AI/Tech Brief — 2026-06-02

TL;DR

Claude Code Update: Version 2.1.160 introduces enhanced security prompts for sensitive file writes and renames the dynamic-workflow trigger to ‘ultracode’.
OpenAI on AWS Bedrock: OpenAI models, including Codex, are now available via an OpenAI-compatible Responses API endpoint on Amazon Bedrock.
Stanford AI Guidelines: Stanford’s CS336 course materials for building autonomous AI agents have gained significant traction, reflecting a high demand for deep, foundational AI knowledge.

Claude Code 2.1.160 Released: Adds security prompts before writing to shell startup or build-tool configs, fixes WSL clipboard issues, and renames ‘workflow’ to ‘ultracode’. This prevents unintended command execution and improves reliability. Source
OpenAI Models Arrive on AWS Bedrock: Codex and other frontier OpenAI models can now be accessed via Amazon Bedrock using an OpenAI-compatible endpoint. This gives developers a local execution path with AWS-managed billing and authentication. Source
Codex App Gains Windows Support: The Codex desktop app’s “Computer Use” feature now works on Windows, allowing it to directly operate desktop applications. This majorly expands the reach of local AI coordination. Source
Stanford CS336 AI Agent Guidelines: Educational resources and best practices for building autonomous agents from a Stanford computer science course have been released openly. This shows a strong community push toward foundational AI infrastructure over wrapper applications. Source
Muxcard - DIY Credit Card Computer: A fully functional, credit-card-sized computer built with off-the-shelf components. Highlights a continued hardware-hacker interest in highly compact edge devices. Source

The Janet Programming Language: Ongoing discussions highlight Janet, a modern Lisp-like language optimized to be small, fast, and easily embeddable for specialized computing tasks. Source

Superhuman AI Job Loss Claims: Various op-eds claiming economists see “zero evidence” of AI job loss are circulating, but these offer no new technical insights or data and can safely be ignored.
Groq and OpenAI Lawsuits: Broader funding rounds and state-level lawsuits (e.g. Florida vs OpenAI) are making headlines but do not impact day-to-day development work today.