests, and merge gates that determine whether the work is correct.\n\nMany reactions on Hacker News focused on trust and questioned the quality of AWS Emulator due to the high number of commits in a short period. However, this overlooks the crucial factor of verification. What matters more is the verification against which the code was validated, and AWS's answer is real AWS.",
  "summary": "This week, Amazon Web Services released AWS Emulator, a local copy of AWS that coding agents can test against. It The post AWS Emulator proves agents need a verification loop. Realism is the hard part. appeared first on The New Stack .",
  "key_points": [
    "AWS Emulator replicates over 200 AWS services for agent testing.",
    "Verification loop enables agents to check their own results, reducing mistakes.",
    "AWS Emulator validates behaviors against real AWS before integration."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}