# Emulated: The Data for Fully Autonomous Software Engineers and Companies — Joseph Wang

## Executive summary

Emulated focuses on creating high-fidelity training data environments that simulate entire companies and complex infrastructure operations, moving beyond simple code diffs or single-node sandboxes. The core argument is that for AI agents to achieve true autonomy in mission-critical systems (like cloud providers), they must be trained on long-horizon tasks involving distributed cluster failures, resource provisioning across VPCs/subnets, managing cost constraints, and reasoning through real-world operational incidents.

## Key takeaways

- The Data Gap in AI Agents: Current benchmarks (e.g., SweBench Pro, Terminal Bench) limit agents to operating within a codebase, failing to capture the complexity of real-world tasks like PM communication, performance testing, or owning underlying infrastructure over years.
- Complexity Requires Full Simulation: Real infrastructure work is not a simple code diff; it involves managing failing nodes, stale deprecated components, live traffic serving, and operational blast radius across distributed clusters.
- Limitations of Single-Node Sandboxes: Standard post-training pipelines often use homogeneous single-node sandboxes. However, real cloud services require simulating resource provisioning (EC2, Cloud Run), VPCs, subnets, and security groups, which necessitates a multi-node sandbox with access to real infrastructure.

## Technical details

- Distributed Systems Simulation: Emulated simulates distributed clusters in single sandboxes, allowing agents to deal with operational issues like network failures between nodes, data corruption (e.g., MVCC issues), and clock skew.
- Cloud Infrastructure Provisioning: Simulating cloud services requires modeling resource provisioning beyond the code base, including VPCs, subnets, security groups, front-end APIs (for throttling/auth), and deployment components for managing versions and rollbacks.
- Operational Constraints: Agents must reason about real-world constraints such as cost management, gradual deployments to limit blast radius, and maintaining service availability while running live traffic through complex migrations (e.g., migrating off old hardware).

## Practical implications

- Build engineers must consider simulating the full lifecycle of a service—from initial resource provisioning (VPCs, subnets) to deployment rollbacks and failure recovery—rather than just testing code changes.
- Testing environments need to evolve from single-container sandboxes to multi-node simulations that model real cloud infrastructure dependencies and failures at scale.
- The focus for AI agent training should shift toward long-horizon tasks involving operational incident response, resource management, and cross-domain reasoning.

## Topics

AI Agents, Distributed Systems, Cloud Infrastructure, Software Engineering, Data Simulation, SweBench Pro, Terminal Bench

Source: https://www.youtube.com/watch?v=zkX03APVj0M