The latest roundup of Artificial Intelligence assisted coding work focuses on fresh experiments with Laravel projects and evolving model workflows. A new version of Laravel evaluations for Artificial Intelligence coding tools puts 6 large language models through 6 different Laravel projects and challenges in a 13-minute video, producing one clear winner that, unexpectedly, is not the Opus model. Additional tips include a short format guide on what to do when writing a long prompt and forgetting to switch the large language model, along with a prompt pattern to automate some of the rules typically stored in CLAUDE.md so that not everything must live in that configuration file.
Project-level workflow changes are explored through a test where CLAUDE.md or AGENTS.md files are shortened or deleted entirely across 5 different Laravel projects, examining how these adjustments affect agent behavior. Another video walks through using Gemini 3.1 Pro inside Google Antigravity on several Laravel projects, showing how the new model behaves in real coding scenarios. There is also a short commentary on an emerging pattern in which website owners will or should serve Artificial Intelligence agents differently from regular users, treating agents as a distinct class of clients in order to work with them more efficiently.
Community updates center on the rapidly evolving Claude Code ecosystem and neighboring tools. One long article emphasizes that prompt caching is everything when building Claude Code based systems. Claude Code is rolling out a Remote Control feature to Max users in research preview that allows starting local sessions from the terminal and continuing them from a phone. Developers can now run Claude Code, Codex, and Gemini side by side in separate git worktrees, and a new guide to agentic engineering patterns outlines practices to get better results from coding agents such as Claude Code and OpenAI Codex. The ecosystem also includes a VS Code extension that turns Claude Code agents into animated pixel art characters, Codex 0.102 with experimental multi-agent support in a TUI under /experimental, and a Claude Code hackathon where 500 builders explored projects with Opus 4.6 and Claude Code. Security capabilities expand with Claude Code Security in limited research preview to scan codebases for vulnerabilities and suggest targeted patches, while Anthropic has sent a legal notice to OpenCodex.com that explicitly prohibits external tools from using or orchestrating Claude Code. New tooling moves Artificial Intelligence coding session context between Claude Code, Gemini, Copilot, Codex, and OpenCode on rate limits, and performance updates from Jarred Sumner highlight that Claude Code’s p99 memory usage dropped by 40x in the last two weeks and by 6x since January, even as the product grows and ships new features.
