AI News June 3 2026 - OpenAI Lands on AWS, Build Day 2 Delivers, and xAI Enters the Coding Agent Race

1. OpenAI GPT-5.5 and Codex Are Now Live on Amazon Bedrock

OpenAI's GPT-5.5, GPT-5.4, and Codex reached general availability on Amazon Bedrock on June 1-2, 2026, following the April 28 limited preview that launched when OpenAI ended its exclusive cloud arrangement with Microsoft. Pricing matches OpenAI's first-party rates exactly - no additional AWS fees - and usage counts toward existing AWS committed spend (EDP agreements). For large AWS customers with multi-million dollar annual commitments, this means OpenAI API consumption can now be absorbed within those commitments rather than paid separately.

GPT-5.5 is available in US East (Ohio) only for now; GPT-5.4 is in both US East (Ohio) and US West (Oregon). Both run via the Responses API on Bedrock's next-generation inference engine with full AWS security controls - IAM, PrivateLink, CloudTrail, encryption. Codex is accessible via CLI, desktop app, and VS Code/JetBrains/Xcode IDE integrations with all inference routed through Bedrock. Amazon Bedrock Managed Agents (powered by OpenAI) and Daybreak on Bedrock are confirmed as the next phase of the partnership.

Full article: OpenAI GPT-5.5 and Codex on AWS Bedrock - pricing, regions, AWS commitment credits, and what comes next ->

2. Microsoft Build Day 2 - Copilot Agents That Debug Your Code and Surface Laptop Ultra

Build Day 2 on June 3 delivered the engineering depth behind the Day 1 platform announcements. The highlight was GitHub Copilot agents for Visual Studio that go beyond code generation: a debugging agent that performs root-cause analysis when a breakpoint is hit (demonstrated finding a race condition in async initialization), a profiler agent that reads Visual Studio Performance Profiler output and generates targeted refactoring suggestions for actual hot paths, and a test generation agent that creates unit tests from live executed code paths rather than static analysis.

The GitHub, Copilot, VS Code, and More live session (LIVE104) showed multi-agent support in VS Code - multiple specialized agents working simultaneously within the editor - and confirmed Windows Agent Framework agents can be invoked directly from the Copilot Chat interface in VS Code. Microsoft also unveiled the Surface Laptop Ultra: a 15-inch device with NVIDIA's RTX Spark system-on-chip that integrates CPU, GPU, and NPU with full CUDA support - designed to collapse the "Mac for portability, NVIDIA desktop for ML" two-device setup into one Windows machine. Phi-4 Mini Reasoning for on-device AI on Copilot+ PC NPUs was also confirmed at Build.

Full article: Build Day 2 - Visual Studio Copilot agents, Surface Laptop Ultra, and Phi-4 Mini Reasoning ->

3. xAI Launches Grok Build Beta - A Third Coding Agent Enters the Race

xAI launched Grok Build in beta - a terminal TUI coding agent running on Grok 4.3 with headless scripting support for CI/CD pipelines and native Agent Client Protocol (ACP) integration for multi-agent orchestration. It is available now to SuperGrok ($30/month) and X Premium+ subscribers. The design is closer to Claude Code (terminal-native) than Codex (desktop app and cloud-first), and it enters the market as beta software against two GA competitors.

Grok Build's genuine differentiator is real-time X/Twitter data access - not web search that surfaces X posts occasionally, but native access to the platform's full data stream. For coding tasks where community signals matter in real time, this is a meaningful advantage. The gaps are also real: beta status, no desktop computer use, a smaller tool integration catalog than Claude Code's 3,000+ MCP servers, and a 256K context window vs Claude Code's 1M beta. xAI is reportedly using only 11% of available Colossus compute for Grok models - the model improvements from a compute ramp are coming, but today's Grok Build is early-stage relative to its competitors.

Full article: Grok Build beta vs Codex vs Claude Code - feature comparison, ACP explained, and the honest gaps ->